Hi Team!

I get a morning brief from Instinct every day and another EOD closer. It might be my favorite super-simple AI use case since I’ve gone back to simple (insert bell curve meme).

With BFCM coming, I keep thinking about the work that never makes the planning deck: chasing final creative, checking whether a discount code applies, and figuring out what the 3PL promised last Tuesday.

There's a surprising amount of admin work that needs to get done.

So here are ten jobs I'd hand Instinct, with the exact prompts I'd use and what I'd connect. The daily brief is something I already use. The others are jobs I'd test, with a human checking anything that affects customers or checkout.

Let's get into it.

This week’s newsie is brought to you by my friends at Horatio!

Is your holiday chatbot stuck in the “Bot Trap”?

A lot of ecommerce brands rolled out shiny new AI bots for the holidays, and with BFCM two weeks out, the cracks are already starting to show.

Here’s how it goes: a shopper asks about a bundle that’s missing an item, or an order that showed up in two boxes instead of one. The bot doesn’t get it, loops them back to the same three answers, and by the time a real person steps in, a simple fix has turned into a customer who’s DONE with you (and probably telling their friends about it).

The answer isn’t ripping out the bot. It’s giving it a safety net: real humans ready to catch the messy, high-stakes tickets before the bot fumbles them.

I teamed up with Horatio to put together a free playbook on exactly that, covering which tickets to keep away from your bot, how to hand off to a human without making customers start from scratch, and how to keep CSAT intact while volume spikes.

Grab it now while there’s still time to put it to work before the rush.

Get the AI Safety Net Playbook

If you’d like to try this yourself, here’s an invite to Instinct if you need one.

1. Start with what broke last year

I'd start with last year's fires. No point making another immaculate BFCM planning deck while the same broken discount code is waiting to happen again.

Text your Instinct: "Scan my email and Slack from November and December 2025 for BFCM problems: broken codes, stockouts, shipping delays and escalations. Group repeats, rank the ten biggest issues, and link the original threads."

Connect first: email and Slack.

2. Get a 7 a.m. brief before the inbox catches fire

From the week before Black Friday through Cyber Monday, I'd want the brief I already use, tuned to the small issues about to become expensive.

Text your Instinct: "From November 20 through Cyber Monday, send me a 7 a.m. brief from my email, Slack and calendar: overnight escalations, repeated complaints, blockers and today's deadlines. Link the source for anything urgent."

Connect first: email, Slack and calendar.

3. Keep an eye on your competitors

Pick five competitors. I want the offer changes and stacking codes, ideally before the group chat starts panicking.

Text your Instinct: "Check [five competitor URLs] every day through Cyber Monday. Report only changes to discounts, bundles, shipping thresholds or stacking codes, with links. If nothing changed, don't message me."

Connect first: Just your email for promo emails.

4. Keep track of who owes you what

An agency says the assets are coming Friday. The 3PL is checking on the cutoff. Nobody seems entirely sure who owns the follow-up. I'd want the outstanding list before one of those promises becomes my problem.

Text your Instinct: "Scan my sent emails from the last 60 days. List outstanding vendor and agency deliverables, with owner, due date and a link to the thread. Flag the three most urgent each morning."

Connect first: email. Add Slack if that's where the promises were made.

5. Write the support macros while everyone is still sane

You know what's coming: the code doesn't work, where's my order, can I change the address, will it arrive by the 24th, and how do I return a gift.

Text your Instinct: "Using last year's tickets and our current policies, draft replies and help-center answers for our five most common BFCM questions. Flag policy gaps. Draft only, don't send anything."

Connect first: A helpdesk export, or email if tickets live there, plus your current policies.

6. Check your own site like a customer

I'd have it test the actual sale path in a fresh browser session. The promotional email is no use if the discount doesn't apply to the product it links to.

Text your Instinct: "Open [PDP], add [SKU] to cart, apply [CODE], and check the discount and shipping promise. Screenshot anything wrong. Stop before payment and flag anything you couldn't test."

Connect first: Store URL, promo codes and expected shipping terms.

7. Catch the promises that contradict each other

The campaign says delivery by the 24th. The FAQ says the 19th. The support macro says three to five business days. Good luck to whoever owns the inbox.

Text your Instinct: "Compare [campaign drafts], [site URLs], [FAQ] and [support macros] for contradictory discounts, shipping dates, returns and subscription terms. Quote both versions and link the sources. Don't edit anything."

Connect first: Campaign drafts, site URLs, FAQs and macro exports.

8. Make someone read the carrier fine print

Holiday shipping dates, carrier exceptions, extended return windows. All the stuff everyone agrees matters and nobody wants to spend Tuesday afternoon reading.

Text your Instinct: "Find the published 2026 holiday cutoff dates for the carrier services we use, and compare the return windows of [three competitors]. Give me a one-page summary with source links, publication dates and exceptions."

Connect first: Public web. Give it your carriers, services and shipping regions.

9. Find out what customers keep asking

Your reviews, surveys and inbound tickets are full of unanswered buying questions. I'd rather find them before spending the BFCM acquisition budget.

Text your Instinct: "Review the attached survey and review exports plus 90 days of customer emails. Rank the ten most common pre-purchase questions, show how often each came up, and tell me which answers are missing from [PDP URLs]."

Connect first: Customer email, review and survey exports, and PDP URLs.

10. Book the January reality check

Book a first read for December 2, then 60-day and 90-day checks. I'd want to know who bought again, what discounts it took, and how much contribution margin was left.

Text your Instinct: "Put a BFCM review on my calendar for December 2, then 60-day and 90-day cohort reviews. In each invite, list second-order rate, contribution margin, refunds and repeat discount use. Compare first-time BFCM buyers with a relevant earlier cohort."

Connect first: Calendar. You'll still need the actual data.

The common thread is the boring work between teams: missing files, conflicting promises, questions nobody answered and follow-ups everyone assumed someone else owned. It's remarkable how much of the holiday rush is spent looking for information that already exists somewhere.

I'd still rather see a short list of things that look wrong than spend Thanksgiving week digging through forty-seven Slack threads trying to find them.

That's it for this week!

Any topics you'd like to see me cover in the future?

Just shoot me a DM or an email!

Cheers,

Eli 💛

Hi Team!

This newsie comes to you from UA85, a 12-hour flight from TLV to EWR after three days in Israel for a wedding. I managed to eat my weight in hummus and rugelach while I was there, so by my standards it was a pretty successful trip.

Black Friday is a few weeks away, and I suspect this might be the hardest BFCM in years.

The spending forecasts actually look pretty healthy. The National Retail Federation expects holiday retail sales to grow 4.2% this year. There's money to be made, and plenty of brands are preparing to spend a fortune convincing customers to give them some of it.

And in our brave new world, AI can (sometimes) produce beautiful ads, completely re-engineer a PDP, etc.

That being said, competitors have the same tools, and customers are noticing how alike everything sounds. From the tone to the visuals, to the same promises being made everywhere.

A Gartner survey recently found that 65% of consumers believe brands are producing too much AI-generated content in their marketing materials, and 57% say this content has made them less trusting of brand messaging overall.

Shoppers will compare dozens of offers and reviews, trying to work out who deserves their money. The signals they once relied on are easier than ever to manufacture.

I think this BFCM will expose the difference between brands that look good and brands that deliver.

There are three places I'd be paying particularly close attention to over the next few weeks.

Let's get into it.

This week’s newsie is brought to you by my friends at Horatio!

If you’ve read the last few issues, you know Horatio has been sharing a lot to help e-commerce brands get through BFCM in one piece: fewer backlogs, faster returns, and fewer chargebacks slipping through the cracks.

But getting through peak season is only half the job. The real payoff is when the customers you win in November are still buying from you in March.

That gets decided when wrong-size gifts start coming back and first-time buyers notice a subscription charge they forgot they signed up for. After weeks of nonstop volume, the instinct is to clear the queue as fast as possible: approve the refund, send the label, close the ticket.

It feels efficient, but every auto-refund that could have been an exchange is revenue walking out the door, usually along with a customer you won’t see again.

The brands that protect their long-term margins treat their queue as a retention opportunity, not a cleanup job. And they set it up before the rush, not in the middle of it.

That’s where the Post-Holiday Retention Guidebook by Horatio comes in. It shows you how to sort post-holiday tickets by root cause, turn returns into exchanges, get ahead of WISMO, and give first-time buyers a real reason to come back.

Grab it now, while there’s still time to put it to work. 👇

Get the Post-Holiday Retention Guidebook

Your best-looking ad has a lot of competition

A growing number of ads in my feed are starting to look interchangeable. Immaculate lighting, immaculate people, and a script you could use for a skincare serum or a water bottle if you changed three nouns.

Now add BFCM, with every brand pushing its biggest sale and customers scrolling through dozens of near-identical promises.

I still stop for the customer unboxing something while a toddler screams in the background, or the founder walking around the warehouse with an iPhone explaining why the product had to change.

A shaky camera is easy to imitate. I care more about what the creator knows. How do the jeans fit after a wash? What's hard to clean? What would they change after a month?

Before BFCM, I'd ask a few creators to talk about the product the way they'd explain it to a friend who was about to spend their own money. Show the setup step that trips people up, the sizing issue, or the part that impressed them.

Then test those videos against your polished creative with the same offer and comparable audiences. Look at conversion and contribution margin alongside attention. There is no prize for making a scrappy ad that gets watched and sells nothing.

To sum it up: authenticity will be the thing that makes you stand out.

Your reviews have a harder job this year

A 5.0 average with four hundred reviews that all use the same three adjectives looks a little sus in 2026.

Sometimes that's generated copy. Sometimes it's just the result of asking everyone "How was it?" and collecting four hundred variations of "Amazing!"

During BFCM, that matters even more. Someone comparing three brands with similar products, similar discounts, and similar shipping offers needs some way to decide which one to trust.

The reviews (and AI review summary) I read tell me the parts the PDP forgot.

"Runs half a size small, order up."

"Took three weeks before I noticed any difference."

A photo of the pan on someone's mismatched stove, with a note about whether the handle gets hot.

A three-star review with a thoughtful brand response can be especially useful when you can see what happened next.

I'd even consider adding a custom question to the post-purchase review request to "What's one thing you wish you'd known before ordering?"

That should beat another wall of "Love it." Make those answers easy to find on PDPs, especially for products you're pushing during BFCM.

And please leave the critical reviews up. Respond with specifics and fix the recurring issues they expose. If fifteen people mention the same sizing problem, updating the size guide is probably worth more than writing the sixteenth polite reply.

The apology is beautiful

A customer reports a missing refund and gets a beautifully written apology in seconds, complete with the assurance that the team is taking it seriously. Nine days and three follow-ups later, they still haven't received their money.

The copy was great, though.

I recently wrote about the rainmaker role and the importance of giving someone enough authority to handle complicated customer moments. What concerns me heading into BFCM is how easy it's become to give customers the feeling of ownership without anyone actually owning the problem.

"I've escalated this to our team" sounds reassuring. Did anybody get the case? Who follows up, or does the customer have to ask again on Thursday?

Now multiply that by holiday order volume, delivery deadlines, and gifts that have to arrive on time.

A missing replacement in July is frustrating. A missing replacement three days before Christmas has a slightly different level of urgency.

I suspect a lot of brands are heading into BFCM with impressive automation metrics and some fairly questionable resolution processes underneath them.

Before the holiday rush, I'd pull twenty escalations that were marked resolved and follow them all the way through. Did the refund actually land? Did the replacement ship? Was the customer told when the promise changed?

I'd be more interested in those answers than an inbox full of perfectly phrased replies.

What I'd be checking before Black Friday

I'd check the creative you are about to spend money on for useful buying information, then look at whether your PDPs and Q&A’s and reviews answer the questions customers repeatedly ask.

Finally, audit the escalations you mark resolved. A closed ticket means very little to someone still waiting on an order.

None of these require a massive project before BFCM. They're worth addressing while there's still room to fix something.

I'm happy for all of us to keep using AI. But customers are getting better at questioning what they see: ads are easy to generate, glowing reviews are easy to write, and apologies arrive instantly.

The product still has to work, the reviews have to help, and the company has to follow through.

BFCM might break sales records. I'd be more interested in which brands keep those customers happy once the offers expire and the orders start arriving.

That's it for this week!

Any topics you'd like to see me cover in the future?

Just shoot me a DM or an email!

Cheers,

Eli 💛

Hi Team!

It’s the last full week of October, and the inboxes can feel it. Every brand you’ve ever bought one thing from is warming up the same email.

Early access, VIP preview, “our biggest sale of the year,” in all caps, somehow bigger than last year’s biggest sale of the year.

Don’t get me wrong, I love a deal. But the same brands that spend eleven months talking about loyalty are about to spend four weeks teaching their best customers to wait for the next code. I don’t think the math on that trade ever gets done.

Let’s get into it.

This week’s newsie is brought to you by my friends at Horatio!

Nothing beats watching the numbers climb on Black Friday. And nothing undoes that feeling faster than opening the support inbox the following Monday.

What rarely makes it into the BFCM planning deck is that your team’s busiest stretch comes after the sale. Packages stall in transit, and anxious customers reach out on email, chat, and Instagram, often about the same order.

Then the returns, gift swaps, and size exchanges start arriving. And right after your seasonal help wraps up, January brings a pile of chargebacks.

Each of these chips away at the revenue you worked hard for. Left unmanaged, the post-peak surge can quietly erase up to 30% of the margin you gained during BFCM.

The tricky part is that most brands don’t see where they’re vulnerable until they’re already underwater. Once response times have stretched from hours to days and your best agents are exhausted, it’s too late to fix the process. You’re just trying to get through the week.

The brands that come out of Q4 with their margins (and their team) intact decided ahead of time what to keep in-house and what to hand off.

That’s why Horatio created The Q4 Survival Guide for E-Commerce Brands.

It covers the three post-peak profit leaks: WISMO, returns, and chargebacks. It also includes a 12-question stress test that shows where your operation is most likely to crack and what to do about it, all without hiring a single new employee.

Take the stress test now, while there’s still time to act on your score. 👇

👉 Get the Q4 Survival Guide

The training program nobody admits to running

Every brand says they want loyal customers. Then they run 40% off every six weeks and act surprised when nobody buys at full price.

What they’ve actually built is a customer who likes the product and has figured out, correctly, that the sale comes on a schedule.

The cycle usually goes the same way. Revenue dips, someone panics, a bigger code goes out, and everyone celebrates when the graph recovers for a week.

Meanwhile, customers who would have bought anyway are learning that full price is for people who forgot to check their inbox.

The coupon tax

Here’s the math we often don’t discuss:

Take a $100 order with a 40% gross margin, so $60 of product cost. Full price, that’s $40 of gross profit. Run 25% off and the same order, same cost, same mix, makes $15.

To book the same gross profit dollars, you now need to sell almost three times as many units. If the code doesn’t get you there, the bigger discount just made you poorer.

And a customer who only shows up at 40% off is telling you what they came for.

When the discount is the right answer

None of this means that there’s no use case for discounting. Clearing genuinely dead inventory is a legitimate reason to run one, and a first-order discount can make sense when the acquisition economics work.

Just remember that the first price someone pays becomes part of how they value the product. And when the same person gets the welcome offer for the fourth time, acquisition economics have become retention economics.

A real thank you that happens once, an anniversary or a milestone, still feels like a gift. Run the same gesture every six weeks, and customers stop reading it as a thank you. The code is a tool, and the problem starts when it’s the only tool you reach for.

What a reason to come back looks like

If the only reason to buy again is a lower price, that’s the whole relationship. The brands people actually come back to give their customers a reason beyond price.

Pull your order data first: what do customers buy after the hero product? If people who buy X usually come back for Y, test it. Run a bundle pairing them, or a gift with purchase aimed at customers who bought X and never came back for Y.

Digital gifts can do the same job when there is no second product to give: a guide, a template, or exclusive content that goes out with the order and still feels like a gift.

Whichever you run, worth testing it against a holdout and look at contribution margin over the next few purchase cycles. If those customers were going to buy the second product anyway, you’ve effectively paid them to do it sooner.

Same with recognition. A customer’s twelfth order should feel different from a stranger’s first. That can look like a note, a small gift, or just a human noticing that they’ve been around for years. An email that says you’ve ordered this twelve times, this one’s on us, lands differently than 15% off today only.

BFCM is for them too

The person who paid full price in September is about to watch you hand 30% off to complete strangers for four straight weeks. The people who give you the most predictable revenue often get the least exciting treatment. And somehow we call this a loyalty strategy.

Start with your own history too. Pull up what you offered last BFCM and the year before: your long-term customers remember it better than you do. If last year was 40% off sitewide, that may be their reference point.

A smaller discount can still win if you explain and test the value of the different offer instead of assuming the bundle speaks for itself.

You don’t have to skip the sale. BFCM is real and the revenue matters. But the customers who already carry your business deserve a version that feels built for them: real early access instead of a subject line that says early access, first dibs on the thing that sells out every year, an actual gift with the order.

Last week was the boring basics: a good product, a fair price, showing up when you said you would. This week is what happens when the promo calendar starts undermining all three.

Before the sale starts, pull up the customers who bought repeatedly over the last twelve months. Look at how often they paid full price, how their behavior changes around your promotions, and what happened to them after last year’s BFCM. If you want customers buying at full price in January, spend a little less time teaching them that November is the only sensible time to shop.

Give your best customers something to come back for besides the next code.

That’s it for this week!

Any topics you’d like to see me cover in the future?

Just shoot me a DM or an email!

Cheers,

Eli 💛

Hi Team!

Welcome back to school. If I ever get to teach a fall class at a university, this is the syllabus. Consider it my TED talk, minus the headset mic and the big red circle on the floor.

Every subscription brand I talk to asks the same question: how do we 10x retention? Nobody likes the real answer, because the real answer is boring. It’s the basics. And in an age of AI, the basics are going to matter more than anything.

Let’s get into it.

This week’s newsie is brought to you by my friends at Horatio!

Picture this: it’s peak season, tickets are flooding in, and your outsourced team is handling refunds, subscription changes, and card updates as fast as they can.

Now picture one of those agents working from a laptop with no screen lock, on a shared login, with no record of what they clicked.

It’s an easy gap to miss. When you’re choosing a partner, cost, speed, and CSAT understandably come first, and workstation setup rarely makes it onto the list.

But when the team handles billing, those details matter, and they usually only come to light during an audit or after a customer complaint.

And since it’s your helpdesk and your customers, that’s where the impact is felt most.

The good news is that you don’t need a compliance background to catch this stuff. You mostly need a list of what to ask, so we wrote one.

The SOC 2 & GDPR Checklist for Outsourced Billing Support covers workstation setup, access controls, and the compliance basics. You can use it to vet a new partner or to check the one you already work with.

If you’re bringing on seasonal agents for BFCM, it’s worth going through before they get access.

👉[Get The SOC 2 & GDPR Checklist]

The 10x question

Nobody asks how to make the product show up on time. Nobody asks whether the shipping estimate on the site is honest. Those questions aren’t sexy.

Sorry, Chad, but retention is mostly boring. The brands that keep customers for years aren’t running super secret black-hat plays. They made a good product, priced it fairly, shipped it when they said they would, and taught people how to use it. Then they did it again the next month.

In an age of AI, the basics matter more, not less. AI will write your emails, answer the bulk of your tickets, and build your segments, and every competitor gets the same toys.

What it can’t do (yet) is roast good coffee, charge a fair price for it, and get it to the door on the day you promised. People just want something reliable. They always have.

The coffee test

Think about your own subscriptions. I get coffee on subscription because I drink coffee every day. If the coffee is good, the price is right, and it shows up when expected, canceling rarely crosses my mind. Which is precisely why I subscribed in the first place. I’d have to go out of my way to leave, and I’m lazy, so I stay.

That’s the whole game. Expectation on one side, reality on the other. When reality matches expectation, retention takes care of itself. When it doesn’t, no winback flow ever written will save you.

The bar is rarely delight; it’s doing what you said you’d do, every month, without making me think about it. The moment I have to think about my subscription is the moment I start doing math on it.

Why subscribers actually cancel

When someone cancels, it’s almost always one of four recurring problems I see behind subscription churn.

There was never a real need. Your 40% off first-order promo + 6 free gifts got them in the door, but they were never going to need you every month. That’s an acquisition problem…

They don’t understand the efficacy of the product. They bought the supplement, the serum, the software, and nobody taught them how to use it, what results to expect, or when to expect them. Confused people cancel.

It wasn’t as expected. The box was late, the flavor was off, the size ran weird. Reality missed the promise.

It wasn’t worth the perceived dollar value. Notice the word perceived. The product can be objectively great and still lose here if the customer never understood what they were paying for.

The first one is an acquisition problem. The other three are basics you can fix before the customer ever complains, and not one of the four gets fixed by a bigger discount.

Fill the valleys first

Brands love building peaks. The big billboard launch, the crazy promo, the gift with purchase, the outlandish surprise and delight program. Peaks are fun to work on, and they screenshot well for the board deck.

Valleys are where retention actually lives. The shipment that went out two days late. The billing email that confused someone into a chargeback. The support reply that answered a question nobody asked.

You can run the craziest promotion in your category and get someone in the door exactly once. If the basics are broken, that’s all you get: one order, one disappointment, one cancellation. A customer who rarely has a bad month has little reason to leave.

The law is catching up

On October 1, New York City became the first city in the country where canceling a subscription has to be as easy as signing up for one, by law.

If a company can take your money with one click, you can leave with one click. Signed up online, you cancel online.

No more sitting on hold for 45 minutes to quit a gym you last visited in January. Fines start at $525 per violation, and the city even launched a complaint portal for it. The projected savings are up to $162.5 million a year, which is a polite way of saying that’s how much the friction was worth.

The FTC tried to do this nationally, and a federal court killed the rule last year, days before it was supposed to take effect. So now it’s happening state by state instead.

More than half the states already have auto-renewal laws; New York State passed its own version last spring, California’s has been in force since last summer, and I’d bet most of the rest follow within a few years.

If your retention plan depends on it being annoying to leave, you are on a clock. TIK TOK.

Your pre-BFCM audit

BFCM is around the corner, which makes this the right week to audit the boring stuff. Consider it homework. It will be on the exam, and the exam is Q4.

Subscription flows. Can a customer skip, swap, or pause without emailing support? If not, you’re generating tickets and churn on purpose. Smart subscription flows let people adjust instead of cancel.

Nurture flows. Look at what goes out after the first order. If the sequence is three discount emails and a review request, that’s not a nurture flow. Teach people how to get value out of the thing they just bought.

Education. Do customers know how to use the product, when to expect results, and what working looks like? Churn reason number two is entirely on you.

Shipping promise. Is your estimate honest, and does it hold? Underpromise by a day and overdeliver. Nobody has ever canceled because a box showed up early.

Consumption and cadence. Look at when customers actually need another shipment versus when you’re sending one. Someone sitting on three unopened bags of coffee probably isn’t excited about a fourth showing up tomorrow. Give them an easy way to adjust frequency before the next charge becomes the reason they cancel.

None of this is a rebrand or a replatform. It’s a week of honest auditing and a few fixes, done before the flood.

Retention was never about the cleverest play. It’s the basics, done well, on repeat, while everyone else hunts for a 10x button.

Class dismissed. Don’t forget the homework.

That’s it for this week!

Any topics you’d like to see me cover in the future?

Just shoot me a DM or an email!

Cheers,
Eli 💛

Hi Team!

A bit different agenda this week. I brought in my good friend and CX leader Michael Bair (former SVP CX, FIGS, etc.) to prep us for BFCM.

Michael is one of the sharpest CX leaders I know, and I am excited for him to share practical CX prep ahead of the most wonderful time of the year.

Let’s get into it.

This week’s newsie is brought to you by my friends at Horatio!

There’s nothing quite like the sinking feeling of a dispute notification landing in your inbox for an order you know shipped.

Let’s be real, chargebacks are a total headache for e-commerce brands. Between unrecognized billing descriptors, shipping delays, and plain old customer confusion, disputes happen to even the best companies.

And they sting more than you’d think. You can lose the product, the revenue, and get hit with a chargeback fee on top. Let enough of them pile up, and your payment processor starts asking uncomfortable questions.

But here’s the thing: once that notification hits, you’re already on the clock. If you’re digging through old emails for tracking numbers or figuring out your response on the fly, you’ve likely already lost the customer and the revenue.

The brands that win disputes (without burning out their team) are the ones that have a system ready before the chargebacks start rolling in.

That’s exactly why Horatio put together The Chargeback Response Playbook. It shows you how to classify disputes in seconds, set up friendly outreach protocols, and assemble rock-solid evidence packets before formal arbitration kicks in.

Grab it now, before you need it. 👇

Download The Free Chargeback Response Playbook

BFCM Stress Test, by Michael Bair

“I cannot remember the books I’ve read any more than the meals I’ve eaten; even so, they have made me.” — Ralph Waldo Emerson

I love reading. I’m on my 26th book this year (Atmosphere, What Kind of Paradise, and All Good People Here are my favorites so far), and I get discouraged when I hear that reading for pleasure is at an all-time low. But then I remember that reading for pleasure isn’t limited to books. It also includes incredible educational and entertaining newsletters like this one.

I’m encouraged, too, when I talk to people who’ve turned a newsletter into a CEO role (shoutout, Zach Bingham) and read news like last week’s announcement that Express Checkout was acquired by Morning Brew (congrats, Nate and Jenna). Passionate, high-quality writing that builds a loyal audience has never been more valuable.

That’s why I’m honored when Eli asks me to guest-write his newsletter. I take that seriously.

With BFCM six weeks away, I want to talk about why it’s your CX stress test. If you’re a founder who’s nervous about whether your CX team is ready for the coming season, this is for you.

First, what qualifies me to talk about scaling? I’ve seen this rodeo a few times. My first startup was acquired in 2011. Soon after, we went from supporting 125,000 customers to 800,000 when our parent company acquired another company and folded it into our brand.

In 2015, I joined my next startup, Inspire, as its 12th employee. Over four years, our member base grew from 10,000 to 150,000, and we handled more than 10,000 retention interactions a month.

At my last full-time role, as SVP of Customer Experience at FIGS, we grew the CX team from 10 to more than 150 agents in North America and supported 700,000 interactions a year, all while going public in 2021 and as a multibillion-dollar company.

Different industries, different problems, but in each one, CS/CX had to scale with the business’s rocket-ship growth.

One of the hardest parts of figuring out whether you’re ready to scale is that you rarely get a test run. In e-commerce, BFCM is a great one. For many businesses, weekly revenue can increase tenfold, putting enormous pressure on pre- and post-purchase support.

Chat volume can quadruple overnight. Email backlogs reach new highs. Phone queues hit wait times you didn’t know were possible. And suddenly, the founder realizes: We were not ready for this.

BFCM is stressful because it exposes and breaks already-fragile people, policies, processes, and platforms. That’s true of CX, but it’s also true across much of your business. So why don’t we plan better for CX?

Some CX leaders are inexperienced with this kind of scaling. They think it’s simply a capacity issue: more orders mean more tickets, so I just need more people, right? Not quite.

When businesses scale, four things happen.

First, problems compound. An address-change email you would normally answer on time gets missed. Now it becomes a cancellation, return-to-sender, or return request. One ticket you’d normally answer in four hours turns into four tickets you’re answering over the next two weeks. Multiply that by ten.

Second, you can’t keep up with manual processes. Reviews go unmoderated, proactive research in Track by Loop doesn’t happen, orders sit in pre-transit, and fraud alerts in Shopify and Loop pile up. The day-to-day work starts falling through the cracks.

Third, processes that require a manager’s review in Slack pile up and get missed. Managers lose their minds trying to keep up with all the questions.

Fourth, and this is a new one, AI scales your incomplete processes at lightspeed. Incomplete process documentation, circular-reference workflows, and informal quality assurance can lead your AI to touch ten times as many tickets as you intended, and you may not be able to turn it off.

Have I caused enough heartburn and anxiety? Don’t worry. You have six weeks. That’s more than a thousand hours. As my son’s soccer team keeps yelling at me: “I’m bout to lock in, Coach Mike.”

Over the next six weeks, work with your CX leadership team on five areas:

People. Make sure everyone on the team has clear roles and responsibilities and knows what great looks like, with metrics to match. Start talking with them today about a six-week roadmap to readiness.

Policies. Review every customer-facing policy- shipping, returns, warranties, and so on- especially how it applies during your BFCM promotion. Is it clear and customer-friendly? Does it make customers want to shop with you again?

Processes. Search your CX knowledge base for “Slack” and count how many processes require someone to ask for permission, help, or access. Eliminate those steps or temporarily suspend them.

Platforms. Audit the readiness of each CX platform: help desk, returns, chargebacks, reviews, and surveys. If you manually review certain return requests, what would happen if those requests increased tenfold? If only one person moderates your social DMs and needs backup, how will the new person get access?

AI. Set up time with your CSM and ask them to audit your configuration. If you use a tool like Gorgias, you can also ask its internal AI agent, Gaia, to audit your setup. Review your settings, handover rules, and knowledge resources. Start QA’ing 10%, then 20%, then 30% of daily tickets, and make changes based on what you find.

Okay, that was a lot. But this is your stress test. It should be a little stressful.

When I first started in CS, before CX was really a thing, a leader’s experience with scaling was measured by the highest number of seats they’d led and interactions they’d handled. Today, I think scalability is about readiness and flexibility.

Readiness means having a plan. Flexibility means being able to handle a sales spike, a promo redesign, a lost pallet (or 10) at the 3PL, an influencer posting about you on TikTok, or any number of events that blow up your plan.

I started with a quote, and I’ll end with one. Mike Tyson famously said, “Everyone has a plan until they get punched in the mouth.” BFCM is coming. Even if it goes well, it’ll feel like a right jab to the face. Get your plan ready, but stay flexible enough to take the jab and throw a hook back.

If all of this feels overwhelming, I’ve got you. Email me at michael@bairconsulting.com or DM me on LinkedIn.

We’ll start training tomorrow.

That’s it for this week!

Any topics you’d like to see me cover in the future?

Just shoot me a DM or an email!

Cheers,
Eli 💛

Hi Team!

We're at a coastal resort in Dubrovnik after a few days at an Airbnb in Kotor Old Town and a quick pit stop in Bosnia to mark #42 off my list.

A trip with the two little brats is more of a change of scenery than vacation, but we’ll take what we can get.

For the points people: the cash price was around $1,500 for three nights. We booked with a 65,000-point boost and a $250 Chase Edit credit. $0 out of pocket.

Being away has me thinking about hospitality, and the part of CX I’d actually want to spend my days doing.

If AI handles 90-95% of CX, the job I'd want is the rainmaker: the maitre d' who can comp the meal or bring out the chef when things really go to sh!t.

You don't touch every table, but you have the levers to pull when someone needs you.

Let’s get into it.

This week’s newsie is brought to you by my friends at Horatio!

If you're planning your holiday support around last year's numbers alone, it's worth taking a second look.

BFCM ticket volume doesn't rise neatly with sales. It tends to spike suddenly, often triggered by the same few culprits: shipping delays, stockouts, checkout issues, or promos that confuse customers.

When response times slip during those surges, the impact shows up quickly in abandoned carts, rising refund requests, and loyal customers losing patience at the exact moment you most need to keep them.

The teams that handle peak season well, without burning out their agents, are usually the ones that spotted their weak points before the rush began.

Horatio put together a quick BFCM Readiness Quiz to help with exactly that.

In just a few minutes, it walks you through your holiday support setup and highlights operational risks that are easy to overlook.

Afterward, you can book a call with a Horatio expert to review your results and fine-tune your plan.

Take the BFCM Readiness Quiz

The maitre d' of CX

Think about a maitre d' who doesn't touch every table, or a restaurant manager who isn't carrying every plate out of the kitchen.

When something goes wrong, they can step in to fix it. They can comp the meal or bring out the chef, depending on what would actually help.

The rainmaker would take over when the normal process has stopped helping. For someone who loves hospitality, there's real beauty in getting to be that person.

This is more hospitality and creative thinking, with less of the BS that comes with general CX work. For those of us who like hospitality (me) , it's the 1% most fun part of the job.

Let them comp the meal

Say an order is missing, and it’s for a wedding this weekend. The customer probably understands that they can get their money back, but they’d also quite like to have the thing they bought for the wedding.

The rainmaker could check whether a replacement is available, speak to the warehouse, and find out whether faster shipping would actually get it there in time.

They could call the customer to work through the options, including being honest when the deadline can’t be met rather than sending another optimistic tracking link and hoping for the best.

Maybe the fix costs more than the original order, or maybe a refund and a straightforward explanation are the best available outcome.

Give someone the authority to weigh those options and make a decision they're comfortable explaining afterward. That means clear spending limits, access to the relevant systems, and someone who can quickly approve an exception when the situation calls for it.

This might be a small share of your tickets, but these are the situations where someone's judgment can change how a customer feels about the entire brand. It's the part I'd actually want to spend my day on.

(Caveat: if you’ve not yet shared tickets with your AI overlords, this is way too expensive of a role to hire on top of a full human team. I think the rainmaker role exists when 80+% of your tickets are being handled or assisted by AI.)

Your rainmaker might already be on the team

You could hire someone from hospitality, but you might also find this person on your existing CX team, secretly losing their mind because they know how to help and keep being told they can’t.

This role is definitely not for people who view CX as punching out tickets. Look for someone who wants to understand what matters to the customer and can stay steady when a conversation gets uncomfortable. They need to know what the business can actually deliver before they promise it.

Sending a discount to someone who wants an explanation can feel like we're trying to get them to go away, even if we've spent money trying to make them happy. Having permission to comp the meal doesn't mean that's always what the person needs.

Give a candidate a messy customer situation and ask them to talk you through it. What do they need to check before deciding what to offer?

Listen to how they think about the person and what happens after the call. A beautifully written apology won't get the wedding order there.

If you've spent years measuring your team on how quickly they close tickets, that's the job you taught them to do. Give them room to work through a messy situation before deciding they aren't suited to it.

Give them time to actually help

Give this person time to get the wedding order sorted before you hand them the next escalation. They need someone who can cover for them when they're tied up with the warehouse or on a long call. Otherwise, you've given them a new title and the same queue.

Start the day with the cases that need attention. The wedding order needs an answer now; a replacement promised yesterday needs checking before you make another promise today. Then get on the phone or work with operations to get the fix moving.

On a quieter day, let them call someone back and find out whether the replacement actually arrived. Don't fill every gap with more tickets. Following through is part of the job you're paying them to do.

If the same delivery promise keeps causing problems, give this person access to whoever can change it.

They should be able to show the team the orders that missed the deadline and ask for the promise to be fixed. Comping the meal gets old if the kitchen keeps sending out the same problem.

That’s it for this week!

Any topics you’d like to see me cover in the future?

Just shoot me a DM or an email!

Cheers,

Eli 💛

Hey team,

In a couple of hours, we head out on a flight to Dubrovnik, rent a car to drive to Kotor, and spend a few days in the Balkans for a quick change of scenery.

The most fun part of this trip is that this is the first trip I have fully entrusted to the bots to book entirely for me… let’s see how it goes 🙃

• Instinct found me a great points flight after checking for United saver redemptions daily for a few weeks.

• Muse used my Chase Ultimate Rewards points to book a “points drop” hotel in Dubrovnik; the cash price was $1,500 and we used 64k points plus the $250 EDIT credit from Chase Sapphire Reserve.

• Muse also went back and forth negotiating with countless Airbnbs until we got a great deal in a top 1% listing in Kotor old town, saving more than 20%.

P.S. First few folks to grab this Muse invite, add code CRX40I in settings for 1B tokens. Instinct invite here too. You’re welcome.

Handing a bot my vacation is easy when the job is clear, and the bot can actually finish it. CX gets messier when the bot writes a beautiful answer, can’t take the action, and still gets counted as a win.

A customer asks for a refund, gets a polished explanation of the refund policy, and comes back two days later to find a human asking for the same order number they already gave twice.

Somewhere, that first conversation is sitting in a report as an AI success.

Let’s get into it.

This week’s newsie is brought to you by Kim.cc.

I’ve been thinking a lot about the gap between *having AI in your support stack* and actually having AI that gets better over time.

NovaaLabs is a pretty good example.

They’re an 8-figure ecommerce brand and had been using Gorgias AI for support for over a year.

And the frustrating part wasn’t that the AI made mistakes. Obviously it will.

*It was that it kept making the same mistakes.*

Their team would catch something, correct it, and then run into a version of the same problem again.

Which raises a pretty important question:

If your team constantly has to watch, correct, and retrain the AI… how much work did you actually automate?

Ahead of BFCM, NovaaLabs moved to Kim.cc.

Within the first month, *45%+ of their support tickets were being handled through Kim, with less than a 3% re-open rate.*

And maybe even more importantly, their internal team did almost none of the heavy lifting to get there.

Kim had CX, AI training, and implementation experts handling the setup, training, QA, and ongoing improvement of the system.

That distinction is interesting to me.

Most AI support products give your team a tool and then leave you responsible for making that tool successful.

Kim is an *AI-native service*, which means the AI handles the tickets. Their experts, also known as "Sentinels", make sure it’s learning, improving, and not repeating the same mistakes over and over again.

So NovaaLabs went from spending time managing the AI to having *45%+ of their tickets handled within a month, <3% reopens, and close to zero implementation lift from their own team.*

Especially heading into BFCM, I think that’s the metric more teams should be looking at.

“How many did it resolve well enough that the customer didn’t have to come back?”

If you’re thinking about AI support before peak season, Kim.cc is worth a look.

Learn more about Kim.cc

Please let me fix the easy thing myself

I’m comfortable letting an agent spend weeks checking flight availability because I have absolutely no interest in doing that myself. The dates and constraints are clear, and there’s a pretty obvious definition of success.

If we land and the hotel has no record of our booking, my enthusiasm for experimentation will depend heavily on whether someone can get us a room.

The same customer can love automation for one part of an experience and need a person five minutes later.

That makes considerably more sense to me than declaring a brand “human-first” or “AI-first” and forcing every situation to fit the announcement.

The self-service experiences at Quince, Wayfair, Target, and Amazon are my favorite examples. For eligible issues, you can go into an order and work through the available resolution options without even opening a conversation with a human.

For a straightforward exchange, I’d take that over explaining the same thing four times to someone who then needs to “check their system.” Let me pick the right size and get the return instructions, and I’m perfectly happy.

This is where anyone starting with AI should slow down for a second, because some of your biggest wins are boring self-service fixes.

Before you build a bot that can discuss your return policy in eleven different tones, make sure customers can actually start a return.

AI earns its keep on the messier requests: someone doesn’t know which order contains the item, explains three problems in one message, or needs help figuring out which option fits.

But it needs the data and permissions to finish. Otherwise, you’ve built something that can confidently describe a refund while being entirely unable to issue one.

The customer still has work to do, however nice the explanation was.

What exactly are we calling resolved?

If you reward a system for keeping conversations away from humans, as we lovingly call “deflection” in this industry, you need to know what happens to the people who never reach one.

Some got exactly what they needed. Others may have given up, tried another channel, or decided they’d deal with it tomorrow.

If your reporting treats a conversation ending without a human as a success, those outcomes can look remarkably similar.

The bot recites the policy, the customer closes the chat, and then emails support the next morning. Unless you connect those interactions, you end up celebrating the first one while paying someone to handle the second.

I don’t think the people building these systems necessarily have bad intentions. They’re often working toward the target they were given.

But I would want “resolved” to involve checking that the action happened and addressed the request. A refund explanation and a processed refund should not earn the same little green tick.

Recontact and reopen rates are useful checks, too, provided you can follow the issue across channels. Silence alone isn’t proof that somebody went away happy.

Read the rest of the message

“Where is my order?” could mean the customer wants a tracking link.

It could also mean the replacement for a missing order has now disappeared, the customer leaves tomorrow, and someone on your team already promised this shipment would arrive in time.

The tag or subject line might be identical. Sending both customers the same tracking response would be a fairly bold choice.

Before the system keeps going, I’d want it checking:

• Does it have reliable information? If the order record and carrier update disagree, what needs investigating?

• Can it take an authorized action that solves the request? Explaining the policy doesn’t count as completing the action.

• Has this already failed? A customer returning with the same unresolved problem should not be sent through the same sequence by default.

• Would a mistake be hard to reverse or carry meaningful risk? Those cases need stronger checks and, where appropriate, a qualified person.

• Has the customer asked for a human? There should be an obvious way to reach one without guessing the phrase that unlocks the door.

Frustration alone shouldn’t decide the route. An annoyed customer might be thrilled with an immediate automated replacement. A calm customer might be asking for something the system has no business approving.

I’d also be careful about creating a setup where becoming more upset is the fastest way to get useful help. Customers will learn that lesson.

A transfer to someone who can’t do anything is just a longer wait

I’ve spent enough time in CX to know how much a good person can do with a complicated situation.

They notice when the proposed fix doesn’t fit, ask the question nobody thought to ask, pull in another team, or make a judgment call within their authority.

Take a gift that arrived damaged the day before someone needed it. Another shipment arriving next week may be useless. A person can work through the alternatives with the customer instead of repeatedly offering the one option the system happens to support.

But that person needs access, authority, and a clear path to whoever can make the decisions they can’t.

If your rep can only repeat what the bot already said, the transfer has mostly changed who is delivering the disappointment.

AI adoption is a terrible excuse to make human help harder to reach. Lovely people are also not a good excuse for making customers wait through a slow, repetitive process. Either way, the customer ends up paying for how strongly you feel about your approach.

The customer should not have to train the next agent

This is the handoff I’d spend real time designing.

Before a person joins, they should see what the customer wants, what already happened, what was tried, and why the system stopped.

If the customer uploaded a photo, bring the photo. If someone promised delivery by Friday, surface that promise. If the bot couldn’t proceed because the records disagree, show the discrepancy.

Keep the original conversation available, too. The rep should be able to check the summary, and an AI suggestion should be clearly identified as a suggestion rather than something already approved.

“Customer is frustrated about delivery” leaves the next person redoing the investigation, now with a customer who has had additional time to become frustrated about delivery.

For a hypothetical damaged replacement, a useful opening would be:

“I can see this is the second damaged item and that you need it before Friday. I have the photo you sent. I’m checking whether we can get a replacement to you in time, and what we can offer if we can’t.”

That person has shown they read the conversation without promising something they haven’t checked.

The customer also needs to know who owns the next step and when they’ll hear back. If the team won’t be available until morning, tell them. “Connecting you now” should involve some actual connecting.

Pick one ugly ticket and follow it to the end

Start with one recurring issue your team already knows gets weird. Review real cases with the people who handle it. Find where the answer is clear, which details change the decision, and what a good escalation looks like.

Then test the ugly versions: out-of-stock replacements, conflicting records, previous failed attempts, customers who can’t use the standard return method, and tickets where the bot gave the right policy but the wrong resolution.

Check whether the action happened correctly, how much effort the customer spent, and whether they needed further help. When something escalates, look at whether the person had enough context and authority to move it forward.

Use what you find. If the same missing information keeps forcing a transfer, fix the information gap. If reps repeatedly wait for the same approval, look at their permissions. Those are decisions someone still needs to own after the software goes live.

I’ve entrusted the bots with quite a lot of this trip. If we arrive in Dubrovnik and the booking has gone sideways, I’m not going to care whether the fix comes from AI, a human, or a guy named Luka at the front desk.

I’m going to care whether someone understands what happened and can get us a room.

We can discuss the future of customer service after we’ve put the bags down.

That's it for this week!

Any topics you'd like to see me cover in the future?

Just shoot me a DM or an email!

Cheers,

Eli 💛

Hi Team,

We’re practically in October… so close to BFCM I can smell it.

I read a post this week from Sean Frank, Ridge’s CEO, that stuck with me. He said revenue would fall 70% if he took his foot off the gas. That’s a hell of a dependency to have.

His point is that selling a wallet today doesn’t automatically help sell the next one. You still have to find and convince another customer, without the network effects you get in some other businesses.

I wouldn’t pretend a better support team could make that dependency disappear. But it did make me wonder how much we spend acquiring customers we then leave to figure out the product themselves. By the time they need something else, we’re back in the auction trying to convince them again.

With BFCM coming, I think that deserves a little attention.

Let’s get into it.

This week’s newsie is brought to you by my friends at Horatio!

Seasonal hiring is a lot more expensive than it looks on paper.

If you’ve ever built out a temporary team for peak season, you know the drill: recruiting on a tight timeline, background checks, training people up fast, and then someone on your team taking on the extra work of managing it all.

And just as everyone is fully ready? Peak season’s over, the team’s gone, and you’re starting from scratch next year.

Once you add up the real cost per agent (recruiting, training, management, churn, and all the little hidden costs) that “budget-friendly” seasonal team isn’t looking so budget-friendly anymore.

Wondering how your setup stacks up? See exactly where you stand with Horatio’s side-by-side comparison chart. And if you’re curious what a hybrid model could look like for your specific order volume, their team’s happy to help you map it out.

👉 See the Real Difference

The first order should help earn the next one

Ridge has built a huge business by getting very good at finding the next customer. Sean’s point is that the work keeps coming: selling a wallet today doesn’t create a network effect that brings tomorrow’s buyer along.

A wallet that lasts for years also shouldn’t need replacing every few months. The next sale might be a gift, another product, or a friend buying because someone they trust recommended it.

We spend a lot of money getting someone to believe the product page. Once they buy, we seem considerably less interested in whether they still believe it.

If you want the wallet customer to consider your other categories, the first purchase needs to give them a reason to trust you with another one. Knowing how to use it, getting a question answered, and finding out whether the guarantee is any use when something breaks all contribute to that decision.

Those moments can also give someone a reason to look elsewhere, without ever opening a support ticket.

The questions your support team never hears

Imagine a customer buying a $160 pair of boots. The ad says they get better the longer you wear them. The product page says they’re built to be resoled instead of replaced.

They arrive in a plastic bag inside a plain box, with no information about break-in or care. He wears them Monday, and by Wednesday they’re hurting his feet. He doesn’t know whether they need time, whether he bought the wrong size, or whether these particular boots just aren’t for him.

So Thursday night he types “do leather boots hurt this much at first” into his phone and spends twenty minutes on a subreddit where four strangers explain his own purchase to him. Two of them recommend a different brand, which is a generous little acquisition service for your competitor.

He had a question about your product, and somebody else was there to answer it.

Later, when the heels wear down, he remembers the resole promise. The link is in the footer and opens a form asking for proof of purchase he can’t find. Shipping is on him, and the turnaround is six weeks.

Six weeks without the boots he wears every day is a fairly compelling reason to buy another pair. He closes the tab and keeps wearing them until he’s ready to replace them.

When that day comes, he opens three tabs. One of them is the brand from that thread.

Nobody escalated anything, nobody asked for a refund, and your CSAT for the quarter is lovely. He never created a ticket, so your support report has nothing to say about any of this. You paid to acquire him, then left him to work out the ownership part with strangers.

A care guide could have explained break-in and when to ask for help with the fit. An order lookup could have spared him the receipt hunt. And if resoling is a reason to buy the boots, I’d want the cost and turnaround visible before checkout, with a process worth using when the time comes.

Nobody puts a car fire in the annual plan

The first purchase can also give people something worth sharing. In her reply to Sean, Alex points to Crocs and Stanley as examples of the brand heat Ridge could try to build, while acknowledging how much riskier that is than a familiar, data-driven acquisition playbook.

You probably remember the Stanley video: a woman’s car burns, her Quencher is still in the cupholder, and there’s still ice inside it. The president responds on TikTok and offers to replace her car.

The product got an absurdly good demonstration of its durability, entirely by accident. Then the company gave people a second reason to talk about it. Replacing the cup would have been understandable. Offering to replace the car made the response part of the story.

It’s a maddening example if you’re trying to plan for it. You can put a media budget in a spreadsheet and argue about the assumptions. Nobody can promise you a Stanley moment in next year’s plan.

I would, however, want to know whether my company could respond to one before the internet moved on. If replacing a damaged order requires three approvals, I have some concerns about our readiness to buy someone a car.

Someone has to notice, someone has to be able to make a call, and the response has to arrive while people still care. Oh, and it doesn’t always need to be a goddamn car. There are much quicker, cheaper ways to do this.

Most brands will never get a Stanley moment. They will get a customer trying to use a repair promise, a gift that arrives damaged, or a question the product page never answered. Those are less spectacular opportunities to give someone a story worth repeating, and they turn up without requiring anyone’s car to catch fire.

I’d happily keep funding the acquisition machine while working on this. I’d also want someone responsible for making sure the experience we’re selling holds up after delivery. Otherwise we’re paying to bring more people into the same disappointing process and wondering why nobody recommends us.

What I’d actually fix before BFCM

I’d work through four things before sending another wave of first-time customers into the same experience:

• Buy and return your own product. Use a new email, pay normally, and give nobody an internal heads-up. See what the customer has to figure out without the benefit of working at your company.

• Find three unanswered first-month questions. Ask your support team, read reviews, then search the questions yourself. If Reddit explains how to use your product better than you do, you have some fairly obvious homework.

• Compare early reviews with later ones. Someone who opened the box yesterday can tell you about delivery and first impressions. Someone who’s used the product for months can tell you what held up, what became annoying, and what they wish they’d known.

• Try using your guarantee. Follow the actual steps. Check the cost, turnaround, and information required. If the customer has to find a receipt your team could look up in thirty seconds, fix that before writing another paragraph about how much you care.

Then I’d check what the rep can do when one of these problems reaches them. If the customer clearly qualifies for a repair, can the rep arrange it? If the case needs judgment, is there someone available to make the call? A good policy is considerably less useful when everyone is waiting for the one person who’s allowed to apply it.

Before BFCM, I’d put the product page next to the experience of actually owning the thing and work through every place they disagree, from the care instructions and repair process to the person who can make a decision when something goes wrong.

You’ve already paid to persuade this customer once. When someone asks them where to buy boots, I’d like them to recommend the pair they own, with a better reason than “the ads were good.”

That’s it for this week!

Any topics you’d like to see me cover in the future?

Just shoot me a DM or an email!

Cheers,
Eli 💛

Hi team,

In case you missed it, the normies are getting AI agents. Maybe not your mom or my mom just yet, but this is getting considerably closer to people whose weekend plans don’t involve the words GitHub commit.

Between Instinct, Grok Bot, and Meta launching Muse, more people will have software that can actually handle something for them while they go live their lives.

If you think “using ai” is asking ChatGPT to rewrite an email, the world is moving fast. These agents can work across connected accounts and browsers, keep going after you step away, and, depending on the product, communicate through text or WhatsApp. Muse and Instinct can even make purchases through Stripe’s Link.

I’ve been using instinct for a few weeks, but in the last week or so, I have had a few experiences that made me wonder how drastically CX will change in the short term.

“Have your agent talk to my agent” is becoming a fairly reasonable household plan.

Let’s dive in.

This week’s newsie is brought to you by my friends at Horatio!

Customer retention isn’t won when things go perfectly. It’s won in the 48 hours before a disruption hits.

We all know the math: a site outage, an inventory glitch, or a shipping delay is statistically going to happen in Q4. But customer sentiment doesn’t drop because of the error; it drops exponentially because of silent errors. If you wait for the tickets to explode before you say something, you’ve already lost the LTV. Winning brands use proactive, omnichannel customer warnings to head off the chaos before it starts.

Horatio’s Human + AI 24/7 Coverage natively plugs into your existing tech stack (Gorgias, Zendesk, Gladly, Kustomer, Fin…you name it!) to flag issues and communicate with customers before your inbox drowns.

Check out their dedicated BFCM hub to map out your proactive strategy today.

👉 Secure Your Trust Window

My dryer has representation now

Our dryer broke, and we had a warranty through Asurion. I gave Instinct the task of getting the claim handled, along with a photo of the appliance sticker.

It found last year’s receipt in my email, went to the warranty website, and encountered the requirement to file claims over the phone.

I have children and a job, so discovering that an appliance requires a scheduled conversation is generally where my enthusiasm for owning a warranty starts to deteriorate.

Instinct went into live chat and worked through it until the claim was accepted, the repair scheduled, and written confirmation arrived that there would be no service fee.

It even attached a screenshot of chat.

As a customer, I was delighted. As someone in CX, I started thinking about all the requests people never bother making.

Brands forecast support demand using behavior that already includes people giving up.

The small charge you never disputed, and the return you abandoned after discovering you needed a printer never became tickets. Those problems existed; your willingness to spend an afternoon on them did not.

Naturally, I started looking for other things I could hand over.

Save 15% or More on Your Car Insurance

I gave Instinct access to my Geico account, and it pulled my coverage, shopped around for the same coverage, and sent me a ready-to-pay quote at Allstate for same coverage that was priced 24% lower. That will save me $1,005 per year.

I understand that shopping around for insurance has always been possible. I also understand that I have spent years being technically capable of doing it while finding almost anything else to do instead.

There’s a meaningful amount of money sitting inside tasks people keep postponing. Once someone else can do the comparison shopping, more of those tasks become worth pursuing.

Agent, Pls Plan Date with My Wife’s Agent

I know, this is bleak… but Instinct also announced a Trusted Person network, where your agent can coordinate with the agents of people you’ve connected with. One possible use is having my agent talk to my partner’s agent to arrange a date night.

Having two virtual assistants negotiate when we can eat dinner together feels like peak Silicon Valley, although I would like to hear their position on finding a babysitter before dismissing the concept.

That feature is restricted to a trusted network. It doesn’t mean my agent can already contact every brand’s agent. But the coordination itself is becoming a product feature, which makes “have your agent talk to my agent” considerably less of a bit.

Check This Flight Every Morning

I was looking at flights to TLV for a wedding in November, and the round trip was $2,000. I asked Instinct to check daily and book if the exact same itinerary dropped below $1,400.

A week later, I received a United confirmation for $1,350.

Now my bot is checking my hotel reservation’s price every night, with instructions to cancel and rebook if a lower rate appears before my free-cancellation window closes.

I used to have time for this kind of thing. These days, remembering that I booked the hotel feels like a reasonable contribution.

From the hotel’s side, though, one reservation could now produce several changes without the guest spending their evenings refreshing a booking page.

Jevons Had a Point

In 1865, economist William Stanley Jevons argued that more efficient steam engines could increase Britain’s coal consumption. Getting more work from each ton made steam power cheaper and useful in more places, expanding demand enough to overwhelm the savings per engine. His original explanation is considerably older than anyone’s AI strategy.

AI creates the same possibility: as individual tasks become cheaper and easier, we find more things worth asking it to do. Checking a flight once becomes checking it every morning. Shopping for insurance graduates from something I should probably do to something that actually happens.

My bet is that support will experience its own version of this.

Think about the small charge you never disputed or the return you abandoned after discovering you needed a printer. Those problems existed, even though your willingness to pursue them expired somewhere between the instructions and dinner.

Brands forecast support demand using behavior that already includes people giving up.

An agent lowers the effort required to pursue those requests. Meanwhile, the brand is (hopefully) using at least some AI to make each interaction cheaper. Your cost per contact could fall while the number of contacts makes the original forecast look rather optimistic.

That doesn’t mean every queue explodes. A hotel rebooking may never generate a support ticket, and a properly resolved claim could eliminate several follow-ups.

It does mean I’d be deeply suspicious of a savings forecast that automates the brand’s side while leaving customer behavior conveniently frozen.

They’ll Still Need Permission to Do Something

My warranty experience involved Instinct working through the support chat. A future version could involve my agent contacting your agent with the receipt, order history, and relevant policy already available.

If your system can check eligibility and complete the appropriate action, that could be a great experience.

If it can only explain the policy and request patience, your team could end up overseeing two pieces of software exchanging increasingly polished messages about a refund neither has permission to issue. A tremendous moment for everyone selling tokens.

Someone still has to decide what can be approved and when a human needs to get involved. A vague response that previously ended the conversation because the customer got busy may now receive a follow-up.

What I Would Do Today

If I were running CX today, I’d test a few controlled requests against our own support experience using a personal agent. I’d want to see whether an eligible request reaches a resolution, whether an ineligible request gets an accurate explanation, and where either one gets stuck waiting for permission.

Then I’d watch completed resolutions and repeat contacts alongside cost per conversation. Handling messages cheaply won’t tell me whether the underlying work is getting finished.

I’m genuinely excited about handing over more of this admin. As a customer, I’ve already found several reasons to keep going.

If I were on the other side of those conversations, I’d leave some room in the forecast for everyone else discovering the same thing.

Have you handed an agent a customer service problem yet? I’d love to hear what happened, particularly if it got further than you expected.

That’s it for this week!

Any topics you’d like to see me cover in the future?

Just shoot me a DM or an email!

Cheers,
Eli 💛

Hi team,

I started tinkering with AI (then fondly called artificial intelligence, its full government name) in 2019 after my friend Zee pushed me to try it. I managed to make my computer sound like it was about to take off, took a break, and got pulled back in when ChatGPT launched in 2022.

I’m a normie. I don’t have a Mac Mini farm, Hermès, etc. I have kids, a calendar I do not fully control, and an alarming number of personal projects in various stages of “almost done.”

I even recently joined letsgo.co as a mentor and have had some genuinely fun conversations helping people automate the most monotonous parts of their day. Once you start looking closely at the work you keep repeating, you find plenty to experiment with.

A lot of the CX folks I speak with still feel behind, though. They’re unsure where to start, and much of what they’ve seen involves AI answering customer tickets, often poorly. I get why that hasn’t made them particularly excited to hand over more of their day.

So I put together four downloadable skills for problems you already understand well enough to have an opinion about the answer.

I think they’ll be genuinely helpful, and would love to hear your pov if you use them!!

Let’s jump into it.

This week’s newsie is brought to you by Kim.cc

I was on a webinar with the Kim team recently, and we spent a fair amount of time talking about where AI should probably stop making decisions on its own.

I’m obviously not anti-AI. I’m about to give you four AI skills to download.

But customer support gets messy very quickly.

A tracking question is one thing. A customer who has ordered six times, received the wrong product, wants a refund that technically falls outside policy, and is now threatening to leave is something else entirely.

Those are the conversations where “the AI followed the policy” is not necessarily the same thing as “the customer got the right outcome.”

And BFCM is a particularly unpleasant time to discover where those gaps are.

You probably aren’t going to replace your helpdesk, rebuild your workflows, retrain your team, and introduce an entirely new support operation a few weeks before peak. I wouldn’t recommend adding that to anyone’s November calendar either.

So Kim is doing something a little different for readers of this newsletter.

They’re picking 25 brands and helping them predict where their current support operation is most likely to break during BFCM.

They’ll look at things like your expected ticket volume, team capacity, current automation, common ticket types, and the places where human judgment still matters, then walk you through where they think the cracks are likely to show up once volume increases.

No requirement to change your stack. Just a chance to stress-test the setup you already have before your customers do it for you.

If you want them to take a look, you can book a 20-minute BFCM Support Stress Test here:

Stress-test my BFCM support →

Skill #1: CX Site Auditor

TL;DR: Reviews your customer-facing pages and messages for conflicting promises and unanswered questions, then suggests fixes and likely owners.

Use when: You want a “stranger” to do a deep dive on your site and audit where you perhaps are making promises you can not keep, or saying different things at different parts of the journey that don’t align.

Start with your store URL, then add emails or pages it can’t access. I’d include the order confirmation, since the customer reading that one has already given you money and would quite reasonably like to know what happens next.

When you get the findings, check the actual passages. “Ships in three days” and “arrives in three days” are different promises, and customers shouldn’t need a working knowledge of your fulfillment operation to decode them.

I’d pick one clear issue, figure out which promise the business can keep, and check everywhere else it appears before changing it. Otherwise, you fix the FAQ and leave the confirmation email confidently telling a different story.

I used this on a beauty brand and it found massive discrepancies on shipping timelines and promises, charging for returns in one place and on FAQs saying returns are free, and much more. It found all of this in under 60 seconds on ChatGPT using this skill.

Download CX Site Auditor

Skill #2: Ticket Second Opinion

TL;DR: Reads a customer conversation alongside your policy, recommends the next action, drafts the reply, and prepares an approval request if needed.

Use when: A ticket has gone back and forth and someone is about to ask for the photo the customer already sent twice.

Give it the conversation and the relevant policy. Then read the recommendation before getting distracted by how nicely it wrote the email.

Did it catch what the customer already provided? Does the proposed action address what they’re asking for now? Is it promising something the rep can actually do?

I’d be perfectly happy agreeing with its recommendation and rewriting the reply myself. Sometimes another read of the thread is useful. You can keep your own way of talking to customers, particularly when the customer has already spent enough time wondering whether anyone is reading what they send.

Download Ticket Second Opinion

Skill #3: Macro Reality Check

TL;DR: Reviews your macros for what the customer still won’t know after reading them, separating wording edits from decisions your team needs to make.

Use when: “Keep an eye on tracking” has somehow become your entire response to a package that hasn’t moved.

Try a macro you use constantly and one you think is already good. I’d be interested in whether it can leave the good one alone. We have enough work without a tool discovering that every sentence needs its personal attention.

For each finding, ask which customer question the suggested change answers. Perhaps the next step is buried, or the timeline needs explaining. Perhaps nobody has decided when the rep can offer a replacement, and the macro has been politely avoiding that subject.

That last finding belongs with whoever owns the policy. Get the decision made before putting a new promise into a reply your team sends all day.

Download Macro Reality Check

Skill #4: Escalation Auditor

TL;DR: Reviews escalation records against your frontline authority policy and identifies patterns worth discussing with your team.

Use when: You suspect reps spend too much time asking permission and want evidence before changing what they can approve.

It starts with the columns you supply and should tell you where missing information limits the analysis. You’ll also need the policy that explains what frontline reps are currently allowed to do.

One thing I’d watch closely: a manager agreeing with the rep’s recommendation doesn’t automatically mean the approval was unnecessary. They may have checked something the export doesn’t capture.

Pull a few of those cases and ask what happened. What did the rep already know? What did the manager verify? Could that information have been available earlier?

You might find a permission worth changing, guidance worth clarifying, or an approval step worth keeping. I’d want that conversation before announcing to the team that a spreadsheet has reorganized their responsibilities.

Download Escalation Auditor

How to use them

For conversation-based use in Claude or ChatGPT:

1. Download the skill you want to try.

2. Attach SKILL.md to a new conversation.

3. Send: “Follow the attached skill instructions. Tell me what information you need from me before starting.”

4. Supply the relevant material: your store URL, macros, a redacted ticket and policy, or escalation records and your frontline authority policy.

Start with something manageable that you know well enough to judge. Your public site or a macro without customer information is an easy place to begin. Before uploading company material, remove customer information locally and follow your company’s AI policy.

Read the output before acting on it, please. These tools recommend and draft, and your team still makes the decisions.

I’d genuinely love to hear which one you try, what it catches, and where it misses something obvious. If you have to explain a detail it should have asked about, tell me that too. That’s useful feedback for making these better.

That’s it for this week!

Any topics you’d like to see me cover in the future?

Just shoot me a DM or an email!

Cheers,

Eli 💛

The 20% Coupon Made It Worse

Hey team,

For starters, this is newsletter #200, which feels insane??? 200 weeks of writing this thing, and it means the world that some of you have been around since day one.

Some of my favorite newsletters over the years have come from random customer experiences I’ve had. The good, the ugly, and the lesson that I learned from them.

This week gave me both: one genuinely impressive recovery experience, and one AT&T experience that made me want to toss my computer through a brick wall.

Let’s jump into it.

This week’s newsie is brought to you by my friends at Horatio!

You-seriously-have-to-read-it BFCM Playbook

You guessed it: revenue-focused. Most brands approach BFCM with a defensive mindset: SURVIVE the queue, and that’s a massive missed revenue opportunity.

When a shopper reaches out during the biggest shopping weekend of the year, you know they are a high-intent buyer holding their wallet (not just a ticket). If your team is playing whack-a-mole with macros, you’re losing money.

I’ve partnered with the team at Horatio to compile real operational frameworks from top-tier scaling brands.

We’ve turned them into an actionable, plug-and-play manual to shift your support from a cost center to a revenue-generating engine.

Don’t try to win BFCM alone. Download the ultimate playbook today, steal the exact operational frameworks top e-commerce brands use, and see if you qualify for a complimentary 15-minute operational capacity audit with their team (give my name and they’ll prioritize you).

👉 Download The Free BFCM CX Playbook

The Tree Saga

I ordered two trees in the mail this week, which is apparently where I am in life now.

A prairie crabapple and a kousa dogwood. Both were shipped in one box, which showed up a few days late, and the prairie crabapple arrived snapped in half.

Generally not the preferred condition for a tree.

Honestly, I get that things can break during shipping; carriers do carrier things. The annoying part was that the entire convenience argument disappeared immediately. I paid to avoid going to the local Lowe’s, waited longer than expected, received a broken tree, called support, and then will have to wait another week and a half for the thing I should have received the first time correctly.

They had phone support, which already put them ahead of half the internet. A person answered, “activated the warranty,” told me the tree would be reshipped, and said someone would follow up.

Operationally, that was the correct first answer. The replacement was moving, the warranty process was triggered, and nobody made me explain the concept of “snapped in half” like I was presenting evidence in a small claims case.

I was still annoyed because the ticket action and the customer frustration were not the same thing. The rep solved the mechanical part of the issue, while the bigger frustration was still sitting there next to my broken tree.

I had lost time, trust, and the entire convenience argument that made me order from them in the first place. How do I even dispose of this neutered tree youngin??

Then came the make-good: 20% off my next order.

That would have landed better if the homepage didn’t already have 20% off sitewide (and ofc, you can’t combine discounts.)

So I asked to escalate, which felt ridiculous and also completely reasonable. The tree was snapped in half.

The Weekened Call Nobody Was Contractually Obligated to Make

The next morning, on a weekend, someone from the CX leadership team called me. They had already refunded 30% of the order. They apologized like a person. They acknowledged the packing issue. They talked through what they could learn from it. They said they would follow up again once the replacement arrived.

That changed the experience.

The original order still failed, and I still had to wait for the replacement. The company did not magically un-snap the tree. The difference was that the escalation finally matched the inconvenience. It added ownership, speed, a real concession, and some accountability for what happened next.

Then, because the CX gods apparently wanted to make the lesson extremely obvious, I had the opposite experience with AT&T.

Page 4, under “News You Can Use”

AT&T still remains undefeated.

My bill went up $15 a month. I found out the way most people find out these things, by looking at a number that felt wrong and going to check. I called, and the rep’s entire explanation was “all plans went up.”

So I filed a BBB complaint, because apparently that’s still a lever that works in 2026, and it escalated all the way to the president’s office.

I want you to sit with that for a second. It took filing a formal complaint with a consumer protection nonprofit to get anyone at AT&T to engage with a $15 pricing question.

And the final answer, from the president’s office, delivered like it was a gotcha, was to check page 4 of my bill, under a section literally titled “news you can use.”

“News you can use” is doing an enormous amount of unpaid legal labor in that document. It is a section designed to be technically true and practically invisible, four pages into a bill nobody reads cover to cover, in a font size that assumes you have both excellent eyesight and nothing better to do on a Tuesday.

Escalation is where the company shows its hand

A lot of brands think escalation means the customer got moved to someone with a better title.

Manager, senior specialist, executive office, retention team. Whatever internal label makes everyone feel like the situation is being handled.

That part barely matters if the person on the other end is only there to repeat the first answer with cleaner grammar.

The useful version of escalation starts with a better question: what did the first answer fail to fix?

In the tree situation, the first rep handled the warranty. That needed to happen. The replacement had to get moving. The escalation handled the part the warranty did not touch: the lost time, the bad packing, the inconvenience, and the fact that I paid for a simpler experience and ended up with a broken product plus another week-plus of waiting.

AT&T did the opposite. The first rep said there was nothing they could do. The executive office called to explain why there was still nothing they were going to do, now with a page number. 🙃

That is a more expensive and high-touch way to annoy someone.

What the complaint was evidence of

The tree company treated the snapped trunk as evidence that something in their process failed. The packaging was wrong, the shipment arrived late, the product showed up broken, and the customer was now doing work he paid to avoid.

They did not treat my annoyance like a mood to manage, rather a signal that the recovery needed to be bigger than a warranty replacement.

AT&T treated my complaint like a customer education problem.

The subtext of “check page 4” was pretty clear: you should have read this already. The price increase existed somewhere in the bill, which meant the company felt covered, and the rest of the call was basically a guided tour through why I had no case.

That is the difference I’d care about heading into BFCM.

Support teams are going to get a lot of complaints over the next few months that are technically messy:

• Promo codes that worked exactly as designed and still confused people

• Shipping cutoffs that were listed somewhere and still missed

• AI answers that technically resolved the ticket and still created a second, angrier ticket

• Customers who were warned in the terms and still feel misled because nobody normal reads terms like bedtime material

The question is what your company thinks those complaints are evidence of.

If every complaint is treated like a customer misunderstanding, your team will spend peak season explaining, defending, and pointing people to the place where the company technically said the thing.

If the complaint is treated like a signal, your team can actually learn from it.

• The promo language was confusing

• The shipping promise created the wrong expectation

• The AI flow closed tickets too early

• The make-good did not match the miss

• The escalation path had no authority behind it

Before BFCM, I’d pull the moments where your team is technically right and customers are still annoyed. Reopened tickets, escalations, promo confusion, AI-resolved conversations that came back, and high-value customers who got the same answer as everyone else.

Then ask what those complaints are being treated as.

• A nuisance to close

• A misunderstanding to correct

• A signal to fix the system

Those three paths create very different customers.

Being right on page 4 does not beat being useful on day one.

And lastly, download the BFCM CX Playbook and review your experience one more time.

That’s it for this week!

Any topics you’d like to see me cover in the future?

Just shoot me a DM or an email!

Cheers,
Eli 💛

Hey team,

We are back in the US after a week of gallivanting through Sicily and Malta, which was beautiful, chaotic, and a helpful reminder that vacation with kids is mostly parenting with extra carbs.

Feeling inspired, slightly tired, and very ready for childcare so I can catch a breather.

Over the last few weeks, I’ve gotten some version of the same question a few times:

Where should I use AI in CX, and where should I avoid?

Most teams start with “what can we automate?” which sounds practical until order tracking, refund exceptions, angry VIP customers, damaged birthday gifts, and cancellation saves all get dumped into the same giant bucket called “support.”

That is how brands get into trouble. AI is great at reading the pile, spotting patterns, pulling context, routing work, and drafting the boring stuff. The wheels come off when it starts making promises, denying refunds, apologizing to angry customers, or deciding whether a long-time customer deserves an exception.

BFCM makes this worse because volume hides bad judgment. A weak AI resolution in August is annoying. The same weak resolution in November gets multiplied across thousands of tickets while everyone is staring at response time, deflection, and queue size like those numbers tell the whole story.

They don’t.

So this week: five CX jobs AI should probably own already, and five I’d keep it far away from for now.

Let’s get into it.

This week’s newsie is brought to you by Kim.cc

BFCM doesn’t create new support problems. It exposes the ones you already had.

The tickets your AI half-resolves in October come back at 2-3x volume in November, on the most expensive week of your year.

A closed ticket isn’t necessarily a resolved ticket, and when the first answer doesn’t land, the brand pays twice: once for the AI interaction, again for the human cleanup. Tickets closed by AI carry an average re-open rate of 50%+. Going into peak, that’s a bill you haven’t seen yet.

Here is an interesting approach that 200+ Shopify brands are taking for this BFCM season: Kim.cc. An AI-native customer support service for ecommerce.

AI handles the volume and Sentinels, trained CX, QA, and prompt experts, own the outcome: they implement the system, step in where judgment is needed, and teach the AI how to handle the hard ones via proper prompt engineering so the same ticket doesn’t come back next time.

This way more issues are resolved the first time, fewer customers coming back (because suboptimal resolutions received by standard AI tools in the market), and an operation that gets sharper as volume climbs instead of buckling under it.

This BFCM brands should really try to look at the problem differently, i.e, the number that matters isn’t how many tickets close. It’s how many stay closed.

You can now book a pre-BFCM re-open rate teardown. This team is pulling a sample of your AI-resolved tickets and show you your real resolution rate — and where peak volume will hurt — before the rush hits.

You can book a 20-min call with their CX solutions expert to get your analysis done:

Learn More About Kim.cc

First, stop worshipping deflection rate

Deflection rate is one of those metrics that looks great in a board deck and gets much weirder the second you read the actual conversations.

Leadership loves it because it is clean “cost savings.” A ticket did not reach a human, the customer got some kind of answer, and the dashboard moved in the right direction. Everyone gets to feel like the operation got more efficient, which is a fun feeling until you remember the customer may have hated every second of it.

A ticket can close cleanly while the customer decides they are done with you.

I remember reading a story about Air Canada slop a few years back. Its chatbot gave a grieving customer bad information about bereavement fares. The company tried to argue the chatbot was responsible for its own actions, which is an incredible attempt at putting a name tag on the Roomba and sending it to court. The tribunal did not go for it.

Then there was the Chevrolet dealership chatbot that got prompt-injected into agreeing to a 2024 Tahoe for $1. Obviously nobody got a Tahoe for a dollar, but the point is still useful. The bot was allowed to speak in a place where the words had financial and reputational weight, and nobody had built a strong enough fence around what it was allowed to say.

AI should help your team see more, move faster, and spend less time doing soul-crushing admin work. It should not be handed the moments where money, emotion, judgment, or public embarrassment are on the table.

1. AI should tag and categorize tickets

This is boring, which is usually a good sign.

Most helpdesks have tags that look like they were built by five teams, two agencies, one intern, and someone who left the company in 2021 but apparently had very strong opinions about capitalization.

AI should clean that up. It should tag tickets by issue type, product, urgency, sentiment, customer type, subscription status, and root cause. It should turn the inbox from “general chaos with search functionality” into something your team can learn from.

No customer sees this or gets offended. Nobody screenshots your bot being weird. Your team just gets better data and fewer tags named “misc-urgent-final-v2.”

Good AI job.

2. AI should summarize messy ticket history

A customer writes in after three previous conversations, two replacements, one refund, and a subscription change nobody fully understands.

The agent opens the ticket and gets to choose between spending seven minutes spelunking through the helpdesk or answering with the confidence of someone who absolutely did not read the file.

AI should solve that.

Give the agent the customer history, previous issues, order timeline, subscription status, prior concessions, and the reason this ticket probably matters. Do not make the human dig through fourteen messages to learn that the brand already messed this up twice.

This is one of the best uses of AI because it makes the rep sharper before they reply. The customer gets a better answer, the agent gets the full picture, and nobody has to pretend that opening six tabs is a customer experience strategy.

3. AI should find patterns across tickets

This is the one I wish more brands cared about.

If ten customers mention leaking caps, someone should know. If five customers say the new formula tastes weird, someone should know. If subscription customers keep saying they have too much product, someone should know before the next retention meeting where everyone pretends churn is mysterious.

AI should be reading across tickets, reviews, surveys, and social comments to flag patterns your team might miss while they are buried in the queue. Broken zippers, melted product, weird smell, missing scoop, late shipments from the same warehouse, confusing onboarding, broken discount code, whatever.

A good CX person can catch some of this manually. They should not have to rely on memory and vibes while 300 tickets are waiting.

Let AI find the smoke and put it in front of the people who know what it means.

4. AI should route work to the right place

A lot of bad CX happens because the ticket takes a scenic tour through the wrong queue before anyone useful sees it.

AI should route billing issues to billing, product-quality issues to the right escalation path, angry subscription issues to the retention team, VIP issues to someone who understands why they are VIP, and low-risk questions to automation.

This will not make a sexy LinkedIn post. It will make the operation better, which is usually more useful.

Good routing means faster replies, cleaner ownership, fewer missed escalations, and less “who owns this?” theater in Slack. It also means your best people spend more time on the tickets that actually need them, instead of being human traffic cops for a helpdesk that looks like it was configured during a fire drill.

5. AI should draft low-risk replies

Drafting is great. Sending is where brands start lighting candles around the helpdesk.

AI should draft the boring replies: order status, return instructions, product FAQs, warranty eligibility, subscription instructions, basic policy explanations. Nobody needs to handwrite “you can start your return here” for the 4,000th time. That is how good agents slowly become houseplants.

The risk level should decide how far the draft goes.

If the customer is calm, the issue is simple, and the answer is low-stakes, let AI help. If the message includes a refund denial, an apology, a dollar amount, a health concern, a damaged gift, a public complaint, or someone who is clearly already angry, do not let the bot auto-send its little masterpiece.

That is how you create screenshots.

Now, the five things I’d keep far away from AI

The simplest rule I’d use: if the moment requires judgment, keep a human close.

1. Refund and return exceptions

If the policy cleanly says yes, automate it.

If the customer is asking for an exception, do not let the bot freestyle.

The policy might say no while the context says obviously yes. A customer outside the return window might be unreasonable, or they might be a great customer with one weird issue your policy did not account for. They might be technically wrong and still worth taking care of.

AI can show the order history, LTV, prior issues, and suggested options. Great. The actual exception should belong to a person until you are extremely confident the logic reflects how your best CX people would make the call.

Most brands are not there. They just have a refund threshold someone invented in a meeting.

2. High-LTV customers

Your top 1% customers should not get the same bot treatment as someone who bought once with a 40% off code and is now trying to return a final-sale item from the Obama administration.

This sounds obvious, and yet plenty of automations do exactly that because the ticket type is the same.

AI should identify the VIP. It should surface customer value, order history, subscription status, prior issues, and anything else that changes how the brand should handle the moment.

It should not quietly decide that your best customer deserves the cheapest workflow because the workflow technically works.

3. Angry customers

AI is very bad at knowing when it is making the situation worse.

It keeps the same tone while the customer gets more annoyed. It apologizes in that weird smooth way where every sentence sounds like it was dipped in customer-service lotion. It says “I understand how frustrating this must be” and then sends a help-center link, which is how brands create screenshots with captions.

Angry customers do not need perfect grammar. They need to feel like someone competent is actually handling the issue.

AI can summarize the problem and suggest next steps. A human should own the reply.

4. Product-quality judgment

AI should flag product-quality issues. It should not be the final judge of whether they matter.

A bot sees the ticket in front of it. A good CX person remembers the last ten and knows when the pattern feels too familiar to ignore.

Let AI group the signals. Let humans decide whether product, ops, quality, or leadership needs to hear about it before Reddit does.

5. Generic cancellation saves

Cancellation is where brands love pretending every problem is a discount problem.

Too much product gets 20% off. Too expensive gets 20% off. Did not see results gets 20% off. Going on vacation somehow also gets 20% off, because apparently every subscription issue can be solved with the same sad coupon.

Too much product needs a pause, skip, or better frequency. Price might need a smaller bundle. Product confusion needs education. Poor fit needs a swap. Travel needs a delay. A customer who is actually done needs a clean cancellation, because making cancellation painful is how you turn “maybe later” into “never again.”

AI can help here when it is designed around the reason, the customer context, and the right save path. It should not treat every cancellation like a discount negotiation with a pulse.

The actual rule

The Extinguisher Rule is the simplest way I can think about this.

AI can help you find the fire. It can read more tickets than your team can read, spot the pattern, summarize the context, route the issue, draft the boring reply, and make the human faster.

When the customer is standing there watching to see if the brand actually cares, keep a person holding the extinguisher.

Not because humans are inherently always magical. Plenty of humans are also bad at this. But your best CX people understand context in a way most AI workflows still do not. They know when the policy answer is technically correct and still expensive. They know when the customer is asking for information versus fairness. They know when a ticket is the first visible sign of something upstream breaking.

Before you buy another AI tool or expand the one you already have, run a simple audit. Pull the last 20 tickets AI resolved end to end. Read them like a customer would. Ask whether you would be comfortable with each exchange being posted publicly. Check whether any of them involved an apology, refund denial, dollar amount, angry customer, VIP, health concern, damaged gift, or product-quality issue.

If more than two make you wince, your deflection rate has probably been eating your judgment while everyone admired the silly little dashboard.

Find the fire with AI. Hand the extinguisher to a person.

That’s it for this week!

Any topics you’d like to see me cover in the future?

Just shoot me a DM or an email!

Cheers,

Eli 💛

If you know me, you know two things take up an unreasonable amount of space in my brain: customer experience and travel.

I’ve written 198 newsletters about the first one. Meanwhile, I’ve been to 40 countries, spent far too much of my life figuring out points, and somehow never really written about the second.

I want to change that.

Every once in a while, I’m going to write about a place we stayed, ate, wandered into, or couldn’t stop thinking about. There will probably be some CX in here because apparently I’m incapable of going on vacation without noticing this stuff, but these are travel stories first.

This is Trip Report #001: Modica.

We got to Modica around 9 pm on Sunday night, which is late enough to arrive at a small hotel and immediately feel guilty about it.

A few hours earlier, Casa Talía had messaged us to say we were going to miss check-in. I apologized, assuming we were about to ruin the evening of some individual working reception.

When we finally pulled up, there was a man in his 60s standing down an alley waving us toward the parking spot.

Polo shirt, chinos, gray hair, small beard. Extremely Italian without making a production out of being extremely Italian.

His name was Marco.

He helped us with our bags and started walking us down the alley toward the hotel, except there wasn’t really a hotel.

There was no lobby or big entrance, and no reception desk with someone waiting to take a credit card for incidentals. Casa Talía is a collection of old little homes built into a hillside overlooking Modica, connected by narrow paths, gardens, stone walls, and lots and lots of stairs.

It reminded me a little of the hidden residential alleys in Jerusalem. You have the feeling that you’ve somehow wandered into somebody else’s neighborhood, except someone has left a very nice bed for you.

We followed Marco to the two-floor house we’d booked for three nights, which cost us 38,000 Hyatt points a night. Cash rates were around €500 a night, and I obviously chose the fake-money option because I have spent an embarrassing portion of my adult life making sure I never have to pay for a hotel.

The house was beautiful. White walls, cave-like ceilings, old stone, restored furniture and almost nothing unnecessary, with a little terrace outside covered in grapevines. You can sit at the table, reach above your head and pull off a bunch of grapes while looking across Modica, which is the sort of thing a hotel could spend an enormous amount of money trying to manufacture and still somehow f$ck up.

It also had a staircase apparently designed for a family with no children and no plans to ever encounter one.

Instead of a banister, ropes hang from the ceiling every foot or so. Architecturally, very cool. As the father of two kids, I have spent considerably less time admiring the architecture.

This was our first introduction to what I ended up loving about Casa Talía. It is beautiful and high-end, but nobody seems particularly interested in sanding away every strange edge until it resembles every other expensive hotel in the world.

And then we found out the guy standing outside at 9 pm waving us into a parking spot was the owner.

Marco Giunta and his wife, Viviana Haddad, were architects living in Milan when they came to Sicily for a summer holiday in 2001. They ended up in Modica almost by accident and stayed for 15 days.

Then they bought a room.

They went back to Milan, returned to Modica, and bought another. Then another. Within eight months they had somehow acquired 12 little properties on the hillside, which is apparently what happens when two architects fall in love with a Sicilian town, and nobody stops them.

Every one of those rooms had originally been somebody’s home. Marco and Viviana restored them, kept the original exterior walls, and connected everything through the garden rather than turning the whole thing into one conventional hotel building.

Which explains why, more than twenty years later, you arrive at Casa Talía and spend the first few minutes trying to figure out where the hotel actually is.

It still feels more like someone’s slightly ridiculous collection of beautiful Sicilian homes than a hotel.

Marco also behaves more like the person whose house you’re staying in than the owner of a fancy boutique property.

He’ll tell you where to eat and, just as confidently, where he wouldn’t bother. He says buongiorno every time he sees you. If you need something, somebody appears. He tells the story of how he ended up here with the enthusiasm of someone who still can’t quite believe this became his life.

And he’s there in the morning.

Breakfast happens outside under olive trees overlooking the old city of Modica. The view is stupid.

The churches and old stone buildings climb up the opposite hillside, framed by branches from the trees above you. Breakfast comes out slowly: bread, yogurt, juice, tomatoes, grilled zucchini, cheese, figs, fruit, jams. Much of it is grown locally or comes from whatever is actually in season nearby.

There isn’t a buffet with 11 kinds of cereal sitting under fluorescent lights. You sit down and eat what Sicily has that morning.

Marco hangs around nearby, talking to guests, checking in on people, and generally holding court.

Then there’s Riz.

I’m not entirely sure what Riz’s official job title is, and after a few days here I’m not convinced Casa Talía believes very strongly in official job titles. He checked us in, serves breakfast, handles the concierge stuff, fixes whatever breaks, and somehow finds time to play with our kids.

He also has an unnerving ability to appear.

At one point, our key wouldn’t open the door to the pool. I sent Riz a WhatsApp, looked back at the door, and maybe 15 seconds later he was standing behind me smiling with a new key.

I have no idea where he came from.

It’s like having a concierge who can apparate, except Hogwarts apparently has worse response times.

That’s basically how it works with Riz. You text him and within 30 seconds there’s a decent chance he’s standing in front of you, smiling, already fixing whatever you were about to ask him to fix.

He is a f$cking legend.

One morning while we were eating breakfast, I noticed that the housekeeping room overlooks the patio.

This means the staff knows when you’ve left your room because they can literally see you sitting under an olive tree eating breakfast. By the time you finish, they’ve usually been upstairs and cleaned it.

I’ve spent a decade around companies buying increasingly complicated technology to figure out where a customer is and what they might need next.

Casa Talía has a window.

There are little things like this everywhere once you start paying attention.

Our terrace has actual grapes hanging over the table. There are fig trees around the property. Old stone paths run between the rooms because the rooms were once separate homes. There are corners of the property where it’s hard to tell where Casa Talía ends and Modica begins.

Very little feels designed around what a hotel is supposed to have. It feels like they started with what was already here and figured out how to let people stay in it.

That applies to the people too.

I’ve stayed at hotels that know an absurd amount about me. They know what floor I prefer, when I last stayed, what status I have, probably what I ordered from room service in 2022. Some of them are incredible. Others manage to know everything about you while making you feel like room 416.

Casa Talía knew almost nothing about us when we arrived. Marco knew we were late, so he waited outside. Riz got to know the kids. Someone saw us eating breakfast, so they cleaned the room.

You start to wonder whether we occasionally make this stuff more complicated than it needs to be.

The WiFi password at Casa Talía is slowliving.

Of course it is.

I initially assumed this was a cute hotel thing, but Marco has been talking about slow living for years. He once described it as forgetting your phone, drinking a glass of wine, looking across at the Baroque buildings and sitting under the olive tree with a book.

There is one problem with this philosophy.

Marco also says he works 24 hours a day.

He said that years ago when asked what running Casa Talía had done to his life. In the same answer, he said his life was much better than it had been before. He gets to live near the sea, see blue sky and sun every day, and spend his time around people who are generally pretty happy to be there.

I love that answer because it makes slowliving slightly harder to explain.

Marco didn’t escape work. If anything, he seems to have created considerably more of it. He just built a life around work he actually wanted to be doing, in a place he didn’t want to leave.

Maybe that’s why nothing here feels particularly slow when you actually need something. When our fuse went out, Riz was there almost immediately. Breakfast appears, rooms get cleaned, bags get carried, kids get entertained.

The slowness seems to be somewhere else.

Most of my professional life has been spent around companies trying to make experiences scalable. You document things, build processes, create roles with clear responsibilities, automate the repetitive parts, and try to make sure the customer gets roughly the same experience regardless of who happens to be working that day.

There are very good reasons for doing all of that. It also removes some of what makes places like Casa Talía special.

If you asked me what the best part of the hotel is, I wouldn’t say the room. The room is beautiful, but I’ve stayed in nicer rooms. I wouldn’t say the pool either. It’s tiny. I wouldn’t even say the view, although the view at breakfast makes it very easy to lose an hour doing absolutely nothing.

I’d probably say Marco.

Which is a terrible answer if your job is figuring out how to scale Casa Talía to 200 locations.

How do you scale the owner standing in an alley at 9pm because a family he has never met is arriving late? How do you operationalize a sixty-something Italian architect telling you where he thinks you should eat dinner?

You could put his recommendations into an app. You could write down exactly what he says when guests arrive. You could create a breakfast SOP, photograph the way the tables are set and train everyone to say buongiorno with precisely the right amount of enthusiasm.

I’m sure you could recreate a lot of it.

I’m not sure you’d end up with Casa Talía.

We spend so much time in business trying to remove dependency on individual people. I understand why. If the company only works because one person is there every day, eventually you have a problem.

Travel has made me wonder whether we occasionally take that idea too far.

Some of the places I remember most are memorable precisely because of one particular person. The restaurant where the owner decides what you should order. The tiny store where someone spends 45 minutes helping you find something that costs $20. The hotel where the guy who owns it is standing outside waiting because he knows you’re arriving late.

Those businesses may be less scalable. They may also be considerably more difficult to forget.

Casa Talía is interesting because Marco isn’t the only one.

Riz isn’t doing an impression of him. He has his own personality and his own relationship with the guests. His job seems to be whatever needs doing at that particular moment.

There’s no obvious attempt to make every person sound or behave exactly the same. They just seem to care about the same place.

Casa Talía doesn’t feel luxurious in the traditional sense. Nobody is polishing marble in the lobby because there is no lobby. The pool is tiny. The stairs remain an ongoing threat to the Weiss bloodline.

But sitting under an olive tree eating tomatoes and cheese while Modica wakes up across the valley, you don’t find yourself wondering what’s missing.

Marco and Viviana started buying these little houses twenty-five years ago. Somehow, after all that time, it still feels like the people here are paying attention.

Maybe that’s what slowliving actually means.

Or maybe I’ve spent too many years working in CX and should be capable of eating a fig without turning it into a business philosophy.

Either way, I’m listening to Marco.

Hi Team!

It’s been a hot minute; hope you are well!

This Saturday night, we head out on Air France business class to Sicily with a layover in CDG, spend the week at hotels that run roughly $700+ a night, take a boat to Malta for two days, then head back to NYC.

ALL WITH POINTS!!

Malta will be country #40, and I genuinely can’t remember the last time I paid cash for a hotel…

If you needed to hear it one more time… go get yourself some signup bonuses and see the world. PLEASE.

Now, back to regularly scheduled programming. In the last few weeks, I’ve seen several CX issues on social handled publicly by the CEO, and it got me thinking.

I’ve heard this story a million times, and even experienced it myself (I am a nightmare customer 🙃)

A customer spends two days arguing with support, gets increasingly pissed, asks for a manager, gets passed around a few times, and eventually lands with someone in the C-suite.

That person reads the thread, apologizes like an actual human, makes a decision, and the customer immediately calms down.

More often than not, the C-level exec does not even offer a discount or a refund; they validate and talk like a human.

So why does this happen? And what can brands do to avoid it?

Let’s get into it.

This week’s newsie is brought to you by my friends at Kim.cc.

Tickets closed by AI have an average re-open rate of 50%+

A closed ticket doesn’t necessarily mean a resolved issue.

When a customer has to come back because the first answer didn’t solve their problem, the brand pays for the same issue twice: once for the AI interaction, and again for the human cleanup.

That makes re-open rate one of the clearest signals of whether support AI is actually working.

The opportunity isn’t to automate more tickets. It’s to resolve more of them the first time.

But resolution requires more than retrieving the right answer from a knowledge base. It requires judgment — knowing how to handle edge cases, exceptions, ambiguity, and situations where the documented answer isn’t enough.

That’s the layer most AI support tools lack.

Kim.cc adds it.

Kim is an AI-native service for ecommerce customer support. AI handles the volume, while Sentinels, trained CX, QA, and prompt experts, implement, train, and continuously improve the system, stepping in where human judgment is needed.

And when a ticket goes wrong, Sentinels don’t just correct that one interaction. They teach the AI how to handle that situation the next time.

That’s how Kim works across 200+ ecommerce brands: continuously improving how tickets are handled so more issues are resolved the first time, and fewer customers have to come back.

Because the real measure of support AI isn’t how many tickets it closes.

It’s how many stay closed.

Check out the Customer Study report

Learn more about Kim.cc

Sometimes they just want you to give a sh!t

I think companies misunderstand what an angry customer is actually asking for more often than we realize.

Sometimes they absolutely want their money back. If the product showed up broken or you charged them for something you shouldn’t have, you certainly should do the right thing.

But plenty of customers aren’t primarily fighting you over $20 or the 20% discount on the “next order.”

They’re fighting because something went wrong, they took the time to explain it, and every response since has made them feel like nobody at the company actually understands why they’re upset.

Someone writes four paragraphs explaining what happened and gets a $5 credit with “we apologize for any inconvenience this may have caused.” Depending on the situation, the $5 almost makes it worse. It’s insulting.

The customer wanted somebody to acknowledge that the experience sucked, and the response feels like you assigned a very small dollar value to making them go away.

Then the thread reaches somebody senior, who writes, “I read through everything, and I’m really sorry we handled this the way we did. I completely understand why you’re frustrated.”

The customer finally feels heard.

Maybe the resolution stays the same. Maybe they don’t get another dollar. Somehow the temperature of the entire conversation changes.

We spend a lot of time training CX teams on rules and policies, and considerably less time teaching people how to recognize what the customer is actually asking for.

Sometimes the right answer is money. Sometimes it’s fixing the product. Sometimes it’s simply having one person at the company say, credibly, “Yeah, I see why you’re pissed.”

We need to hire folks with EQ and let them cook.

You trained everyone to escalate

Escalations are useful when something genuinely requires more authority or expertise. Every company will have situations where the stakes are high enough, the request is unusual enough, or the risk is significant enough that somebody senior should make the call.

The problem starts when escalation becomes the default destination for anything that doesn’t fit perfectly inside the playbook.

An agent gets a request that feels reasonable but falls slightly outside policy, so they ask a manager. The manager has more authority, but the safest decision is still to follow the existing process.

The customer gets another version of the same answer, gets angrier, and by the time someone senior enters the thread they’ve spent two days trying to convince the company that the issue deserves an actual decision.

I’d pay very close attention to what changes once the executive responds.

In a lot of cases, the executive actually has less context than the original agent. They didn’t spend the last day inside the thread, and they probably know less about the day-to-day mechanics of the policy.

They simply feel comfortable deciding that this particular situation deserves an exception, or admitting that the company handled something badly without worrying about whether the wording is approved.

The agent closest to the problem can have the most context while holding the least authority to act on it.

As the issue moves upward, the person making the decision often knows less about the conversation but feels far more comfortable making the call.

That is how you end up with a CEO who looks weirdly brilliant at customer support.

And it should infuriate you.

You threw AI at every problem

AI is genuinely good at a lot of support work now. WISMO, return status, address changes, subscription updates, basic product questions, and all the repetitive work that used to eat huge chunks of a team’s day should increasingly be handled without a human.

I’ve said this in the past… but that changes what the human job looks like.

Once the easiest tickets disappear, a larger share of the conversations reaching your team are messy. Something went wrong twice. The customer already tried the obvious fix. The normal policy doesn’t quite fit. There’s history involved.

The technically correct answer can still make the situation dramatically worse.

Those conversations require better judgment than the average support ticket did a few years ago.

A lot of companies upgraded the technology while leaving everything around the human team basically untouched. They kept the same hiring profile, the same QA process, the same approval tree, the same bullshit macros, and the same fear of somebody making an exception leadership didn’t explicitly approve.

So software takes the questions that barely require judgment, while people inherit more of the conversations where judgment is the entire job.

Then we give those people even less room to use it.

I’m pretty sure we have that backwards.

Your whole CX team should not sound the same

I can’t say I hate a macro or canned response per sè. Nobody should spend their Monday typing the same return instructions 80 times because someone decided handcrafted support emails are more authentic.

But somewhere between “we should have a consistent brand voice” and “we should save agents some typing,” many companies decided that every person on the CX team should sound exactly the same.

I don’t think that’s a good thing.

You hired seven different humans. They have different personalities, different ways of explaining things, different senses of humor, and different instincts for dealing with people. Then you handed all seven the same paragraph and told them not to touch it too much.

Customers can feel that.

If I write three paragraphs about a very specific situation and get a reply that could have been sent to 5,000 other people that morning, I have no evidence anybody actually read what I wrote.

The macro should get the agent 80% of the way there. Let them do the remaining 20% themselves.

There should still be guardrails. I don’t need every agent improvising your warranty policy based on their mood that afternoon.

But there is a huge amount of space between letting everyone say whatever they want and making every person at the company sound like the same damn macro.

AI handling more of the routine work makes this even more important. If a customer has reached a human because the situation is complicated enough to need one, sounding like a human seems like a reasonable place to start.

Sometimes they have permission and still don’t give a sh!t

I also don’t want “empower your team” to become the explanation for every bad CX organization, because I’ve seen brands give people plenty of freedom and still get painfully mediocre support.

The team can make exceptions. There’s budget for surprise and delight. Leadership has explicitly told them to waive fees, send replacements, spend a little extra, and use their judgment when the situation deserves it.

Then you read the tickets, and everyone is still sending the same macro, nobody follows up, and a customer who has ordered 14 times gets treated exactly like someone placing their first order.

At some point, that stops being a permission problem.

Sometimes you hired people who are good at working a queue but don’t naturally care enough to look beyond the ticket in front of them.

The same goes for CX leadership. Someone can keep response times down, maintain the macros, run the weekly meeting, and send you a perfectly nice CSAT report while having almost no POV on whether the actual customer experience feels good.

That matters even more as AI takes over the repetitive work. The human job is increasingly made up of messy situations where the policy won’t tell you exactly what to do.

I’d rather hire someone who is curious enough to understand what happened, cares enough to want to fix it, and has the judgment to know what a reasonable fix looks like.

You can teach someone your policy and give them a bigger budget. Teaching them to actually give a shit is a lot harder.

Your customer has no idea who any of you are

Companies underestimate how absurd their internal processes can feel from the outside.

Inside the company, the ticket moved to Operations because Operations owns that workflow. Finance has to approve something, the manager is out today, the warehouse hasn’t responded yet, and the original agent went off shift.

To the customer, they emailed one company on Monday and have spent the rest of the week explaining the same problem to that company.

They don’t know which department owns replacements, and they definitely don’t care why somebody in Finance needs to approve $60. They want one person who knows what happened and stays attached to it until it’s over.

Executives tend to be good at this because they collapse all of that internal complexity. They take responsibility for the company as a whole and tell the customer what happens next.

Your frontline team should be able to create that same feeling without personally running Finance, Operations, and the warehouse.

Here are the 8 things I’d change

If escalations keep landing with leadership, I’d start here:

• Pull your last 20 to 30 executive escalations and compare the first replies with whatever finally resolved them. Look at what materially changed once someone senior got involved. Sometimes the remedy will be different. In other cases, the biggest change will be tone, clarity, ownership, or somebody finally feeling comfortable making the decision.

• Find the exact response where the customer became meaningfully more upset. Most ugly threads have a moment where the original problem becomes secondary and the customer starts getting angry about how the company is handling it. Figure out what your team said right before that happened.

• Stop automatically reaching for discounts when somebody gets angry. Figure out what they actually want first. If you owe them money, give them the money. If they’re looking for accountability, a tiny coupon in place of a real acknowledgment can feel more insulting than useful.

• Write down what agents can actually do without asking permission. Refund limits, replacements, shipping upgrades, fee waivers, credits, phone calls, and common policy exceptions should have clear boundaries. If your team has to guess where their authority ends, they’re going to send anything remotely unusual upstairs.

• Look for decisions managers approve almost every time. If managers keep rubber-stamping the same type of $50 or $75 exception, move the decision closer to the person who already has the customer context. Your approval history is giving you a pretty good map of authority you no longer need to centralize.

• Let people personalize the macros. Every meaningful response should contain enough of the agent’s own language that the customer knows somebody actually read the conversation. Consistency matters, but if every person on your team sounds interchangeable, you’ve probably taken it too far.

• QA judgment alongside compliance. An agent can follow every policy, use every approved phrase, and still make the customer much angrier. Review whether the response moved the situation forward, whether the agent understood what the customer needed, and whether a reasonable decision earlier in the thread could have prevented the escalation.

• Turn good judgment into training. When an agent handles something beautifully or an executive saves a difficult conversation, break down what worked. Show the team how the person thought about the situation, which details mattered, what they were willing to do, and why. The same type of issue should require less seniority the next time it happens.

I think senior leaders should keep reading customer emails. They should answer some too. There are few better ways to understand what your company actually feels like from the outside.

But customers should feel those standards long before anyone with “Chief” in their title has to enter the thread.

If the apology gets warmer, the ownership gets stronger, and people become much more willing to make a reasonable decision every time another title gets added to the email chain, I’d look pretty hard at what the organization taught the first person who answered.

Your customer probably didn’t need the CEO. They needed somebody who was allowed to solve the problem.

That’s it for this week!

Any topics you’d like to see me cover in the future?

Just shoot me a DM or an email!

Cheers,

Eli 💛

Hi Team!

Welcome to the summer chaos where pickup, drop-off, and camp laundry somehow fill half my spare time.

This is newsletter #197, and while I’ve covered more topics than I can count, there’s one topic I get more emails and consulting requests about than almost anything else:

Increasing subscription LTV.

More brands than ever, especially in CPG and supplements, are pushing subscription as the default and doing whatever little hacks they can come up with to juice recurring revenue. Some of it works, but some of it feels like it was brainstormed in a room where nobody ever tried to cancel anything online.

Most subscription retention problems are expectation problems dressed up as billing problems.

Brands treat retention like a dunning issue or a cancellation flow project. Delay the churn event, add a discount, make the portal a little more annoying, force the customer through three screens and hope they get tired…

By that point, the customer already decided.

They’ve been annoyed for weeks because the product timing, quantity, or frequency stopped matching their life. The shipment shows up and they already have three bottles in the cabinet. Or they ran out two weeks ago and forgot to order. Or they didn’t see results and assumed the product wasn’t for them.

At that point, the retention moment has passed, and the cancellation is more of a formality

This week, I want to talk about five levers that actually move subscription LTV, and why most brands are optimizing the wrong part of the funnel.

Let’s get into it.

This week’s newsie is brought to you by my friends at Hoop!

(Hoop is one of the newest tools I found and one I highly recommend checking out)

Brands pour a huge chunk of their budget into acquiring new subscribers. On average, 50 to 70% of those subscribers churn before month 2.

Stop and consider that for a moment: you paid an expensive CAC for that revenue, and most of it walks out through a cancellation ticket or self serve flow that gets closed with a polite confirmation. You don’t know exactly why each subscriber cancelled, and you didn’t make a reason based offer to help them stay.

So what can you actually do today to keep more of those customers for longer? That’s exactly what Hoop does, out of the box. Built by the team behind Trello, it runs the save conversation on every cancellation:

- Asks why before offering anything: then matches the offer to the specific reason (product education, price, a product swap instead of a cancel, and fully customizable based on your brand)

- Reads full subscriber context: from systems of record like Shopify and Recharge or Skio

- Works inside Gorgias, Zendesk and Kustomer: No new tools, no migration.

MoonBrew, a scaling sleep supplement brand, went live in about a week and hit 24/7 coverage within a month, for roughly the cost of one agent instead of five.

Their team stays on the complex, high-touch conversations while Hoop handles cancellation requests and boosts save rate.

Since launch, not one customer has asked to “talk to a human.”

Want to see how much revenue you could be keeping?

Take the free audit: Hoop reads 90 days of your helpdesk tickets and hands you your baseline save rate, why subscribers actually leave, and the annual revenue you’re leaving on the table.

Get your free audit →

The first purchase doesn’t need to be a subscription

Most brands are obsessed with converting the first purchase into a subscription because it makes the dashboard look clean. Higher attach rate, prettier LTV projection, easier forecasting.

The problem is that forcing a subscription on the first purchase can suppress AOV, attract discount-chasers, and set the wrong expectation from day one. Someone who says yes because you offered 20% off isn’t the same as someone who says yes because they already know they want the product every month.

Sometimes the better long-term customer is the person who buys a larger one-time bundle, uses the product for six weeks, understands the value, and subscribes later with actual intent.

The subscription funnel gets optimized to death while the OTP path is treated like a waiting room. Brands spend all their retention energy on cancel flows and almost none of it on turning high-intent one-time buyers into subscribers after they’ve proven they like the product.

That customer often has better intent and fewer “wait, why am I being charged again?” vibes.

Save offers should match the cancellation reason

A lot of cancellation flows are basically Mad Libs. Customer clicks cancel, brand says, “Sorry to see you go, here is 20% off.”

Doesn’t matter if the reason was price, quantity, timing, product fit, travel, moving, or the customer having six tubs of greens powder in the pantry. Same sad offer every time.

A discount only helps when price is the actual problem. If the customer has too much product, 20% off is just a cheaper way to make the problem worse.

If someone has too much, offer pause or skip. If price is the issue, offer a smaller bundle or lower frequency. If they didn’t like the product, offer a swap. If they’re traveling, delay the shipment. If they didn’t see results, reset expectations or send usage content.

The save offer should solve the reason they’re leaving. Otherwise you’re just bribing them into one more billing cycle they’ll resent.

Frequency is retention

This is probably the most underrated lever in subscription retention, and the one that kills LTV for consumable brands.

Most subscription churn is less of “I hate this brand” and more of “I have too much of this product sitting in my cabinet and I can’t keep up.”

Funny enough, this was our #1 churn reason at OLIPOP way back.

This is huge for supplements, coffee, skincare, pet, household, beverage, pantry, basically anything that gets used up.

If you ship a 30-day supply every 30 days and the customer only uses 22 days’ worth, they’re building backstock after every shipment. After six months, they have a small museum of your product in the cabinet. That’s when they cancel.

Let customers set frequency. Make skip easy. Use first-order behavior to suggest the right cadence. If someone buys a 60-day supply as their first order, please do not default them into a 30-day subscription like a lunatic.

OTP customers aren’t failed subscribers

I saw Olivia Kory from Haus post a clip from Andrew Faris on this, and it put words to something I’ve been seeing in audits.

Subscription customers look amazing in the first 90 days because LTV front-loads nicely. Recurring revenue, predictable orders, easy modeling. The dashboard gets all dressed up and tells everyone the strategy is working.

Then you look six or twelve months out and the gap between subscribers and non-subscribers can get a lot less dramatic.

One-time buyers often have higher AOV on the first purchase because they weren’t chasing a subscription discount. They have lower first-order friction because they didn’t have to opt into recurring billing. And if your replenishment marketing isn’t terrible, they come back.

The mistake is treating subscription as the only retention path and OTP customers as failures.

OTP customers can be retained. They just don’t show up as neatly in the subscription dashboard. You have to work a little harder to bring them back, but the economics can be just as strong, and sometimes cleaner because they’re not locked into a cadence that stops matching their life after three months.

The OTP funnel is under-optimized because subscription makes the spreadsheet feel safe. That doesn’t mean it’s always the better customer experience.

Cancellation is not the first retention moment

Most brands wait until someone opens the cancellation portal to “do retention,” which is completely backwards. By the time a customer clicks that button, they’ve already made the decision. The portal is just where they go to make it official.

The better retention levers happen way earlier. Onboarding sets expectations. Usage education helps them get results. Delivery timing prevents backstock. Failed payment handling catches accidental churn. Support moments build or destroy trust.

All of that happens before cancellation.

Save flows still matter, especially when they respond to the actual reason someone is leaving. The bigger win is making sure the cancellation data changes the business. If 40% of cancellations are “too much product,” the problem is not your save offer. The problem is your default frequency.

By the time the customer is in the cancellation portal, the brand is already playing from behind.

Most subscription brands are optimizing the cancellation moment and ignoring the decisions that happened before it: first purchase path, save logic, frequency setup, OTP retention, and the early customer experience.

That’s why so many subscription programs look good on paper and bleed out in month six.

That’s it for this week!

Any topics you’d like to see me cover in the future?

Just shoot me a DM or an email!

Cheers,

Eli 💛

Hi Team!

It’s been a great week. We made it through the vicious heatwave in NY that made it feel like a European summer, and I restocked points and miles big time as well.

• Chase Sapphire Reserve 150k bonus (current offer)

• Chase Business Ink 100k (current offer)

• Amex Gold 150k (my wife and I both snagged a few months back)

We wiped through 300k points on our St. Kitts and Panama trip, but were able to recoup another 445k since then, with another 100k coming down the pipe :)

Sicily and Malta are booked for August.

Croatia, Montenegro, and Bosnia are booked for October.

The points-and-miles gods are shining on me this year. ☀️

A few weeks back, we talked about segmentation and why most brands are grouping customers by what they bought instead of who they are.

This week, I want to get a bit deeper and more practical: what useful personalization data actually looks like and what a few brands are doing differently when they build around it.

Most brands have gotten very good at using data that doesn’t actually tell them much about the customer. A customer viewed a hoodie, so the hoodie now follows them around the internet like a Victorian ghost.

That’s behavioral recall, and the issue isn’t that the data is wrong. It’s that the data is too thin to build anything meaningful on top of.

Clicks, purchases, abandoned carts, PDP views, discount usage, and maybe a quiz response from 2021 that nobody’s touched since. Helpful signals, but a very shallow foundation.

I want to talk about what doing this better actually looks like.

Let’s get into it.

This week’s newsie is brought to you by OuterSignal!

Their order tells you what they bought. OuterSignal shows you who they are.

OuterSignal tells brands 10x more about their customers than they know right now — age, occupation, property value, social following, family status, interests, education, and more — and uses that to identify VIP customers, build precision segments, and personalize outreach at scale.

Whether it’s flagging celebrities, athletes, investors, podcasters, buyers at large retail chains, or simply surfacing micro-influencers within a target cohort, OuterSignal finds, flags, and activates the customers that matter most with unprecedented precision.

OuterSignal will research 1,000 customers for free, and is offering friends of Eli 50% off their first two months when they upgrade.

Tell them Eli sent you!

Get 50% Off Outer Signal

Shallow signals make shallow personalization

Most brands personalize off whatever data is easiest to grab: viewed product, purchased product, clicked email, added to cart, browsed collection. That data tells you what happened inside your store or email program. It doesn’t tell you much about the customer’s life outside of it.

Your behavioral data sees the same action no matter who’s behind it, so the brand sends everyone the same ‘Still thinking it over?’ email and calls it personalization.

The data is technically accurate. It’s also too thin to make the message feel like it was written for a human.

This is how brands end up with lifecycle programs that look sophisticated in Klaviyo and feel completely generic to the person reading them. The logic tree is impressive.

The message is still basically, “You looked at this. Please buy it.” Very moving stuff.

What Gratsi figured out

I recently came across Gratsi, a DTC wine brand selling boxed wine that does not look like something you’d hide under the sink before guests come over.

They had the same problem most brands have: people were buying, subscribing, clicking, and coming back, but the customer file still didn’t explain much. “Wine buyer” is technically true, but it doesn’t get you very far.

A retiree hosting friends, a health-conscious professional looking for a lower-sugar option, a founder sending client gifts, and a parent trying to get through Friday night without reading 19 tasting notes from a guy named Sebastian can all look identical in Shopify.

Gratsi used OuterSignal (this week’s sponsor) to understand their core personas. OuterSignal enriched their customer data and surfaced different identity-based groups inside the customer base: Active Retirees, Health-Conscious Professionals, Creative Entrepreneurs, and a few others.

Then Gratsi took one campaign and changed the message based on those groups. The send time, audience size, subject line, and product stayed consistent. The email itself changed based on the reason someone might care.

• Active Retirees got messaging around slowing down and enjoying the moment.

• Health-Conscious Professionals got messaging about zero sugar, zero guilt, and feeling better the next morning.

• Creative Entrepreneurs got a different angle entirely.

The persona-matched messaging drove a 47.2% lift in revenue, 54.1% higher click rate, 56.1% more subscription conversions, and 41.7% more orders. Open rates were basically flat, which is the part I care about most. The lift happened after the open, when the message matched the person reading it.

Purchase history is a lazy creative brief

This is where lifecycle teams get stuck. The brief says “past wine buyers,” which could mean dinner party hosts, weekday drinkers, collectors, gift buyers, wellness shoppers, empty nesters, suburban parents, founders, retirees, or someone who bought one box because an ad caught them at 11:43 PM after two glasses of someone else’s wine.

Same thing in apparel. “Bought a dress shirt” tells you almost nothing useful. One customer is an executive wearing it twice a week. Another is a groom buying it for one occasion. Another needs something affordable and wrinkle-free for work travel.

Same SKU, completely different customer, and if those three people get the same email, one might convert, one ignores it, and one unsubscribes because the brand keeps talking like they’re shopping for a life they don’t have.

The product is the shared detail. The reason for buying is where the money is.

Better data makes the copy less lazy

A lot of weak marketing starts before the copywriter ever touches the brief. The brief is boring because the segment is boring, and the segment is boring because the data underneath it is too basic.

The “insight” is that someone purchased Product A, so now the brand would like them to purchase Product B. Groundbreaking stuff. Somewhere, a lifecycle marketer is bravely dragging a product block into Klaviyo.

Identity data gives the creative team something real to work with. A health-conscious professional should get a different wine message than an active retiree. A new parent should get a different apparel message than a founder. A micro-influencer has a different value to the business than a one-time buyer at the same AOV.

This is also why brands end up over-discounting. When you don’t understand the customer, price becomes the easiest lever. So the brand sends 15% off, then 20% off, then a final final final chance email that somehow returns every month like a raccoon in the attic.

Better customer data gives you more to work with: status, use case, lifestyle, timing, identity, actual reason to buy. The discount becomes a tool instead of the whole damn strategy.

That’s it for this week!

Any topics you’d like to see me cover in the future?

Just shoot me a DM or an email!

Cheers,

Eli 💛

Hi Team,

HOW IS IT JULY ALREADY?

Before I jump in, two important things:

1. If you wish you traveled more and want to travel for free this summer, get yourself a Chase Sapphire card and check out seats.aero to start looking at transfer options for your summer vacay. Anyone can do it, GO SEE THE WORLD. I will keep saying this until I am blue in the face. It’s free money.

2. If you have been to Kotor or Dubrovnik lately and have suggestions on where to stay, please let me know. Debating an October sprint with the kiddos. :)

Now for this week, I want to dive into maybe the hottest and most unnerving topic discussed here, which will certainly get me in trouble.

AI in CX.

Something I’ve been thinking about lately:

I still have yet to meet a single customer who told me they loved an AI support experience.

Every brand I talk to is implementing AI all over their CX stack. Your efficiency metrics look incredible on paper. Response times are down, handle times are shrinking, and your team can process way more volume than they could a year ago.

But nobody is talking about the fact that your customers hate a lot of it, and according to data from this week’s sponsor, >50% of bot-resolved tickets get reopened.

Not all of it and maybe not always, but enough of it that a meaningful chunk of them are one bad bot loop away from abandoning your brand entirely.

They are not even waiting long enough for the CSAT survey to hit, they already slammed the phone down and are jogging to your competitor instead.

This week, I want to talk about where AI actually works in CX and where it actively destroys trust, brought to you by real data.

Let’s get into it.

This week’s newsie is brought to you by Kim.cc.

Nearly 60% of consumers welcome AI for simple tasks, but demand a human when the stakes are higher.

Everyone keeps saying customers want AI support.

Turns out that’s only half true.

Kim.cc surveyed 1,000+ U.S. consumers to understand what people actually think about AI customer service.

Some findings were baffling:

• 50% blame company leadership, not the AI, when a chatbot makes an error.

• Over one-third have already abandoned a brand because of a bad AI support experience.

• Nearly half repeatedly type “agent” just to escape the bot.

The takeaway isn’t “don’t use AI.”

It’s that customers want AI for simple questions, and humans the second things get complicated.

Kim.cc is built around that exact idea: AI customer support with a human eye.

Through their Sentinel service, AI handles the work, while trained experts supervise the system - reviewing quality, handling exceptions, and continuously training the AI. They work with 200+ Shopify brands to help them move faster without turning support into a locked room with a chatbot in it.

If you’re building customer support for your Shopify store, this report is worth reading.

Check Out The Report

Learn More about Kim.cc

I Watched Brands Get This Backwards

I’ve started using two questions before letting AI touch anything, because I’ve seen enough brands roll out AI support, celebrate the efficiency gains, and then wake up a few weeks later to reviews about how impossible it is to get help.

The two questions:

1. Can the AI execute this task flawlessly?

2. Are the stakes low enough that if it goes slightly wrong, the customer doesn’t lose money, miss something important, or feel ignored during a crisis?

If the answer to both is yes, AI is probably fine. If the answer to either is no, you need a human close by.

Simple tracking requests, basic FAQs, store hours, return policy lookups. These are low-stakes, high-repetition tasks where speed actually matters and accuracy is mostly binary. Nobody needs to wait 18 hours for a human to say their package is still in New Jersey.

That’s where AI works.

According to Kim.cc’s survey, 44.5% of consumers welcome AI for simple tracking, and another 42% accept or tolerate it. For basic info like store hours, return policies, or FAQs, 42.9% welcome AI and 43.2% accept or tolerate it. Pretty clear green light.

But the second the issue gets expensive, emotional, or messy, people want a human. Only 15.9% welcome AI for financial transactions like refunds, billing disputes, or plan upgrades. 51.7% demand a human. For high-emotion issues like a terrible experience or flight cancellation, 59.5% demand a human.

Essentially, the lower the stakes, the higher the tolerance. The higher the stakes, the faster people bail. I know this might sound obvious to most of you all, but then half the brands I talk to are “99% AI,” so apparently we still need to say it.

This is where brands keep getting themselves into trouble. They see AI work for “Where is my order?” and somehow decide it should sit in front of every customer issue.

That is like saying because a microwave is good enough for leftovers, it should handle your wedding catering. Useful tool, wrong damn job.

The Stakes Are The Strategy

A tracking question and a refund dispute both count as one ticket. A return policy lookup and a damaged item escalation both count as one ticket. A subscription skip and an angry cancellation after three bad deliveries both count as one ticket.

But they are not the same thing. Some tickets are low-stakes and annoying. Others are trust moments.

If I call Delta to change my seat and an AI bot can move me from a middle seat to an aisle seat in 30 seconds, incredible. Give me the bot 10/10 times. I do not need a human, I do not need fake empathy, and I sure as hell do not need someone to say, “I understand how frustrating middle seats can be.”

Just move me away from 26B and let us both continue our lives.

But if I have an emergency, my flight leaves in 30 minutes, and I need to change the name on a ticket or get rebooked before I miss something important, please do not make me explain my life to a bot that can only offer three buttons and a cheerful apology.

Same airline, same general support category, completely different stakes.

That’s the part most AI CX rollouts miss. The better question: what happens if AI gets it wrong?

If your AI screws up a tracking update, the customer is mildly annoyed. If your AI screws up a billing dispute or gives wrong information about a return policy, the customer loses money or time, and now they’re gone.

The survey found that when AI gives bad information that costs customers time or money, 50.3% blame company leadership for deploying an unready tool. Only 17.7% blame the AI itself.

Customers see bad AI support as a deliberate cost-cutting decision. Once they see it that way, the damage compounds. They don’t think the bot is broken. They think the brand is cheap.

The Gatekeeper Problem

The Kim.cc survey calls one of the worst AI behaviors “The Gatekeeper”: refusing to give customers an option or button to speak to a human.

It is brutal because it is exactly what customers hate, and exactly what many brands are optimizing for. One of my least favorite metrics in CX is “deflection rate.”

Imagine telling a customer the goal of your support bot is to deflect as many people with issues as possible from getting to their ideal resolution in a painless manner.

Sounds less sexy on the vendor slide.

Most AI is designed to solve the issue before escalating. On a slide, that sounds reasonable. Why route to a human if the bot can handle it?

That assumes the bot actually can handle it. More importantly, it assumes the customer trusts the bot to handle it.

Those are different things.

When the bot misses once, maybe the customer rephrases. When it misses twice, the customer starts getting annoyed. By the third attempt, the customer has usually decided two things: the bot is useless, and the brand is making it hard to get help on purpose.

According to the survey, 47.6% of consumers have typed “agent,” “human,” or “representative” repeatedly to bypass the bot.

That is a damn revolt.

A customer should not have to trick your support system to reach someone who can make a decision. They should not need to type nonsense, threaten to cancel, pretend they want to buy something expensive, or call a separate phone number because the chat experience turned into a hostage situation.

It is an obstacle course with a CSAT survey at the end.

If your AI cannot solve the issue in two prompts, it should offer a human. Not after five attempts. Not after the customer starts typing in all caps. Not after they threaten to report you to the Better Business Bureau like it is 2006.

Two prompts.

Fake Empathy Makes It Worse

Customers hate fake empathy.

When a bot says, “I’m so sorry to hear that, I understand how frustrating that must be,” 32.3% of consumers say they find it manipulative or annoying because a machine cannot actually feel empathy. Another 17% say it actively angers them because it feels like a fake delay tactic. Only 18% find it comforting.

So roughly half of customers are actively put off by the exact language many brands are adding to make AI feel warmer.

AI cannot feel empathy, and customers know this. Pretending otherwise does not make the interaction better. It makes it feel dishonest.

If the bot is going to apologize, it should apologize for the delay or inconvenience and then immediately solve the problem.

“I’m sorry this is taking longer than expected. Let me pull up your order details now.”

That works because it is transactional and honest.

“I totally understand how frustrating this must be for you” does not work because the customer knows the bot does not understand anything. It is reading a script.

The brands that do this well do not try to make AI sound human. They make it sound fast and accurate.

The bot is a tool, not a therapist. Let it be a good tool.

What Actually Works

If the task is simple, repetitive, and low-risk, AI is fine. Order tracking, FAQ lookups, store hours, basic account changes, return policy questions, address changes before fulfillment. Let the bot handle it. Speed actually wins there.

When the task involves money, emotion, urgency, or any level of interpretation, route to a human quickly. Don’t make the customer fight for it. Don’t make them type “agent” five times. Build the escape hatch into the first or second prompt, and make it obvious.

Stop trying to make your AI sound empathetic. It’s a bot. Let it be fast, accurate, and honest about what it is.

Solve the problem or get out of the way. That’s the whole job.

Right now, too many brands are using AI like a blanket solution, and the data is screaming that it’s backfiring.

Your efficiency metrics look great. Your customer sentiment does not.

Fix the framework before you lose more people.

Again, highly suggest checking out the report from Kim.cc

Check It Out

That’s it for this week!

Any topics you’d like to see me cover in the future?

Just shoot me a DM or an email!

Cheers,

Eli 💛

Hi Team!

Spent a few days in Panama end of May, and 10 days in Israel at the beginning of June. Got hit with jet lag, a stomach bug, and the strong willingness to be a proper homebody. (at least until the Toronto trip in July and the Sicily/Malta trip in August 🙃)

Something that has come up a lot in conversations over the last few weeks with brands I’m chatting with is segmentation and its importance in overall retention efforts.

Most DTC brands have gotten really good at segmenting customers by order history. You’ve got your one-time buyers, your repeat customers, your high AOV spenders, maybe a VIP tier if you bought more than three times or crossed some dollar threshold.

The problem is that none of those segments tell you anything about the person. A C-suite executive and a college student can both be two-time purchasers who spent $80. Your segments treat them the same because your data only knows what they did, not who they are.

So you send them the same email, the same ad, the same offer, and one converts while the other churns.

This week, I want to talk about why segmenting on purchase behavior is a weak foundation for retention and growth, and why the brands pulling ahead are segmenting on identity instead: occupation, lifestyle, life stage, the stuff that actually explains why someone bought in the first place.

Let’s get into it.

This week’s newsie is brought to you by one of the coolest new tools I’ve personally ever seen and one that explicitly solves the problem we’re talking about today. Meet OuterSignal.

OuterSignal helps brands transform basic order data into fully researched, actionable customer profiles.

Right now, all you get is a name, email, and address. Imagine if every order also had age, gender, occupation, social following, property value, and 100+ other signals.

With OuterSignal, you can:

1. Segment your existing customers by data-driven personas – including occupation, life stage, lifestyle, and buying motivations– allowing you to personalize email, SMS, and customer experience to each persona

2. Know why each customer bought, so you can predict who’ll repeat, win back the right churned customers, and stop blasting everyone the same offer.

3. Spot VIPs in your customer base (celebrities, athletes, investors, influencers, buyers), and treat them like it.

Leading brands like Create, Jolie, Magic Mind, and Mizzen+Main trust OuterSignal to power their customer personalization and increase their ROAS by up to 160x.

OuterSignal will research 1,000 customers for free, and is offering friends of Eli 50% off their first two months when they upgrade.

Tell them Eli sent you!

Learn More

Your segments are SKU buckets wearing a costume

Most brands segment like this: anyone who bought Product A goes in Bucket A, anyone who bought twice goes in the repeat bucket, anyone who spent over $200 goes in the high-value bucket. You might dress it up with a name like “Core Enthusiast” or “Premium Loyalist,” but under the hood it’s still a filter on order count or SKU.

But something most retention folks learn over time is that buying twice doesn’t tell you if someone’s about to buy a third time or ghost you forever. It tells you what someone did after they decided to buy, but it doesn’t tell you why they bought, what problem the product solves in their life, or whether they’ll buy again.

A menswear brand could have 50,000 customers who each bought a dress shirt, and some of them are executives who wear it to board meetings, some are grooms who need it for a wedding, and some are retail workers who need something cheap and wrinkle-free for their shift.

Same SKU, totally different reasons for buying, completely different retention curves. Messaging all of them the same way means you’re treating the person who needs you every week the same as the person who needed you once.

The brands doing this differently segment on who people are

Mizzen+Main sells performance menswear. They could segment on repeat buyers or AOV, and they did for years. But when they enriched over 100,000 historical customer records with actual identity data, they found something better: a massive cluster of C-suite executives who had purchased in the past but weren’t responding to broad campaigns.

They built an audience around that profession, that lifestyle, that identity, and ran it on direct mail over BFCM. It became their highest-performing campaign in company history. Over 20x ROAS, $50 AOV lift above baseline, their number one campaign for the entire peak season. Same product. Same platform. Different understanding of who was buying it.

Gratsi, the DTC Italian wine brand, ran a controlled A/B test on this exact question. They took a single email campaign and split it: half the list got a generic message, the other half got persona-matched creative based on identity data like Active Retiree, Mindful Achiever, Creative Entrepreneur, Health-Conscious Professional. Same subject line, same send time, same list size. Open rates were nearly identical, which means the audiences were consistent.

What changed was what happened after the open. The persona-matched emails drove 47% more revenue, 54% higher click rates, 56% more subscription conversions. The generic email said “It’s Always a Good Time for Gratsi.” The retiree email said “You’ve Earned the Right to Slow Down.” The health-conscious professional got “Zero Sugar. Zero Guilt. Zero Hangovers.” Same wine. Same list. Just talked to them like humans instead of order IDs.

Identity explains behavior, behavior doesn’t explain identity

If you only know someone bought your product twice, you have no idea if they’re about to buy a third time or if they’re done. You’re flying blind. If you know they’re a new parent, a remote worker, a retiree, a competitive athlete, suddenly you have context. You know what problem your product solves in their life, how to talk to them, what else they might need, when to expect them to come back.

A beverage brand might see that someone subscribed and then canceled after two months. If you only have purchase data, that person looks like a churn risk or a lost subscriber. If you know they’re a shift worker who drinks your product before early morning warehouse shifts, you can message them differently than the yoga instructor who drinks it pre-class. One of them churned because the flavor got boring, the other churned because their schedule changed. You can win back one of those people, but you need to know which is which.

You already have the list, you don’t know who’s on it

Magic Mind, the productivity drink brand, installed identity enrichment on their customer base and found out that Kim K had been a customer for nearly two years. No partnership or sponsorship. She had just been buying organically. They sent a care package, she posted it to her Instagram story, hundreds of millions of impressions. The going rate for a sponsored post from her account is in the millions. They got it for the cost of a care package because they knew who was already buying.

Every brand has some version of this. You have customers who are executives, influencers, journalists, investors, people with audiences or purchasing power or strategic value that goes way beyond their LTV. But if you’re only looking at order data, they’re invisible. They look like every other two-time buyer who spent $60, and you’re sending them the same abandoned cart email as everyone else.

Behavior is the what, identity is the why

If you’re only segmenting on the what, you’re optimizing for a pattern you don’t actually understand. You’re building retention programs and acquisition audiences on top of a foundation that can’t tell you why someone bought in the first place or whether they’ll buy again.

The brands winning on this are enriching their customer data with occupation, lifestyle, life stage, and household attributes. They’re segmenting their email lists by persona, not by product. They’re building lookalike audiences based on who their best customers are as people, not what they happened to purchase. And the results aren’t incremental: Gratsi saw 47% more revenue from the same list, Mizzen+Main’s executive audience did 20x ROAS and became their top audience over the biggest shopping weekend of the year, Jolie’s identity-based lookalike drove 412% more sales than the average of everything else they were running.

You can keep segmenting on order count and SKU mix. Most brands are still doing that because it’s easier and the data’s already in Shopify. But easier doesn’t mean better. And the gap between the brands segmenting on purchases and the brands segmenting on people is getting wider every quarter.

Your customers are not their order history. They’re people with jobs, routines, problems your product solves in ways you probably don’t fully understand yet. Segment like it.

That’s it for this week!

Any topics you’d like to see me cover in the future?

Just shoot me a DM or an email!

Cheers,

Eli 💛

Hi Team!

This newsie comes to you from a quiet corner in Yotpo’s new office in Ramat Gan, Israel.

The 10-hour flight with two little kids was exactly as chaotic as you’d imagine. Ari somehow managed to sleep less on the plane than his nap at home, which I didn’t realize was mathematically possible.

On the bright side, we’re here, caffeinated, and I’ve already consumed an irresponsible amount of delicious coffee.

Now onto this week’s topic.

Last week, I reviewed a handful of retention programs across brands doing anywhere from $10M to $100M+ annually.

Different categories, teams, and growth stages.

Same problem.

Everyone obsesses over the traffic spike. Almost nobody obsesses over what happens after.

A July 4th sale gets weeks of planning. Teams debate offers, paid budgets increase, creative gets refreshed, and landing pages get rebuilt. Everyone focuses on maximizing the five days the promotion is live. Then the sale ends, and everyone moves on.

Strange when you think about it.

If 100,000 shoppers visit your site during a sale and 97% don’t buy, that’s 97,000 people who raised their hand and showed interest. Most brands spend all of their energy generating that attention and very little energy recovering it. That approach is backwards.

The best retention programs I’ve seen treat the sale as the acquisition window. The revenue window is what comes next.

This week, I want to talk about how brands extend a five-day promotion into two or three weeks of additional revenue, and why July 4th is the last major opportunity to get this right before back-to-school and Labor Day.

Let’s get into it.

This week’s newsie is brought to you by Instant.

Here’s the part of a sale almost nobody plans for: the two weeks after it ends.

Picture a shopper who lands on your site during the July 4th sale on a Friday afternoon. She looks at the linen dress, clicks into the sage green, reads a few reviews, adds it to her cart, then gets pulled away and leaves.

The sale ends Monday. She still hasn’t bought. In most programs she now gets the same “you left something behind” email as the person who bounced off a TikTok link nine seconds in.

That’s the gap.

Instant identifies the shopper, remembers it was the sage green and not the black, knows she read the reviews and got close, and sends a follow-up that reflects all of it: the right product, the subject line most likely to get her to open, sent at the hour she actually tends to buy, with an offer only if she needs one.

If she ignores the first email, the next one changes. Maybe it leads with reviews. Maybe it handles fit. Maybe it shows the matching piece. Maybe it saves the incentive for later instead of training her to wait for 15% off.

That’s the difference between a flow that fires and a flow that sells. It’s also why Instant works on both halves of a sale: campaigns create the peak, flows monetize the tail.

On campaigns, that means you stop building ten versions of the same promo by hand. You build one multi-step campaign, let Smart Rules swap the content by shopper, and let Instant write a unique subject line for every single recipient.

The proof is hard to ignore. SKATIE went live in mid-February and crossed $1M in incremental revenue from Instant in four months. They hit 61x ROI in month one, 141x in month three, and are now at 180x in June and still climbing. Open rates across their flows are up about 12% and click rates nearly 18%.

The time savings are real, too. Their lifecycle manager used to spend roughly two hours a day building one email. Now they can build a week of campaigns in about 30 minutes. That’s why some of the fastest-growing brands like David Protein, Naturium, Liquid IV, Jenny Bird, Third Love and TRX made the switch to Instant.

If you’re a Shopify brand doing $300K+ in monthly site revenue, book a demo by June 21st. Instant will get you live before your July 4th sale, and you’ll get a $100 gift card of your choice for showing up.

BOOK A DEMO + GET A $100 GIFT CARD

Most Brands Stop Selling Too Early

One thing I’ve learned after years in retention is that teams consistently underestimate how much revenue is sitting inside existing demand.

The instinct is always to chase more traffic. More Meta spend, more influencers, more creators, more campaigns. But when I audit retention programs, I rarely find brands that have exhausted the opportunity already sitting in front of them.

A shopper visits three product pages, comes back twice, adds a product to cart, and leaves. That’s a follow-up problem, not a traffic problem.

The irony is that recovering interested shoppers is often far cheaper than acquiring brand-new ones. You’ve already paid for the click, earned the attention, and the shopper has already told you what they’re interested in. Yet many brands treat the interaction as over the moment the session ends. The best operators don’t.

Your Best Sale Revenue Often Comes After The Sale

Most teams think about sales in terms of a start date and an end date. Customers don’t.

A shopper who visits on July 3rd and buys on July 10th still came because of your July 4th promotion. The calendar changed, but unlikely that the intent did.

Here’s what nobody talks about: most core flows stop at 1-3 emails, which is why they go quiet right after the sale ends. Extend those same flows to 5 touches, gated on engagement so only the people who haven’t converted keep hearing from you, and the flow keeps working for another two to three weeks.

And the cost of doing that is almost nothing. An email costs cents to send, and the revenue per recipient across a strong flow can run $5 to $25 over its life. So the downside of a few more well-timed, personalized touches is rounding error, and the upside is no longer having to drive 90% of your revenue in the first 5 days of the month and anxiously weighing a summer savings push late July to save your number.

The brands that understand this don’t measure success exclusively by sale-week revenue. They measure how much of the demand they captured over the entire lifecycle of the event. That’s a meaningful difference in how you staff your team, how you build your flows, and how you think about what success actually looks like.

The Traffic Is Expensive. The Follow-Up Is Cheap

This is where many brands leave serious money on the table.

I still see companies spending hundreds of thousands of dollars driving traffic into experiences supported by abandonment flows that haven’t been touched in years. Same subject lines, timing, creative, and message for every shopper.

Think about what that actually is: you are renting back attention you already own.There’s nothing wrong with retargeting, but if someone already visited your site, browsed and added to cart, email and SMS should be your highest ROI channels for bringing them back. The intent is already there. You’re capturing demand that already exists, not creating it from scratch.

Every shopper your flows recover is a shopper your paid team doesn’t have to buy twice.

Why AI Is Suddenly Making This Easier

If extending flows and personalizing every touch is so obviously worth it, why doesn’t everyone do it? Because by hand it’s brutal.

Doing this properly means building dozens of flows, each with variants for different behaviors and segments, writing the copy, choosing the product, setting the timing, and then maintaining all of it as your catalog and pricing change. That’s the work that lives on every retention team’s roadmap and never gets done.

That’s why I’ve been paying attention to what companies like Instant (this week’s sponsor) are doing. The interesting part is that the system is continuously adapting based on actual shopper behavior: it writes the copy, pulls the exact product the shopper looked at from your live catalog, sets the send time per person, and runs more A/B tests in a month than most teams run in a year.

A shopper who viewed the same product page three times and added it to cart doesn’t need the same discount-heavy email as someone who bounced immediately. The high-intent shopper might convert with a softer nudge or even a reminder. The low-intent shopper might need social proof or a limited-time offer. AI can make those distinctions automatically instead of requiring a retention manager to manually build and maintain dozens of segments. AI makes those calls automatically, so that you can spend your time on where curated human attention actually moves the needle.

That shift matters because it means brands can finally personalize at scale without requiring massive teams or complex workflows, across both flows and campaigns. The follow-up becomes smarter, the conversion rates improve, and the cost per recovery drops.

What This Looks Like In Practice

A beverage brand I talked to recently ran a Memorial Day promotion that generated roughly $800K in revenue during the five-day sale window. Strong performance by any measure.

But when they looked at the full two weeks following the sale, they realized their abandonment flows and post-sale follow-up had generated an additional $300K from shoppers who visited during the promotion but didn’t buy immediately. That’s nearly 40% more revenue from the same traffic, with almost no additional acquisition cost.

The difference was that their follow-up was personalized based on actual browsing and engagement behavior, not generic cart reminders sent to everyone who abandoned.

Another apparel brand treated the window after their sale as part of the same event and extended a five-day promotion into nearly three weeks of selling. Shoppers who didn’t convert during the sale got follow up tailored to what they viewed, how many times they came back, and whether they’d engaged with earlier emails. Total event revenue came in more than 50% higher than the sale week alone.

These aren’t outliers. This is what happens when you treat the sale as the beginning of the revenue opportunity instead of the end.

July 4th Is Your Last Big Test Before Labor Day

Most retention teams are already thinking about the back half of the year. Back-to-school, Labor Day, Black Friday, Cyber Monday. The biggest revenue periods on the calendar

July 4th is your last real rehearsal before all of it. It’s your chance to find out whether your flows, your personalization, and your post-sale follow up can actually carry demand past the promotion window, while the stakes are still relatively low.

So this is the week to fix it. If your flows haven’t been updated in six months or longer, update them. If your core flows stop at 1-3 emails, extend them and gate the later touches on engagement. If you’re still sending the same generic cart abandonment email to every shopper regardless of their behavior, add the nuance. If you’re measuring success purely by sale-week revenue and ignoring what happens in the two weeks after, now is the time to expand your reporting.

The traffic you generate during July 4th is expensive. The follow-up is cheap. The brands that win are the ones that treat both as part of the same system.

That’s it for this week!

Any topics you’d like to see me cover in the future?

Just shoot me a DM or an email!

Cheers,

Eli 💛

Hi Team!

Sorry for the brief absence. After 3+ years of rarely missing a week, I needed a short break.

Naturally, I used that break in the least restful way possible: spending an ungodly amount of time in Claude Code and Codex, dragging my life into the future.

I vibe-coded a bunch of tools for work, tinkered on too many side quests, and somehow found the brain capacity to finally redesign the newsie.

I brought more of my travel obsession into the look and feel, which makes the whole thing feel a bit more Eli-centric, including at least one boarding pass reference nobody asked for.

Outside of that, we celebrated Noah’s 5th birthday with a bouncy house, pizza, and a superhero theme, which is basically the holy trinity of five-year-old joy.

We’re also taking a last-minute points trip to Panama, 30k round trip on United via Air Canada points, and then prepping for 10 days in TLV in June.

On the work side, we launched Yotpo Discover, and we’ve been working on a few new ideas I’ll share here soon.

This week, we’re talking about why “having flows” is not the same thing as having a real retention program, why most brands are missing high-intent shoppers, and how to think about flows as a revenue channel instead of a checkbox.

Let’s get into it.

This week’s newsie is brought to you by Instant.

Most email flows were built for a world where personalization meant adding a first name and maybe showing the product someone left behind.

That bar is too low now.

A shopper who spent 20 minutes across three sessions looking at the same product should not get the same follow-up as someone who bounced after nine seconds from a TikTok click.

Instant helps brands fix this with AI-powered flows built around what shoppers actually do on site. It identifies shoppers, tracks the products they viewed, the variants they came back to, the carts they abandoned, and how close they were to buying. Then it sends personalized flows based on that behavior.

The product in the email matches what they were actually looking at. The timing reflects when that specific shopper is most likely to act. The offer is calibrated based on whether they need a nudge or would likely buy at full price.

The results are hard to ignore. Karen Kane added $1.1M in incremental revenue in 90 days. The Collagen Co. drove $10.6M. Linda’s saw a 109x ROI. July drove $3.9M. Nakie drove $2.7M.

Instant is quick to start with, easy to expand, and built to handle work your team never gets to. Most brands go live in under 30 minutes and see the lift within days.

It is as easy to turn Instant email flows on as it is to pause them. No long-term contract, no drawn out setup, and no extra lift from your team.

Book a quick demo with Instant before Memorial Day and get 50% off your first 60 days.

GET 50% OFF YOUR FIRST 60 DAYS

The “we have flows” problem

Ask a retention team about their email program and you will usually hear the same few things. Welcome sequence is live. Cart abandonment is live. Post-purchase is live. There is probably a winback flow somewhere in the account, though nobody is completely sure when it was last updated or whether it still references a product that got discontinued.

The team feels good about it because someone did the actual work to build those flows, and the program really is better than whatever came before it. Both things can be true. The program can be better than it was and still not be good enough.

A flow can be live and still miss most of the intent happening on your site. It can drive revenue and still underperform badly. It can look perfectly healthy in a dashboard while leaving money behind because the message, timing, offer, and product recommendation are too generic.

I once looked at a brand where more than half of email revenue was coming from the welcome flow. The team was proud of that number, but it should have made everyone uncomfortable. If the welcome flow is carrying the whole program, the rest of the lifecycle is probably underbuilt: browse abandonment is weak, winback is a blanket discount, price-drop isn’t running, and post-purchase is either too generic or pushy.

The brand thinks email is working because one room in the house looks nice. The rest of the house is empty.

Most brands stop at the obvious flows

Welcome, cart, checkout, post-purchase. Those are the basics, and that is also where most brands stop.

Customer behavior is more specific than those four buckets. Someone may view the same SKU five times. Someone may buy from one category and repeatedly browse another. Someone may purchase three times, then slowly go quiet. Those are all signals, and most programs treat them like noise.

What you end up with is a whole layer of mid-intent shoppers who are warmer than cold prospects but not close enough to convert on their own. The program has nothing thoughtful built for them.

A good retention program responds to meaningful behavior across the whole shopping journey. If a previously active customer starts going quiet, the winback should reflect what they used to buy and whether they are worth protecting margin for. None of this is new in concept. Every retention team has some version of it on a roadmap. The reason it stays on the roadmap is that building it properly is a pain. Each flow needs logic, copy, creative, segmentation, and ongoing maintenance. That overhead is what kills execution.

Campaigns build context. Flows close the loop.

Campaigns and flows should not have the same job.

Campaigns are for brand moments: product launches, seasonal pushes, editorial notes, big promos. They go out because the brand has something to say. Flows fire because the customer did something: viewed a product, browsed a category, abandoned a cart, stopped buying.

That means the message should be much closer to the behavior. A browse abandonment email does not need to open with a brand manifesto. The shopper was just on the PDP. They need the product they viewed, a clean reason to reconsider it, and an easy path back. A cart abandonment email does not need three lifestyle modules. Maybe they need shipping clarity. Maybe they need reviews. The job is to close the loop, not tell the brand story again.

Too many brands treat flows like smaller campaigns. They make the creative prettier, soften the message, add more brand language, and end up with something that looks nice but does not do the job. Pretty is nice. Useful converts.

Discounts are covering for weak relevance

A lot of brands use discounts to cover for lazy timing and generic follow-up. Someone abandons cart, so they get 10% off. Someone goes quiet for 60 days, so they get 20% off. Someone views a product twice, so they get routed into the same offer path as everyone else who did not convert.

This works just well enough to become dangerous. The dashboard shows revenue, the team feels validated, and nobody asks how many of those customers would have bought anyway, or whether what they actually needed was better timing, social proof, or simply the exact variant they had already shown interest in.

Basically, you trained your best customers to wait you out.

Lazy discounting erodes margin on conversions that were already going to happen, and once you set that expectation across your list, it is expensive to walk back. A loyal customer who buys every six weeks may not need a coupon on day 42. A first-time shopper who came back to the same PDP three times may need proof more than they need 10% off. When every shopper gets the same offer, you give away margin to people who did not need it and still fail to convert people whose actual objection was something else. Discounts are not bad. Lazy discounts are bad.

Open rates are not the scoreboard

Most retention teams are looking at the wrong numbers and calling it a strategy review.

Open rate has been unreliable since Apple’s privacy changes. In a lot of cases you are counting server pings as proof of engagement. CTR is more useful but still does not tell you whether the program is improving customer value or just catching shoppers who were going to buy anyway.

The better number is revenue per recipient, not per send, per person on the list. If it is growing month over month, your program is compounding. If it is flat or declining, something structural is broken and no subject line test is going to fix it.

The other number worth tracking honestly is incremental revenue: what you generated because the email was sent to someone who would not have converted without it. Email platforms are very good at taking credit. The actual lift is usually smaller than the dashboard suggests, and the programs that hold up under honest analysis are the ones where behavioral personalization is driving real additional conversions, not just getting credit for ones that were already happening.

What to audit this week

Start with the percentage of site revenue coming from email. If it is under 35% across campaigns and flows, your retention setup is missing the mark.

Then pull your email revenue breakdown by flow. If more than half is coming from the welcome series, you do not have a full lifecycle program. That means the same gaps are hiding in plain sight: the browse flow is not using the exact SKU, variant, reviews, or collection someone cared about. Cart and checkout are stopping too early instead of running a real sequence with smarter timing, proof, shipping clarity, and offer logic. Winback is just a blanket discount. Post purchase is not creating the next order.

Most brands will find the same thing: the foundation exists, but the coverage is thin.

Once you see it, the work becomes clearer. If someone looked at the same product three times, they should not get a generic “you left something behind.” If someone already bought the item, they should not get the new customer offer. If someone buys every 45 days, they should not get a panic discount on day 12. If someone is one click away from buying, do not bury the product under a paragraph about your mission.

Your customers are already telling you what they care about through their behavior. Most brands are only listening to the loudest signals. The brands that win will be the ones that hear the quieter ones, too.

The truth is, you probably already know this. These improvements are on your roadmap, but they keep getting pushed back. That’s where AI changes email and why you should take the leap: it gives your team the bandwidth to finally execute.

That’s it for this week!

Any topics you’d like to see me cover in the future?

Just shoot me a DM or an email! Cheers, Eli 💛