The fundamentals of DTC growth, from the first ad to the fifth order.
Most DTC brands don’t fail at one thing. They do eight things at about sixty percent, and the misses stack: a creative that wins cheap clicks, a page that loses them, a welcome series stuck in draft, a list that shrinks every month, and a dashboard that says it’s all working.
Those two numbers are the book in miniature. The first says that an ordinary estimate of what ads did and what an experiment showed they caused can be wildly different, at a company with some of the best analysts in the business. The second says that after you pay for the click, seven in ten shoppers who reach a cart still leave. Growth is the whole machine: what you pay to get someone in, what you keep when they buy, and whether they come back. A brand can be excellent at any one of those and still lose money.
This is the book I wish I’d been handed on my first day running growth for a consumer brand. It covers the fundamentals, in the order they compound: the numbers first, then winning the stranger, then the offer, then the channels you own, then how to know what’s working. Each chapter is short. Each ends on a task.
A brand can be excellent at any one part of the machine and still lose money on every customer.
The three books before this one each went deep on one part. The Second Order is about the second purchase. Close the Loop is about customers bringing customers. The First Offer is about the offer that wins the first order. This one is the floor under all three. If you’ve read them, the chapters on the offer and on cohorts will be a quick review, and the rest is new.
Start with The Machine Audit. Your score names the part to read first. Or follow a path:
Eight tools run in the page. Nothing you type leaves your browser.
Examples that open with Say or Picture use made-up round numbers. Every source is listed in Appendix C.
What this book argues, and what would prove each claim wrong.
A position says what would prove it wrong. Test each on your own business.
Five parts. Each one feeds the next, and each one can quietly break the others.
Picture the business as a pipe with five joints. Money goes in at one end as ad spend. Profit comes out the other as repeat orders. Every joint leaks.
The whole book is above and always will be. These are the same chapters addressed individually, for linking to one idea rather than to ninety.
| Part | The question it answers | The number that tells you | Where it leaks |
|---|---|---|---|
| 1. The numbers | What does one order leave after every cost? | Contribution margin per order | Returns, shipping and discounts nobody priced |
| 2. The stranger | What does a new customer cost, really? | New-customer acquisition cost, blended | Paying for buyers who would have come anyway |
| 3. The offer | Why this, why now, why you? | First-order conversion and second-order rate by offer | Offers that win bargain hunters |
| 4. The owned channels | Do they come back without another ad? | Second-order rate and revenue per subscriber | Missing flows, a shrinking list, the spam folder |
| 5. The proof | Which of this is working? | Cohort contribution at 90 and 180 days | Reports built on calendar months and platform claims |
The joints interact, and that’s the reason to learn all five. A deeper discount lifts part 2 and quietly lowers parts 1 and 4. A stronger welcome series makes paid acquisition look better, because the platform takes credit for orders the email closed. A deliverability problem shows up as “campaigns stopped working” and gets fixed with more campaigns, which makes it worse.
Most teams are organized by channel: a paid person, an email person, a site person. Each optimizes their joint. Nobody owns the pipe. This book is written for whoever should.
Twelve checks, one per joint that leaks most. About an hour, most of it spent finding out who knows the answer.
The audit isn’t a report card. It finds the one or two joints that are leaking most, so the next quarter goes there instead of into whatever the loudest person on the team cares about.
Open your store admin, your ad accounts, your email and SMS platform, and your finance sheet. Score each check 0 to 2: 0 if it failed or nobody can answer it, 1 if partly true, 2 if clean. “We could find out” is a 0. Don’t read the bands until all twelve are scored.
One rule makes this work: the score goes to whoever can answer from the data today, not whoever thinks they know. If the email lead says the welcome series is live and the platform says two of three emails are in draft, the platform wins. I’ve audited accounts where a team was sure its most important automation was running and it had been sitting in draft for months.
“We could find out” scores a zero.
Score as you go; your band appears when all twelve are in.
| Score | What it means | Read next |
|---|---|---|
| 20–24 | The machine runs. Your job is proving which parts add the most. | Part five, starting at Cohorts, Not Calendars |
| 14–19 | The machine runs and leaks at one or two joints. Fix the zeros first. | The chapter linked from your lowest check, then The Weekly Rhythm |
| 8–13 | You have channels, not a machine. Nobody owns the pipe. | Part one, starting at One Order’s P&L |
| 0–7 | You’re spending before you’ve priced what you’re buying. | One Order’s P&L, then The First Ninety Days |
Fix zeros before ones, and fix them in check order. The first three checks price the business. Until they’re answered, every decision in parts two through four is a guess about whether you can afford it.
One pattern shows up often enough to name. A brand scores well on the paid checks, badly on the owned-channel checks, and has no cohort report. That brand is renting every customer twice: once to acquire them and again, through retargeting and discount codes, to get them back. The fix starts in part four.
Revenue is what the customer paid. Contribution margin is what you kept. Every other number in this book is priced in the second one.
Ask a founder what an order is worth and you’ll usually hear the average order value. Ask finance and you’ll hear gross margin. Both are wrong for running growth, because both leave out the costs that move when marketing moves.
Contribution margin is what one order leaves after every cost that exists only because the order happened. It’s built in three layers, and the names vary by company. These are the ones I use.
| Layer | Start from | Subtract | What it tells you |
|---|---|---|---|
| Net revenue | List price of the items | Discounts and codes; add shipping charged to the customer | What the customer actually paid |
| CM1 | Net revenue | Product cost, including inbound freight and duties | Whether the product makes money at all |
| CM2 | CM1 | Pick and pack, outbound shipping, packaging, payment fees, returns and refunds | What the order leaves to pay for marketing |
| CM3 | CM2 | Marketing spent to get the order | What the order leaves to pay for the company |
CM2 is the number growth should run on. It’s what an order can afford to spend on acquiring the next one. Gross margin stops at CM1 and ignores shipping and returns, which in apparel and heavy goods can be the difference between a profitable order and a loss.
Gross margin tells you the product is fine. CM2 tells you the order is.
Picture an $80 basket. The customer used a 15% code, so they paid $68, plus $5 shipping you charged: $73 in. The product cost $22. Picking, packing and packaging cost $5.50, the carrier charged $9, and the card processor took 3%, about $2.19. One in ten orders like this comes back for a refund, and half of what comes back can be resold.
Take the refunds out first. On average this order brings in $65.70, and the product it really costs you, after resold returns go back on the shelf, is $20.90. CM1 is $44.80, which looks healthy. CM2 subtracts $5.50, $9 and $2.19, leaving about $28.11, or 43% of net revenue. If you spent $35 on ads to win this order, CM3 is about −$6.89. The order lost money, and the ad platform reported a return on ad spend of 2.1.
Now look at where each marketing decision lands. The 15% code cost $12 of revenue and about $10.44 of CM2, more than the carrier. Free shipping would cost about another $4.35. A looser return policy moves the returns line. A bundle that raises the basket to $110 probably adds more CM2 than any code takes away, because pick, pack and payment don’t scale with the price.
New customers, what each one cost, what their first order left, and how many come back. Almost everything else is a detail of one of these.
A DTC business can be written as one sentence: this many new customers, at this cost each, leaving this much on the first order, with this share coming back. If you can fill in that sentence every Monday, you can run the company. If you can’t, no dashboard will save you.
| Number | How to count it | Why it’s here |
|---|---|---|
| New customers | People whose first-ever order was this week. Not orders, not sessions, not “new visitors”. | It’s the only measure of whether the business is growing its base. |
| New-customer cost | All marketing spend this week divided by new customers this week. | It’s what you pay per unit of growth, with nothing hidden. |
| First-order contribution | Average CM2 on those new customers’ first orders. | It’s what each unit of growth pays back on day one. |
| Second-order rate | Share of a cohort that orders again within 90 days, then 180 and 365. | It’s what turns a costly first order into a profitable customer. |
Notice the second number uses all marketing spend, including brand, retention tooling and agency fees, divided only by new customers. That’s deliberate. It’s the harshest version, and it’s the only one that can’t be gamed by moving spend between labels. Keep the channel-level versions too; just don’t run the business on them.
Fill in one sentence every Monday: how many new customers, at what cost, leaving what, with how many coming back.
Return on ad spend is each platform’s claim about the revenue its ads caused. Every platform counts the same order if it touched the customer, so the platforms’ claims added together often come to more than the business actually sold. The reported number also mixes new and returning customers, so an ad account can post a better ROAS by retargeting people who were going to buy anyway. Chapter 8 covers what those claims are worth.
Two blended ratios fix most of this. Marketing efficiency ratio, or MER, is total revenue divided by total marketing spend. New-customer MER is revenue from first orders divided by total spend. Neither pretends to know which ad did what. Both go up when the whole machine works better, and new-customer MER can’t be propped up by harvesting existing customers.
One more ratio belongs on the page, as context rather than a target: the share of revenue that comes from returning customers. Too low, and the business is renting every order from an ad platform. Too high, and it’s living off a base it has stopped refilling. I look at the trend more than the level. A share that climbs while new customers fall is the classic early warning: the business looks stable because old customers are still buying, and the base is quietly shrinking under it.
The four numbers are most useful when they move in different directions, because each combination points somewhere specific. New customers up and new-customer cost up together is normal scaling; you’re buying further into the audience. New customers down and cost up means the creative or the offer has stopped working, and the fix is in part two or three. First-order contribution down with everything else steady is almost always discounting or product mix: someone added a code, or a cheaper product started winning in the ads. Second-order rate down for a recent cohort is the one to worry about most, because by the time it shows up in revenue, several more cohorts like it have already been bought.
Lifetime value tells you what a customer might be worth someday. Payback tells you when you get your cash back. Only one of them pays the supplier.
Lifetime value is the most quoted number in DTC and one of the least useful for running the business week to week. It isn’t wrong. It answers a question you rarely need answered.
It has four problems. The horizon is unknown: “lifetime” for a brand that’s three years old is a forecast about years that haven’t happened. It ignores time: a dollar of contribution in month 30 doesn’t pay for inventory in month 2. It’s an average, and customer bases are lopsided, so the average customer barely exists. And it’s fragile. Nudge the assumed retention rate a few points and lifetime value doubles, which is how a spreadsheet justifies an acquisition cost the bank account can’t.
Payback asks a narrower question: how many months until a new customer’s contribution, first order plus reorders, covers what you paid to acquire them? It uses only numbers you’ve already seen happen, and it maps straight onto cash.
Lifetime value is a forecast. Payback is a receipt.
The strongest position is paying back on the first order. Warby Parker’s 2021 registration statement said it was “profitable on a customer’s first order” Filed. Casper’s 2020 registration statement described what it called “first purchase profitable” e-commerce economics Filed. Both are worth reading together, because they show the same claim can sit inside very different businesses. Chapter 5 is about the second one.
Most brands don’t pay back on the first order, and that’s fine if the reorders are real and fast. What matters is choosing a ceiling, the longest payback you’ll accept, and holding the media budget to it. The right ceiling depends on how much cash you have and how sure you are of the reorder curve. A consumable with a steady replenishment cycle can carry a longer ceiling than a durable whose repeat rate is still a guess.
Picture two brands that each pay $60 to acquire a customer and each earn $150 of contribution per customer over three years. The spreadsheet says they’re identical: 2.5 times return on acquisition cost. Brand A’s customer leaves $40 on the first order and then about $5 a month in reorders; it pays back in the fourth month. Brand B’s customer leaves $10 on the first order and reorders once a year; it pays back in year two. Brand A can double its spend next quarter from its own cash. Brand B has to raise money to grow at all, and if the reorder assumption is wrong, it finds out two years and many cohorts later.
A famous brand, a product people love, and marketing that ran at 35% to 43% of revenue in every year it reported before going public. The numbers show why.
Casper sold mattresses online, made the category feel modern, and became one of the best-known DTC brands of its decade. When it filed to go public in January 2020, the registration statement laid out its economics in more detail than most brands ever share. They’re worth reading as a lesson in what a product’s replacement cycle does to everything else.
| Year | Revenue | Sales and marketing | As share of revenue | Net loss |
|---|---|---|---|---|
| 2017 | $250.9M | about $106.8M | about 43% | ($73.4M) |
| 2018 | $357.9M | $126.2M | 35.3% | ($92.1M) |
| 2019 | $439.3M | $154.6M | 35.2% | ($93.0M) |
FiledCasper Sleep Inc., Form S-1 (January 2020) for 2017 and 2018; full-year 2019 results released March 19, 2020. Shares of revenue are my arithmetic Derived.
Revenue grew about 75% in two years. Losses grew with it. Sales and marketing spending never fell below 35 cents of every dollar of revenue.
The registration statement put the core problem in one line: the traditional replacement cycle of many of its products was “longer than Casper’s existence” Filed. A mattress is bought once in many years. That means almost every sale has to come from a new customer, and every new customer has to be bought.
Casper knew this and built around it. The filing describes a “Sleep Economy” worth hundreds of billions of dollars, and a product line that reached into pillows, sheets and bedding. It reported that more than 16% of its direct customers had come back to buy again Filed. For a mattress company that’s a real achievement. For a business spending more than a third of revenue on marketing, it wasn’t enough. Casper also described its e-commerce economics as “first purchase profitable,” by its own definition, and still lost $93 million in 2019. Whatever the first order covered, it didn’t cover the rest of the business.
When the product’s clock is measured in years, the first order has to pay for itself. Nothing after it arrives soon enough.
Casper priced its IPO on February 5, 2020 at $12 a share Filed, after cutting the expected range from $17–19 Reported. The pricing valued it at roughly $575 million, against a $1.1 billion valuation in its last private round Reported. In November 2021 it agreed to be taken private by Durational Capital at $6.90 a share, and the deal closed in January 2022 Filed. The brand was later sold on to Carpenter Co. Reported.
It would be easy to tell this as a story about bad marketing. It isn’t. Casper’s marketing was famous for being good. The lesson is about fit between the growth model and the product’s clock. A consumable reordered monthly can afford to lose money on the first order, because payback comes in weeks. A product replaced once a decade can’t, because payback may never come from that customer at all.
| Product clock | Examples | Where payback has to come from |
|---|---|---|
| Weeks | Coffee, supplements, pet food | Reorders. You can afford a thin first order if the reorder curve is proven. |
| Months | Skincare, apparel basics, refills | A mix. The first order should cover most of acquisition; the second order finishes the job. |
| Years | Mattresses, furniture, luggage | The first order, plus accessories and referrals. Price acquisition on the first order alone. |
The oldest finding in brand research, and the one retention marketers argue with most. They’re both right, about different things.
I’ve spent most of my career on retention, so this chapter is an argument with myself. The evidence says that brands grow mainly by winning more buyers, not by making existing buyers more loyal. Retention still matters enormously. It just matters for a different reason than most retention people claim.
In 1963 a sociologist named William McPhee described a pattern in popular culture: less-known entertainers had fewer fans, and those fans also liked them slightly less. Andrew Ehrenberg and colleagues later showed the same pattern, which McPhee had named double jeopardy, across many consumer categories. Smaller brands have far fewer buyers, and those buyers are also slightly less loyal Published. Byron Sharp’s How Brands Grow (2010) made it famous outside academia.
The practical meaning is this: when a brand grows, it’s mostly because it reached more buyers, and loyalty rises a little as a side effect. Loyalty doesn’t come first. Research from the same institute estimates that winning new buyers matters roughly twice as much for growth as reducing defection Published.
Brands grow by winning more buyers. Retention decides whether those buyers are worth winning.
If growth comes from penetration, why build a retention program at all? Because retention changes what each new buyer is worth, and therefore how much you can afford to pay for one. A brand whose second-order rate goes from 20% to 30% hasn’t grown by itself. It has raised the price it can pay for the next thousand new customers, which lets it outbid competitors for the same attention. Retention is the fuel. Reach is the engine.
This also explains a common trap. A team fixes its flows, repeat revenue rises, total revenue looks healthy, and new-customer counts quietly fall because paid spend was cut to “let retention carry it.” A year later the returning base has thinned, and there’s no new base behind it. The share of revenue from returning customers going up is only good news if new customers aren’t going down.
Les Binet and Peter Field studied the UK’s IPA effectiveness database and found that, on average, the most effective campaigns split budget about 60:40 between long-term brand building and short-term activation Published. Their 2018 update put it at 62:38, and the IPA is clear that the right split varies a lot by category. Treat it as an average from mostly larger advertisers, not a rule for a two-year-old DTC brand.
The useful idea underneath is that some spend creates demand and some spend harvests it. Most DTC ad accounts are almost entirely harvest: retargeting, lookalikes of buyers, and prospecting optimized for a purchase within a week. That works until the pool of people already primed to buy is used up, and then costs rise for no visible reason.
Most creative briefs say what the ad should look like. The ones that produce winners also say what it’s selling, why it matters, and to whom.
Creative is the biggest lever left in paid social. The platforms have taken over targeting and bidding; what you control is what the ad says and who it says it to. Yet most briefs I see are a format request: “three UGC videos and two statics for the spring launch.” That brief can produce a good ad by luck. It can’t produce a lesson.
A brief that teaches has four parts, and each one is a decision you can test separately.
| Part | The question | Example |
|---|---|---|
| Offer | What exactly are we selling, at what price and on what terms? | The starter set, full price, free exchanges for 60 days |
| Angle | Why would this person care, in their words? | “I’ve tried five of these and they all broke” |
| Audience | Who is it for, and how much do they already know? | People who’ve bought a cheaper version and been let down |
| Format | How does it show up in the feed? | A 20-second side-by-side durability test, filmed on a phone |
Format is last on purpose. It’s the part everyone can see, so it gets all the attention, and it’s the part least likely to be why an ad won. Two ads in the same format with different angles usually differ more than two ads with the same angle in different formats.
Test the message before the format. The format is how you say it; the angle is what you’re saying.
In 1966 the copywriter Eugene Schwartz published Breakthrough Advertising, and two of its ideas still decide whether an angle works Published. The first is the state of awareness: how much the reader already knows about their problem, the kinds of solutions and your product. People who don’t yet feel the problem need the problem named; people who already know your product need a reason to act today. Modern marketers usually label the stages unaware, problem-aware, solution-aware, product-aware and most aware. Those labels are later shorthand for Schwartz’s descriptions, but they’re useful.
| What they know | Lead with | A typical miss |
|---|---|---|
| Nothing about the problem | A story or moment they recognize | Leading with the product’s features |
| The problem, not the solutions | Naming the problem better than they can | Leading with price |
| The kinds of solution | Why this kind is better, with a demonstration | Claiming to be the best without showing it |
| Your product, not convinced | Proof: reviews, comparisons, guarantees | Explaining what the product is again |
| Your product, ready to buy | The offer and a reason to act now | Another brand video |
The second idea is market sophistication: how many competitors have already made your claim. In a new category, stating the benefit plainly works. Once everyone has said “the most comfortable,” you need a bigger claim, then a new mechanism, then an identity the buyer wants to belong to. Schwartz numbered five stages. You don’t need the numbers. You need to notice when your best claim has become the category’s wallpaper.
Not from the brand deck. From customers, in their own words. Read the 3-star reviews of your product and your competitors’: they say what people hoped for and what almost stopped them. Read support tickets from people who haven’t bought yet. Read the comments under your ads. Put the exact phrases in the brief. The best angles I’ve seen were a customer sentence that nobody on the team would have written.
A one-page brief template is in Appendix B.
Every ad platform reports the sales it touched. Only an experiment tells you the sales it caused. Three studies show how far apart those can be.
An ad platform’s dashboard answers the question “which sales did our ads come near?” You want the answer to a different question: “which sales would not have happened without the ads?” The gap between the two is where most wasted ad money lives.
In 2012 eBay’s economists ran a field experiment on their own search advertising. First they stopped buying ads on searches for the word “eBay” on MSN and Bing. Almost nothing happened. The people who would have clicked the ad clicked the free listing right underneath it instead, and organic search recovered 99.5% of the traffic Published. eBay had been paying for visits it would have received anyway.
Then they switched off non-brand search ads, the ads on searches like “used guitar,” in 68 of the 210 US TV markets for about two months, and compared those markets with the rest. The ads did bring in new and infrequent buyers. But most of the spend went to frequent eBay users who would have bought regardless. The measured return on that spending was −63%. An ordinary, non-experimental analysis of the same data had estimated +4,173% Published. That’s Blake, Nosko and Tadelis, published in Econometrica in 2015.
An observational estimate said +4,173%. The experiment said −63%. Same ads, same company.
A few years later, researchers at Northwestern and Facebook took 15 large US advertising experiments run on Facebook, where the true effect was known because a random control group never saw the ads, and asked how close standard non-experimental methods would have come. In half the studies, the methods were off by a factor of three Published. In one checkout study the experiment measured a 73% lift; simply comparing people who saw the ads with people who didn’t gave 316%. The methods mostly overestimated, but not always (Gordon, Zettelmeyer, Bhargava and Chapsky, Marketing Science, 2019).
The reason is simple. The people an ad system chooses to show ads to aren’t a random sample. They’re the people most likely to buy. Compare them with everyone else and you measure the targeting, not the ad.
Randall Lewis and Justin Rao looked at 25 large advertising experiments with US retailers and brokerages and found a sobering problem: individual sales are so noisy that even big experiments often can’t tell a good campaign from a break-even one Published. The median confidence interval on return was more than 100 percentage points wide. Only 3 of the 25 had enough data to show that a campaign earning a strong +50% return beat break-even. Their conclusion was that informative tests can need more than ten million person-weeks.
You don’t need ten million people to use this. You need to stop treating small differences as findings, and run tests long enough, on big enough groups, to see the effect you care about.
In April 2021 Apple released iOS 14.5, which asked iPhone users whether apps could track them across other companies’ apps and websites. Many said no. The ad platforms lost much of the data they used to connect an ad to a purchase. In February 2022 Meta’s chief financial officer estimated that Apple’s iOS changes would cost the company on the order of $10 billion in revenue that year Published. Since then, platform-reported results have leaned more heavily on modeled conversions: estimates of sales the platform believes it caused but can’t observe. The dashboard became more confident and less checkable at the same moment.
You paid for the visit. Seven in ten carts still get left. The fixes are dull, cheap, and worth more than most ad tests.
Every dollar spent in the ad account arrives at a product page. If that page and the checkout behind it lose a few more people than they should, every campaign costs more than it has to, and no amount of creative testing will show you why.
The Baymard Institute has averaged 50 studies of online cart abandonment and puts the rate at about 70% Reported. Some of that is people browsing with no intent to buy. Baymard also surveys US shoppers who abandoned a checkout about why, setting aside the “just browsing” answers.
| Reason for leaving the checkout | Share of shoppers |
|---|---|
| Extra costs too high (shipping, tax, fees) | 40% |
| Delivery was too slow | 20% |
| Didn’t trust the site with card details | 19% |
| Had to create an account | 18% |
| Checkout too long or complicated | 17% |
| Website errors or crashes | 17% |
| Returns policy wasn’t good enough | 13% |
| Couldn’t see the total cost up front | 12% |
ReportedBaymard Institute, cart abandonment rate statistics, 2025. Shoppers could give more than one reason.
Read the list again with your own site in mind. Almost every line is something the shopper learned too late: the shipping cost, the delivery date, the return terms, the need for an account. The cheapest checkout fix is usually to tell people earlier.
Most checkout losses are surprises. Tell people the cost, the date and the terms before they reach the checkout.
Northwestern’s Spiegel Research Center found that a product with five reviews was about 270% more likely to be bought than one with none Reported. Displaying reviews raised conversion more for expensive items (380%) than for cheap ones (190%). And purchase likelihood peaked at an average rating between 4.0 and 4.7, then fell as ratings approached a perfect 5.0. Shoppers distrust perfection. Don’t hide your 3-star reviews; answer them.
A 2020 study by Deloitte Digital for Google followed 37 brands and found that a 0.1 second improvement in mobile site speed went with 8.4% more conversions and 9.2% higher order values for retail sites Reported. It’s observational and it was commissioned by Google, which has its own stake in a faster mobile web, so don’t bank the exact figures. The direction is not in doubt, and most DTC stores carry apps and scripts that no one has audited in years.
A product page answers four questions in the order a stranger asks them: what is it, is it for me, can I trust it, and what happens if I’m wrong. Most pages answer the first at length and the last in the footer. Put the answer to “is it for me” in the first screen on a phone: who it’s for, the problem it solves, a photo of it in use. Put trust next: reviews, the count and the average. Put the delivery date, the return terms and the total cost near the button. Everything else, the brand story, the materials, the founder’s note, goes below for the people who scroll.
A loved product, a loyal base, a celebrated IPO, and revenue that fell every year after 2022. The numbers suggest a brand that stopped winning enough new buyers.
Allbirds went public in November 2021 as one of the most admired brands in DTC. Its wool sneakers had a devoted following and its customers came back. Four and a half years later the company sold its brand and intellectual property for about $39 million. The path between those two points is a lesson in what loyalty can and can’t do.
The registration statement described a base most brands would envy. About 53% of 2020 net sales came from repeat customers. Repeat customers spent more than 25% more in their second year than in their first. About 80% of repeat orders included a different item from the customer’s first purchase Filed. This was a brand with a real second order.
Allbirds priced its IPO at $15 a share on November 2, 2021 Filed. The stock closed its first day at $28.64 Reported.
| Year | Net revenue | Net loss |
|---|---|---|
| 2021 | $277.5M | ($45.4M) |
| 2022 | $297.8M | ($101.4M) |
| 2023 | $254.1M | ($152.5M) |
| 2024 | $189.8M | ($93.3M) |
FiledAllbirds, Inc. full-year results. From 2022 to 2024 revenue fell 36% Derived.
In September 2024 the company did a 1-for-20 reverse stock split to keep its Nasdaq listing Filed. In March 2026 it agreed to sell its intellectual property to American Exchange Group for about $39 million Filed, and the remaining public company later renamed itself and changed business entirely Reported.
On the company’s March 2023 earnings call, co-founder and then co-chief executive Joey Zwillinger said: “We overemphasized products that extended beyond our core DNA” Reported. He named specific new shoes and seasonal colors, and added that the company’s “overinvestment on newness came at the expense of focus on consumers who are loyal to our brands.”
A loyal base is a floor, not an engine. It holds revenue up while you figure out how to win the next buyer. It doesn’t win them for you.
It’s tempting to call this a retention failure. The filings suggest the opposite: the repeat behavior was strong. What the business needed after going public was more new buyers at a cost it could afford, and a reason for people who didn’t already love wool sneakers to try the brand. The product extensions were an attempt at that. By the company’s own account they pulled focus from the loyal buyers, and the falling revenue says they didn’t win enough new ones.
This is double jeopardy from chapter 6 in real life. A brand’s existing customers can’t grow it forever. Growth has to come from penetration, and penetration has to come from something new buyers want, sold at an acquisition cost the first order can support.
Warby Parker went public about five weeks earlier, in a direct listing. Its registration statement said it was “profitable on a customer’s first order” Filed. Revenue went from $393.7 million in 2020 to $771.3 million in 2024, and in 2025 the company reported its first full year of GAAP net income, $1.6 million on $871.9 million of revenue Filed. It’s not a flawless comparison: glasses are a prescription need with a built-in replacement cycle, and Warby leaned heavily on physical stores. But the difference in the first sentence of each company’s economics, first-order profit versus repeat-driven value, is the difference that mattered.
The offer gets a whole book of its own. Here are the parts every operator needs, and where to go for the rest.
The offer is what sits between the ad and the order: what you’re selling, at what price, on what terms, with what promise. It’s usually decided by whoever writes the next ad. It should be decided once, on purpose, by whoever owns the numbers in part one.
A discount is only one of the five, and usually the most expensive one to use.
In 2006 the marketing professor Michael Lewis studied customer records from a newspaper and an online grocer and found that customers won with deeper discounts were worth less over time. A 35% acquisition discount produced customers worth about half as much as full-price customers Published. Three field studies published in 2004 by Eric Anderson and Duncan Simester at a catalog retailer found close to the opposite for new customers: deeper first discounts led to more buying later Published. The two fit together once you ask what the discount teaches, and to whom. The First Offer spends a chapter on it.
An offer is a price for the first order and a lesson about every order after it.
The full treatment, with eight tools and the research behind each point, is in The First Offer.
A subscriber list decays every month. If sign-ups don’t outrun the decay, every campaign reaches fewer people than the last one did.
The email and SMS list is the only audience a DTC brand owns outright. No auction, no algorithm, no rent. It’s also the asset most brands measure worst. They track total subscribers, which only ever goes up, instead of engaged subscribers, which is the number that pays.
People stop opening, change addresses, lose interest, or move your mail to a folder they never read. Some of that is your fault and some isn’t. Either way, a share of your engaged list goes quiet every month. The list only grows if new engaged sign-ups outrun that decay.
Total subscribers only goes up. Engaged subscribers is the number that pays.
From my work: when I audit an account, the most common list problem isn’t a weak offer. It’s a popup that was switched off during a redesign and never switched back on, or a stack of old forms of which only a couple are live. Check what’s actually running before you redesign anything.
The tool shows something counterintuitive. A list that isn’t growing is usually fixed faster by slowing decay than by adding sign-ups. Decay is driven by what you send: too many promotions, sends to people who stopped reading long ago, and nothing worth opening between sales. Chapter 15 is about the send mix. Chapter 13 is about the part of decay that’s really mail going to spam.
An email in the spam folder costs the same to send and earns nothing. Deliverability is a weekly number, not a one-time setup.
Deliverability problems are sneaky. Nothing breaks. Revenue per send slides a little each month, the team blames the creative, and sends more to make up the difference, which makes it worse. By the time someone checks, a large share of the list has been reading you in the spam folder for a quarter.
Starting February 1, 2024, Google required anyone sending more than 5,000 messages a day to Gmail addresses to meet a set of rules Published. Yahoo announced matching requirements. Once you cross the threshold, Google treats you as a bulk sender permanently.
Most email platforms handle the header and the link. The domain records are yours. So is the complaint rate, and it’s the one that moves. Set up Google Postmaster Tools on your sending domain today; it’s free and it’s the only direct view you get of how Gmail sees you.
Your spam complaint rate is the most important email number most teams have never looked at.
In September 2021 Apple shipped Mail Privacy Protection with iOS 15. When it’s on, Apple Mail downloads an email’s images in the background, whether or not the person reads it Published. Opens from those readers are recorded automatically. Open rates rose across the industry as people updated their phones, and they’ve been unreliable ever since.
This matters in three places. Don’t judge subject lines on open rate alone. Don’t define “engaged” by opens only; include clicks, site visits and purchases. And don’t let a healthy-looking open rate reassure you about deliverability. An account can show strong opens and a collapsing click rate at the same time, and the click rate is the one telling the truth.
It rarely looks like a problem. Open rates hold, because of the Apple change. Click rate drifts down a little each month. Revenue per campaign falls, and the team decides the creative is tired. In Postmaster Tools, domain reputation slips from high to medium, and the spam rate line touches 0.1% on the days you mail your widest segment. If you see that pattern, the fix isn’t new creative. It’s narrowing who you send to for a few weeks, letting the engaged readers rebuild your reputation, and widening again slowly.
Automated messages triggered by what a customer just did. They’re a small share of sends and a large share of revenue, because they arrive when someone is paying attention.
A campaign is a message you decide to send to a group on a date. A flow is a message a customer triggers by doing something: signing up, leaving a cart, receiving an order, going quiet. Flows arrive at the moment of highest attention, which is why they earn so much per message.
Klaviyo’s 2026 benchmark report, across more than 183,000 of its customers, found that flows produced about 41% of email revenue from just 5.3% of sends, and earned roughly 18 times the revenue per recipient of campaigns Reported. Their order rate was 2.11% against 0.16% for campaigns. Omnisend reported a similar pattern in its 2025 data: automations were 2% of sends and 30% of email revenue Reported. Both are vendors with a stake in email, and both averages include a lot of neglected accounts. The direction is the point.
Flows are the floor. Campaigns are the ceiling. Build the floor first.
| Flow | Triggered by | Its job | First thing to get right |
|---|---|---|---|
| 1. Welcome | A new subscriber | Turn a sign-up into a first order, and teach what the brand is | That it’s live. Then deliver the promised offer in the first message, immediately. |
| 2. Cart and checkout abandonment | A started checkout or a cart with an email attached | Recover orders that were nearly placed | Timing. Test the first message at about an hour against later. |
| 3. Post-purchase | A first order | Make the first experience good and set up the second order | Branch on the first product. A skincare buyer and a gift buyer need different next steps. |
| 4. Browse abandonment | A known subscriber viewing products without adding to cart | A gentle reminder, often with reviews or a comparison | Suppress people who just got a cart email. |
| 5. Winback | A customer passing their usual reorder window without buying | Bring lapsed customers back before they’re gone | Time it from your reorder curve, not a round number. |
| 6. Sunset | A subscriber with no engagement for a long stretch | Last chance to re-engage, then suppress | That it exists. It protects deliverability for everyone else. |
Once those six are healthy, add the ones your category earns: replenishment reminders for consumables, a VIP series for your top customers, back-in-stock, price drop, birthday, review requests and cross-sell.
It’s the flow most worth getting right, because every subscriber passes through it exactly once. The first message goes immediately and delivers whatever the sign-up promised, with no preamble. The next few, over the following week or two, do the work of a good salesperson: say what the brand is and isn’t, show the product that most first-time buyers choose, answer the most common objection from your reviews, and show proof from customers like the reader. Exit anyone who buys into the post-purchase flow straight away. If the sign-up offer has an expiry, say so once, near the end, and mean it.
From my work: the flows most accounts have are fine on paper and broken in the wiring. When I audit an account, these are the faults I find most often.
Group flows in your platform by the stage of the customer’s life they serve: getting a first order, recovering an order, after the order, and bringing someone back. When they’re organized that way, gaps are obvious.
Revenue per recipient: revenue attributed to the flow divided by the people who entered it. It lets you compare a welcome series with a winback flow fairly, and it shows when a change helped. For the flows that matter most, add a holdout, a random slice of people who enter the flow and receive nothing, so you know what the flow adds beyond orders that would have happened anyway. The Second Order covers holdouts on flows in detail.
Campaigns earn less per send than flows and matter anyway: they’re how you reach everyone who isn’t in the middle of something. The calendar decides whether they build the list up or wear it down.
Most brands run campaigns the way they run sales: reaching for one when the month looks soft. Sends cluster around promotions, go quiet between them, and the list learns that email means a discount. That’s the fastest way to make a list worth less every year.
From my work: at least one campaign a week to your engaged list, planned four weeks ahead, is the floor for almost any DTC brand with a list worth mailing. Brands with more to say and a highly engaged list can send several times a week. Below weekly, the list forgets you between sends, and every send works harder to be noticed.
Planning a month ahead isn’t bureaucracy. It’s what lets campaigns work with the flows instead of against them, lets creative be made properly rather than the night before, and lets the promotions be deliberate instead of reactive.
A campaign calendar that’s all promotions teaches the list to wait. One that’s never promotional leaves money on the table in the weeks when people are ready to buy. When I rebuild a calendar that has become nearly all flash sales, I start from roughly 60% promotional and 40% everything else, then adjust by what the engagement data says. “Everything else” is anything worth opening without a code: how the product is made, how to use it, customer stories, new arrivals, the founder’s view on something in the category.
A list that only hears about sales learns to wait for them.
The non-promotional sends aren’t charity. They keep engagement up between sales, which keeps you out of the spam folder, which makes the promotional sends reach more people. And they sell. A product story often outsells a discount among customers who were going to pay full price anyway.
The subject line has one job: to earn the open from someone who is scrolling past forty other senders. Say what’s inside, specifically. “New: the travel size is back” beats “You won’t want to miss this.” The preview text is a second line, not a repeat; use it to add the detail the subject line had no room for. Judge subject lines on clicks and orders per recipient, not opens, since opens stopped being reliable in 2021. And keep a house style, so the inbox sees the same brand every time: mine is title case subject lines and lowercase preview text, which reads as a headline and a whisper.
Plan the year’s sales in January: which moments, how deep, for whom. A small number of real sales, announced and ended on time, trains customers to act when you ask. A constant stream of “flash” sales trains them to never pay full price. The First Offer has a chapter on the sale calendar and what JCPenney learned about untraining a discount.
You don’t need two hundred segments. You need about a dozen, built on three questions: are they reading, where are they in their life with you, and how much are they worth?
Segmentation gets sold as sophistication. In practice most accounts have either none, one list sent everything, or dozens that nobody remembers building. The useful middle is small and boring and pays for itself every week.
Engagement segments decide who gets a send. Define engaged by clicks, site visits and purchases as well as opens, because of the Apple change in chapter 13.
Adjust the windows to your send frequency and product clock. A brand that sells a yearly purchase can keep people longer than a brand that sends daily.
Lifecycle segments decide what a send says. The same product launch should read differently to someone who’s never bought and to someone on their sixth order.
| Stage | Who | What they need to hear |
|---|---|---|
| Subscriber, never bought | Signed up, no order yet | Why this brand, proof, a reason to make the first order |
| One-time buyer | Exactly one order | How to get the most from what they bought, then the natural next product |
| Repeat buyer | Two or more orders, recent | What’s new, what goes with what they own, recognition |
| Top customers | Your highest spenders | First access, a human voice, no discounts they don’t need |
| Lapsing | Past their usual reorder window | A reminder of what they liked, and a reason to come back |
| Lapsed | Well past the window | Winback, then less frequent contact |
RFM scores every customer on recency (how recently they bought), frequency (how many times) and monetary value (how much in total). Rank customers into five equal groups on each and you get scores like 5-5-5 for your best and 1-1-1 for your coldest. It’s decades old, it needs nothing but your order history, and it’s often more useful day to day than a predictive score, because anyone on the team can see why a customer landed where they did.
Three questions, about a dozen segments. Anything more has to prove it changed a send.
RFM earns its place in two decisions. Who gets recognition instead of discounts: your high-frequency, high-value customers, who will buy at full price if you let them. And who is slipping: customers with high frequency and value whose recency is falling, the most valuable people to reach before they’re gone. Chapter 20 shows how concentrated revenue usually is among them.
A segment earns its keep if it changes what somebody receives. If a segment exists but every send goes to the same audience anyway, delete it. Before building a new one, write down which send it will change and how you’ll know it worked.
Text messages get read. That’s why they’re regulated more tightly than email, and why a sloppy program can cost more in legal fees than it ever made.
SMS is the most intimate channel a brand can use. It sits beside messages from family. Used well, for moments that matter, it’s the highest-intent channel you have. Used as a second email list, it’s intrusive, expensive per message, and a legal exposure.
This chapter is an operator’s summary of the rules as of September 2026, not legal advice. The rules differ by state and change often. Have counsel review your consent language and sending rules before you scale.
Under the US Telephone Consumer Protection Act, marketing texts sent with an autodialer require the recipient’s prior express written consent Published. In 2021 the Supreme Court, in Facebook v. Duguid, narrowed what counts as an autodialer, and some argued that removed most SMS platforms from the rule. Don’t build on that argument. Several states have their own versions of the law, carriers have their own rules, and written opt-in remains the industry standard. Collect it clearly: what they’ll receive, roughly how often, that message and data rates may apply, and how to stop.
US carriers require businesses texting from ordinary ten-digit numbers to register their brand each messaging campaign through The Campaign Registry, a system known as A2P 10DLC. Unregistered traffic is blocked; Twilio has blocked it since September 2023 Reported. Your SMS platform handles most of this. Make sure it’s done, and that what you register matches what you actually send.
Honor STOP and its common variations immediately, and confirm the opt-out with one final message. If you sell subscriptions, cancellation belongs in the same spirit. The FTC’s “click to cancel” rule was adopted in October 2024 and vacated by a federal appeals court in July 2025. In March 2026 the FTC started the rulemaking again Published. It still enforces existing law against subscriptions that are hard to cancel, and several states have their own auto-renewal rules. Make cancelling as easy as signing up, whatever the federal rule says this year.
Every SMS you send should be one the customer would be glad to get. The law just makes that expensive to forget.
The best SMS sign-ups come from people who already gave you their email, on the second step of the form described in chapter 12, with their own reason to say yes: early access, restock alerts, or a small extra on the first order. Keyword sign-ups on packaging and at events work too. Don’t buy lists, don’t import phone numbers collected for order updates into marketing, and don’t pre-check a consent box. Each of those is a fast way to a complaint.
Four numbers for campaigns, four for flows, one for the balance between them. The lines I use to decide in five minutes whether an email program is healthy.
When I open an email account for the first time, I don’t start with the calendar or the creative. I pull the last 30 days of campaigns and flows and check eight rates against three bands. It takes five minutes and it tells me where to look for the rest of the day.
These are my working thresholds, built from auditing accounts, not an industry standard. The spam lines are anchored to Google’s published limits. The rest are where I’ve found the difference between accounts that are working and accounts that are drifting. Treat them as a starting point and adjust for your category.
| Last 30 days | Healthy | Watch | Act now |
|---|---|---|---|
| Campaigns | |||
| Click rate | 1.25% or more | 0.5 to 1.25% | under 0.5% |
| Placed order rate | 0.5% or more | 0.2 to 0.5% | under 0.2% |
| Unsubscribe rate | under 0.5% | 0.5 to 1.25% | over 1.25% |
| Spam complaint rate | under 0.1% | 0.1 to 0.2% | over 0.2% |
| Flows | |||
| Click rate | 4% or more | 2 to 4% | under 2% |
| Placed order rate | 1% or more | 0.5 to 1% | under 0.5% |
| Unsubscribe rate | under 0.5% | 0.5 to 1.25% | over 1.25% |
| Spam complaint rate | under 0.1% | 0.1 to 0.2% | over 0.2% |
From my workRates are per recipient. Spam lines sit inside Google’s bulk-sender guidance: below 0.10%, never 0.30% or higher.
Two notes on reading it. Flows are held to a higher bar than campaigns because they’re triggered by something the customer just did; a flow performing at campaign levels is broken. And my healthy line for campaign order rate is well above the average in Klaviyo’s 2026 benchmarks, which put campaigns at 0.16% Reported. That’s deliberate. The average includes many accounts mailing their whole list several times a week. Near the vendor average, you’re typical, and by my lines that still means work to do. Above 0.5%, you’re healthy.
A flow performing at campaign levels is broken, not average.
One more number: the share of email revenue that comes from flows. From my work, in a healthy account flows earn about half or more. Far below that usually means flows are missing, broken or in draft. Very far above it can mean campaigns are being under-used, or that the list has been mailed so hard that only triggered messages still get read.
Two field experiments with loyalty cards explain more about why customers come back than most loyalty software. People keep going when they can see they’re getting closer.
Loyalty programs are usually sold as a discount with a login. The research behind the ones that work says something more interesting: the reward matters less than the feeling of getting closer to it.
In 2004 Joseph Nunes and Xavier Drèze handed out 300 loyalty cards at a car wash. Every card needed eight purchased washes to earn a free one. Half the cards had eight empty slots. The other half had ten slots, with two already stamped. Same effort, same reward. Only the starting point looked different.
34% of the customers with the pre-stamped card finished it, against 19% with the plain card. They also came back sooner, about three days less between washes on average Published. Nunes and Drèze called it the endowed progress effect: people work harder toward a goal they believe they’ve already started.
People finish what they believe they’ve already started.
In a paper published the same year, Ran Kivetz, Oleg Urminsky and Yuhuang Zheng studied a café reward card that gave a free coffee after ten purchases. Customers bought more often as they got closer to the reward: the gap between visits shrank by about 20% from the first stamp to the last. Customers given a twelve-stamp card with two stamps already filled finished faster than those with a plain ten-stamp card, 12.7 days against 15.6 Published. The goal gradient, an idea psychologists first described in the 1930s, works on coffee drinkers too.
A calendar month mixes customers who arrived yesterday with customers who arrived three years ago. A cohort follows one group from its first order forward. Only one of them can tell you whether anything you changed worked.
Most ecommerce reporting is by calendar month: revenue in March, repeat rate in March, customers in March. It feels natural and it hides almost everything. March’s revenue includes new customers you paid for last week and loyal customers you won in 2022. If the welcome offer got worse in January, March can still look fine, because the old customers are carrying it.
A cohort is everyone whose first order fell in the same period. Follow a cohort forward and you can see what those customers did in their first 30, 90, 180 and 365 days. Compare cohorts and you can see whether customers you won this spring are behaving better or worse than those you won last spring. That’s the only fair test of a change in offer, creative, channel or flows.
Calendar reports tell you how the business did. Cohort reports tell you whether it’s getting better at what it does.
From my work: the most common cohort mistake I see is a report that’s labeled 90-day, 180-day and 365-day lifetime value but is really three calendar windows stacked back from today. The giveaway is the customer count. In a true cohort view, the count falls as the window gets longer, because only customers acquired at least 365 days ago can have a 365-day value. If the count rises with the window, it’s not a cohort report. The First Offer walks through how to rebuild it; Appendix A has the query.
Sort every customer you’ve ever had by lifetime spend and the shape is almost always lopsided. Picture a brand with 20,000 customers and $2 million of lifetime revenue.
| Group | Customers | Share of revenue | What it means |
|---|---|---|---|
| Top 1% | 200 | 15% | Named people. Treat them like accounts. |
| Top 10% | 2,000 | 45% | The base the business stands on. |
| Bought once | 14,000 | 35% | Seven in ten customers, a third of revenue. |
PictureMade-up round numbers with a shape I see often. Pull your own; the tool below uses them.
That shape suggests two separate programs, with different owners and different numbers. A top-customer program, whose goal is to lose none of them. And a one-time-buyer program, whose goal is to move as many as possible to a second order.
The first program is small in headcount and large in money. Give it a named owner and a list of actual people, not a segment definition. Watch recency on every one of them: when someone who normally orders every six weeks hasn’t ordered in ten, that’s a phone call or a personal note, not an automated winback with a code. Give them first access to launches, a direct line to a person, and the occasional unexpected thank-you. Measure it on one number, the share of last year’s top customers who are still active this year, and aim to lose almost none.
The second program gets easier to fund once you price it. Compare the average lifetime spend of customers with exactly one order and customers with exactly two. The difference is the step-up: roughly what a one-time buyer is worth if they place a second order. Multiply by the number of one-time buyers and a reactivation rate and you have a dollar figure for the program.
Treat that figure as a ceiling, not a forecast. Customers who reorder on their own are the more committed ones, so a customer you nudge back is probably worth less than the average two-order customer. Measure the real effect with a holdout, as in chapter 8, and use the step-up to decide whether the program is worth building at all.
| Back | People | Revenue | Net |
|---|
A test that wins without a written reason teaches you to repeat a result you don’t understand. A test that loses with one teaches you something you’ll use for years.
Most teams test. Few learn. The difference isn’t statistics; it’s paperwork. A test that started with a written hypothesis and a success metric chosen in advance produces a lesson whether it wins or loses. A test that started with “let’s try a new subject line” produces a winner at best, and nobody can say why.
Before anything launches, fill in five lines.
After the test, add two more lines: what happened, and what changed because of it. The card goes in a test log that anyone on the team can search. A template is in Appendix B.
The test log is worth more than any single winner in it.
Test where the money is and where the effect is likely to be large. From my work, the order usually looks like this: offers and flow timing first, because they move orders directly; flow structure and segmentation next; creative angles in paid social alongside; subject lines and button colors last, because the effects are small and hard to measure since opens became unreliable.
Lewis and Rao’s 25 advertising experiments in chapter 8 are the warning. Individual behavior is so noisy that small tests can’t see small effects. Three practical rules follow.
In a study published in 2000, Sheena Iyengar and Mark Lepper described a jam-tasting table in a grocery store. When it offered 24 jams, more people stopped, but only 3% of those who stopped bought. When it offered six, 30% bought Published. “Choice overload” became one of the most quoted findings in marketing. Ten years later Benjamin Scheibehenne and colleagues combined 50 experiments on the same question and found the average effect was essentially zero, with big differences between studies Published. Too much choice can hurt, under some conditions. It isn’t a law.
The lesson for your own tests is the same. One win is a hint. A result that holds when you run it again, on a different cohort, in a different month, is a finding.
From my work: one new test launched every two weeks across the program, each with a card, is a pace most teams can sustain and it compounds. Twenty-five tests a year with written lessons is a playbook no competitor can copy, because it’s about your customers.
The fundamentals don’t fail in a crisis. They drift, one missed check at a time. A fixed weekly, monthly and quarterly rhythm is what stops the drift.
Every number in this book is easy to check once. The discipline is checking them every week, in the same order, whether or not anything seems wrong. The weeks when nothing seems wrong are when the welcome series quietly breaks.
| Line | Number | From |
|---|---|---|
| 1 | New customers, new-customer cost, first-order contribution, second-order rate for the cohort from 90 days ago | Chapter 3 |
| 2 | MER and new-customer MER, with the four-week trend | Chapter 3 |
| 3 | Share of revenue from returning customers | Chapter 3 |
| 4 | Engaged subscribers, and net change this week | Chapter 12 |
| 5 | Spam complaint rate, from Postmaster Tools | Chapter 13 |
| 6 | Revenue per recipient for welcome, cart and post-purchase | Chapter 14 |
| 7 | Tests live, and any that finished, with the lesson | Chapter 21 |
Seven lines. If a line moved more than you’d expect, it gets one sentence of explanation and one owner. If nothing moved, the meeting takes ten minutes. That’s the point.
The weeks when nothing seems wrong are when things quietly break.
Once a month, a longer report answers four questions in this order. What happened to the four numbers, and why. What the cohort table says about customers won in the last three months compared with the same months last year. How the owned channels did against the health thresholds in chapter 18. What the tests taught, and what’s next. Keep it to a few pages. A report nobody reads is worse than none, because it lets everyone believe someone is watching.
Keep a dated log of every change to the machine: a new offer, a flow edited, a popup changed, a budget moved, a price raised. One line each. When a number moves three weeks from now, the changelog is how you find out why in five minutes instead of arguing about it for an afternoon. From my work, it’s the single cheapest habit that separates teams who learn from teams who guess.
Once a quarter, re-score The Machine Audit. The scores stay in your browser from last time. Compare. Any check that went down gets an owner and a date. Any check that’s been at 2 for a year might be ready for a harder test: a holdout on a flow, a geo test on a channel, a price test on an offer.
Whether you’re new to the job or starting over, here is the order I’d fix the machine in. Stop the bleeding, build the floor, then build what compounds.
The temptation in a new role is to start with the most interesting problem. Resist it. Start with what’s losing money every day it stays broken, because those fixes pay for the patience the rest needs.
Stop the bleeding, build the floor, then build what compounds. In that order, every time.
The fundamentals don’t change as a brand grows. Which one is the bottleneck does. Four stages, the number that matters most in each, and the mistake I see most often.
A brand with its first few hundred customers and a brand with a million shouldn’t read this book the same way. Everything in it applies to both. The order doesn’t. Find your stage by what’s true of the business, not by revenue, and read the chapters it names first.
What’s true: Most orders come from people who know the founder, a first community, or a handful of ads being tested. There isn’t enough data for cohorts to mean much. The product may still be changing.
What’s true: Paid acquisition works at a small budget. A few thousand customers exist. The question is whether they come back without being paid for again.
What’s true: The machine works and the constraint is how fast it can grow. Paid spend is rising. Costs per new customer are starting to creep up. More people are involved, each owning one channel.
What’s true: Returning customers are a large share of revenue. Growth has slowed. The team is big, the reports are many, and the business can look healthy for a long time while the base quietly shrinks.
The fundamentals don’t change as you grow. The bottleneck does.
What whoever owns the whole machine needs on the first day.
Whoever owns growth, a new hire, an agency, or you on a Monday you’ve decided to start over, needs six things on day one. Without them, the first month goes on finding out what they should have been handed.
The books and papers this one leans on, and what to take from each.
And the research: Blake, Nosko and Tadelis (2015) on paid search at eBay; Gordon, Zettelmeyer, Bhargava and Chapsky (2019) on advertising measurement at Facebook; Lewis and Rao (2015) on why ad effects are hard to measure; Nunes and Drèze (2006) on endowed progress; Kivetz, Urminsky and Zheng (2006) on the goal gradient; Iyengar and Lepper (2000) and Scheibehenne and colleagues (2010) on choice overload; Ehrenberg, Goodhardt and Barwise (1990) on double jeopardy. Full references are in Appendix C.
Andrew Lauchner runs Growth Legend, embedding inside consumer brands to own lifecycle, email and SMS, and revenue operations. He wrote The Second Order, on turning first-time buyers into second-time buyers; Close the Loop, on getting customers to bring the next customer; and The First Offer, on the offer that wins the first order.
As Senior Director of Growth and Retention Marketing at Gallery Furniture, he rebuilt the customer journey and the sales playbooks together. He has worked on growth and retention at Binance and 3Commas, and has been Head of Growth and Retention at Greatness Wins and at Nexus Agriscience.
The methods and thresholds marked “from my work” come from auditing and running client accounts in 2025 and 2026. Clients aren’t named and their numbers aren’t here.
“Andrew led retention, lifecycle, and email/SMS, but what separates him from most in this space is how deeply he understands the role retention plays in the overall growth engine.”
Akram Khan, Head of Marketing at Gallery Furniture, senior to Andrew but didn’t manage Andrew directly
Andrew answers every note from people running the machine, including those looking for someone to own it. Write to andrew@growthlegend.com or message him on LinkedIn.
The fields the machine needs, and five pulls that build every table in this book.
Every pull below runs on an orders table, an order-lines table, a refunds table and a customers table. Resolve identity first: customer ID, then normalized email, then phone. A guest who later made an account is one customer. Counting them twice turns a repeat buyer into two one-time buyers and breaks every cohort number.
| Field | On | Rule |
|---|---|---|
first_order_at | Customer | Date of the first paid order. Set once, never overwritten. |
first_products | Customer | Product IDs in the first order. |
is_first_order | Order | True only for the customer’s first paid order. |
net_revenue | Order | Paid, after discounts, plus shipping charged, before tax. |
cm2 | Order | Net revenue minus product cost, pick and pack, shipping paid, payment fees and refunds, from finance’s cost tables. |
sales_channel | Order | Own site, marketplace, retail, wholesale. Cohort work uses own site unless stated. |
holdout_digit | Customer | Random 0 to 9, set once at profile creation, used to hold out flows and campaigns. |
-- cumulative CM2 per customer by months since first order
WITH firsts AS (
SELECT customer_id,
DATE_TRUNC('month', MIN(ordered_at)) AS cohort
FROM orders WHERE status = 'paid' AND sales_channel = 'own_site'
GROUP BY customer_id
),
ages AS (
SELECT f.cohort, f.customer_id, o.cm2,
DATE_DIFF('month', f.cohort, DATE_TRUNC('month', o.ordered_at)) AS m
FROM firsts f JOIN orders o USING (customer_id)
WHERE o.status = 'paid'
),
sizes AS (SELECT cohort, COUNT(*) AS customers FROM firsts GROUP BY cohort)
SELECT a.cohort, s.customers, k.m AS months,
SUM(CASE WHEN a.m <= k.m THEN a.cm2 ELSE 0 END) / s.customers AS cum_cm2_per_customer
FROM ages a
JOIN sizes s USING (cohort)
CROSS JOIN (SELECT 0 AS m UNION ALL SELECT 1 UNION ALL SELECT 3
UNION ALL SELECT 6 UNION ALL SELECT 12 UNION ALL SELECT 24) k
WHERE DATE_DIFF('month', a.cohort, DATE_TRUNC('month', CURRENT_DATE)) >= k.m
GROUP BY a.cohort, s.customers, k.m
ORDER BY a.cohort, k.m;
The WHERE line is the one that stops the window trap from chapter 20: a cohort only reports a month it has lived through. Check it by eye. Reading down any column, the youngest cohorts should be missing, not zero.
Four one-page forms. Copy them into whatever your team already uses.
OFFER What exactly, at what price, on what terms:
ANGLE Why this person cares, in a customer's words:
Source of the words (review, ticket, comment):
AUDIENCE Who, and what they already know:
[ ] unaware of the problem [ ] know the problem
[ ] know the kinds of fix [ ] know us, not convinced
[ ] ready to buy
FORMAT How it shows up in the feed:
TESTING Which ONE of offer / angle / audience / format changes:
JUDGED ON New-customer cost after __ days, not clicks
NAME offer_angle_audience_format_date
BECAUSE WE SAW the observation: WE BELIEVE THAT the change, for whom: WILL MOVE one metric: BY AT LEAST smallest change that would make us act: MEASURED OVER duration and group sizes, fixed before launch: HOLDOUT yes / no, and how chosen: RESULT (after) WHAT CHANGED (after) what we now do differently
FLOW | STATUS (live/draft/off) | TRIGGER | EXCLUSIONS | MESSAGES
| SMS STEPS | REVENUE PER RECIPIENT, 90 DAYS | HOLDOUT | LAST CHECKED
1. The four numbers: this month, last month, same month last year. Why. 2. Cohorts: customers won in the last 3 months vs the same months last year, at 30 and 90 days. 3. Owned channels: the nine health numbers, with status. 4. Tests: what finished, what it taught, what's next. 5. Changelog highlights, and what we'll change next month.
Every external source, by chapter. Web sources were read in September 2026.