The offer that wins a stranger decides the customer you get.
Most brands design the first offer to win the first order. The first offer also decides who shows up, what they think your price is, and whether they ever pay it.
In 2006 a marketing professor named Michael Lewis looked at customer records from a newspaper and an online grocer. The deeper the discount that won a customer, the less often that customer bought again. A 35% acquisition discount produced customers worth about half as much, over time, as customers who came in at full price.
Two years earlier, Eric Anderson and Duncan Simester ran three field experiments at a catalog retailer and found close to the opposite: a deeper first discount made new customers buy more later, not less. Chapter 4 puts the two studies side by side. Both are right, and the reason why is the most useful idea in this book.
JCPenney learned the other half of it in public. In 2012 the company stopped running sales and printed honest everyday prices instead. Customers who had been trained for decades to wait for a coupon stopped coming. Net sales fell from $17.3 billion to $13.0 billion in one year. Chapter 6 takes that apart.
An offer is a price for the first order and a lesson about every order after it.
The Second Order was about getting a first-time buyer to buy again. Close the Loop was about getting that buyer to bring someone. This book goes back to the start: the offer that turns a stranger into a first-time buyer, judged by what that buyer does next. It covers what most offer writing covers, from value equations to bundles to testing on Meta. It adds what most offer writing skips: cohorts, contribution margin after the first order, and the second offer.
Start with The Offer Audit; your score names the part to read first. Or follow a path:
Every chapter ends on a task. 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 customers.
Ten checks, starting from memory. If nobody can say what last quarter’s first offer was, you have promotions, not an offer.
Before anyone opens a report, ask three people one question. The founder, whoever buys the media and whoever runs email each write down, in one sentence, the offer a stranger sees before their first order. Then put the three sentences side by side.
Most teams get three different answers. The ads say 20% off. The popup says 15% off plus free shipping. The founder says the offer is the product. Nobody is wrong, and that’s the finding: a brand with three first offers has no first offer. It has a set of promotions that happen to run at the same time.
Then open your store, your ad account and your email platform, and score each check 0 to 2: 0 if it failed or you can’t answer it, 1 if partly true, 2 if clean. Don’t read the bands until all ten are scored.
A brand with three first offers has no first offer.
Score as you go; your band appears when all ten are in.
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.
| Score | What it means | Read next |
|---|---|---|
| 17–20 | You have an offer. Prove what it adds before you change it. | Part five, starting at Test the Offer, Not the Ad |
| 12–16 | The offer exists and leaks after the first order. | Part four, starting at The Offer After the Offer |
| 7–11 | You have promotions, not an offer. | Part three, starting at Price It Before You Run It |
| 0–6 | The first offer is choosing your customers for you. | Two Studies That Disagree, then Find the Door |
Fix zeros before ones, in check order. The early checks feed the later ones: you can’t price an offer until you know which product it sells, and you can’t judge an offer on its cohort until you know what a cohort normally does.
Put your two lowest beside the three sentences from check 1. If the sentences disagree and check 5 is a zero, the brand is paying to acquire customers it has never priced. The rest of this book fixes that in order.
Most teams say offer and mean discount. An offer is five separate decisions, and the fifth is the one almost nobody makes on purpose.
Ask a room what the offer is and someone will say “20% off.” That’s one setting on one of five dials. It’s like being asked what car you drive and answering “blue.”
The confusion is expensive. A brief that says “new customers, 20% off” leaves the creative team guessing which product to show, what to promise and who it’s for. The ads come out generic, the test comes back flat, and the team concludes the discount wasn’t deep enough. The next brief says 30%.
The first four decisions win the first order. The fifth decides whether there’s a second.
Here are offers this book takes apart, laid out on the five decisions. Read down the last column.
| Offer | Product | Door | Mechanic | Next step |
|---|---|---|---|---|
| Dollar Shave Club, 2012 | Razor and blades | A club, not a razor | A dollar a month, plus shipping | The subscription itself: the next box ships unless you stop it |
| Warby Parker Home Try-On, until 2025 | Frames | Try before you pay | Five frames at home, free | A prescription order, then a store nearby |
| Costco | Everything in the warehouse | Membership | A yearly fee, low prices inside | Renewal, which most members do |
| Chewy Autoship | Pet food and supplies | The food your pet already eats | A discount for scheduling repeat delivery | The schedule; most of Chewy’s sales now come from Autoship customers |
| Gallery Furniture, 2022 | A mattress or furniture, $3,000 and up | A Houston Astros promotion | Your money back if the Astros won the World Series | A store visit worth talking about, and a reason to buy now |
Dollar Shave Club and Chewy built the next step into the mechanic. Costco sells the next step as the product. Warby Parker’s next step moved into its stores, which is why it could retire the try-on (chapter 11). Gallery Furniture’s offer, which I helped run while leading retention there, was built for one moment, and its next step was the story people told about it.
Acquisition and retention usually sit on different teams with different scorecards. The paid team is judged on the first order, so it builds the first four decisions. The email team inherits the customer afterwards, so it writes the reorder campaign. Nobody owns the handoff, and the customer meets the full price for the first time in a reorder email written by someone who didn’t choose the first discount.
The fix is organizational before it’s creative: one person signs off on all five decisions at once. That person writes the offer in two sentences (chapter 23 gives the template) and the paid and retention teams both work from it.
Split last year’s new customers by first offer: product, door and mechanic. If customers from your different offers reorder at the same rate over twelve months, the offer isn’t selecting anyone in your category, and you can treat it as a conversion lever only. Most brands that run this find a spread of several points between their best and worst first offers.
Four ideas explain almost every offer that works. None of them is “make it cheaper.”
A stranger buys when what the offer is worth to them clears what it costs them, and the step is easy enough to take right now. Every idea in this chapter is a way of moving one side of that sentence.
Alex Hormozi’s $100M Offers (2021) gives the cleanest version. Value, to the buyer, is the dream outcome times their belief they’ll actually get it, divided by how long it takes and how much effort and sacrifice it asks:
value = (dream outcome × perceived likelihood)
÷ (time delay × effort and sacrifice)
Price sits outside the fraction. It’s what the buyer compares value to. A discount lowers the bar; everything else in the equation raises the value. For physical products, the four parts translate directly:
| Part | Raise it with | DTC example |
|---|---|---|
| Dream outcome | A promise in the customer’s words | “Put together in thirty seconds,” not “premium jewelry” |
| Perceived likelihood | Proof and risk reversal | Reviews from people like the buyer; a fit guarantee |
| Time delay | Speed, or a first result sooner | Rush shipping; a starter size that shows results in a week |
| Effort and sacrifice | Remove steps and decisions | A pre-built bundle; a quiz that picks the shade |
Notice how many of these cost less than a discount of the same pull. A fit guarantee costs you the returns it causes, which (chapter 11) are usually fewer than the sales it adds. A quiz costs a build. A discount costs margin on every order, including the ones that would have happened anyway.
BJ Fogg’s behavior model, B = MAP, says a behavior happens when motivation, ability and a prompt converge. Low motivation can be carried by high ability: a one-tap reorder gets done on a lukewarm Tuesday. High motivation survives low ability for a while, then gives up.
Offers usually attack motivation because that’s what discounts do. Most first-order losses are ability losses: a size chart nobody trusts, a shipping cost revealed at checkout, an account creation wall. Chapter 10 sorts objections into these two piles, because the fix for each is different.
Eugene Schwartz’s Breakthrough Advertising (1966) sorted prospects by what they already know: unaware of the problem, aware of the problem, aware of solutions, aware of your product, and most aware, ready for a deal. A discount speaks to the last group. It says nothing to the first three, who don’t yet want what you’re discounting.
This is why a discount-led first offer tends to recruit people who were already close to buying. It converts them cheaply and tells you little about whether the offer can reach anyone new. It also selects for the kind of customer who responds to discounts, which is the subject of Part two.
Nobody knows what a necklace should cost. They know what it costs compared with something else on the page, in their head, or in last month’s sale email. Dan Ariely’s Predictably Irrational (2008) reports a classroom test built on an old Economist subscription page: web only at $59, print only at $125, print and web at $125.
Nobody chose print only. Its job was to make print and web look like a free extra. The same logic runs through bundles (“the set is $12 less than buying separately”), size ladders (the middle size is chosen because the large exists) and sale prices (the struck-through number is the comparison). It also explains JCPenney: take away the comparison and the honest price looks like no deal at all.
A price is only ever cheap or dear compared with another price.
Daniel Kahneman and Amos Tversky’s work on prospect theory found that losses weigh roughly twice as much as equal gains. For offers, that has one practical meaning: removing a possible loss (“if it doesn’t fit, the return is free”) moves a hesitant buyer more than adding an equal gain (“$10 off”). Risk reversal is loss aversion put to work.
Before you reach for a discount, write the offer’s weakest part. If the promise is vague, the dream outcome is low. If there are no reviews from people like the buyer, likelihood is low. If the first result takes a month, time delay is high. If the checkout asks for an account, ability is low. A discount papers over all four. Fixing the weak part costs less and doesn’t teach anyone to wait.
One found that deep first discounts halve a customer’s value. The other found they raise it. Both are right, and the reason is the most useful idea in this book.
Every argument about discounting ends in anecdotes. Someone remembers the promotion that brought in a wave of customers who never came back. Someone else remembers the launch discount that built the brand. Two field studies, run on real customers over years, show why both memories can be true.
Michael Lewis, a marketing professor, published “Customer Acquisition Promotions and Customer Asset Value” in the Journal of Marketing Research in 2006. He had customer-level records from two businesses: a newspaper that sold subscriptions at different introductory discounts, and an online grocer.
In both, the deeper the discount that won a customer, the less often that customer bought again and the less they were worth over time. His summary is blunt: a 35% acquisition discount produced customers with about half the long-term value of customers acquired without a promotion.
His explanation is uncertainty. Some prospects aren’t sure the product is for them. At full price they don’t try it. A deep discount lowers the cost of finding out, so they try it, and many find out it isn’t for them. The discount didn’t make those customers worse. It let in people who would have correctly stayed out.
Eric Anderson and Duncan Simester ran three large field experiments at a catalog retailer of durable goods, with between 56,466 and 297,405 customers each, and tracked purchases for 22 to 28 months. Some customers got deeper discounts than others, at random. The results, published in Marketing Science in 2004, split cleanly by who the customer was:
| Customer | Effect of a deeper discount on later buying | Why, as the authors read it |
|---|---|---|
| First-time customers | More future purchases | A low first price set favorable expectations about the retailer’s prices |
| Established customers | Fewer future purchases, and cheaper items | They bought ahead and became more price-sensitive |
So a deep first discount can build a better customer, and the same discount shown to an existing customer can make a worse one. The second half of that finding is the one most brands ignore: every sitewide sale that reaches your established customers is teaching them to wait.
Put the two studies next to each other and the difference is what the customer was unsure about.
For most DTC brands, the stranger’s uncertainty is about the product. Will it fit, will the shade match, will it work on my skin, is it worth this much. That’s Lewis territory, and it’s why deep discounts so often produce the wave that never comes back. Where the uncertainty is about value in a category the buyer already understands, like pet food, coffee or razors, a strong introductory price behaves more like Anderson and Simester’s catalog.
A discount answers a question. Make sure it’s the question your customer is asking.
Utpal Dholakia, also in the Journal of Marketing Research in 2006, followed customers of a service business and separated those who joined on their own initiative from those who were induced to join. The self-starters behaved more like relationship customers, and they reacted badly to reminder coupons, which read to them as pressure. Rewards that recognized them, like automatic enrollment in a rewards program, did the opposite.
That matters for the second offer. A customer who found you and paid full price didn’t need a coupon to come in, and sending one may be the thing that teaches them to need it (chapter 17).
Put in what you know, or best guesses, for a full-price first offer and a discounted one. The tool follows both cohorts for twelve months and tells you what repeat rate the discounted cohort would need to break even.
Pull customers acquired in the same quarter, split by the depth of their first-order discount: none, under 20%, 20% or more. If the three groups reorder at the same rate over twelve months, your category behaves like the catalog, not the newspaper, and you can discount the first order more freely, to new customers only.
Marketing spend rose tenfold in two years. The customers it bought were worth less than the ones who came first.
Blue Apron filed to go public in June 2017. Its registration statement is one of the few public documents where a consumer subscription company showed investors how much each year’s new customers spent. Read with the marketing line beside it, it’s a warning about what happens to an offer at scale.
The marketing line is plain. Blue Apron spent $14.0 million on marketing in 2014, $51.4 million in 2015 and $144.1 million in 2016.
The filing also showed average net revenue per customer in each annual cohort’s first six months. As reported from the filing: $402 for customers acquired in 2014, $451 for 2015 and $387 for 2016.
Put the two charts together. Between 2015 and 2016 the company nearly tripled its marketing, and the average customer it acquired spent 14% less in their first six months (derived: $387 against $451). Growth came from buying more customers, and each one was worth less than the last year’s.
At scale, the same offer reaches people who were less sure they wanted it.
The IPO priced in June 2017 at $10 a share, below the $15 to $17 range first set, valuing the company at just under $1.9 billion. In 2023 Wonder Group agreed to buy it for $13 a share, about $103 million in equity value (both reported by CNBC and TechCrunch). There were many reasons, from competition to operations. The cohort line was visible in the filing, and it pointed the same way.
Outside analysts tried to back out what the filing didn’t say. Daniel McCarthy, a marketing professor who models customer bases from public disclosures, estimated from the cohort data that roughly seven in ten new customers were gone within six months, and put acquisition cost in early 2017 at around $147 per customer. Those are estimates from a model, not disclosures; the direction is what matters here.
The filing doesn’t break out what Blue Apron spent on first-box discounts, so this isn’t a claim about any one promotion. It’s a claim about what happens to any first offer as spend rises. The first customers a brand acquires are its most certain: people who wanted exactly this. Every extra dollar of spend reaches someone a little less sure. Lewis’s uncertain customer, from the last chapter, is who you meet at the margin.
If the first offer leans on price, scaling spend recruits more of the people price is persuading, and fewer of the people the product is persuading. The six-month revenue line bends down while the customer count line bends up, and the growth chart looks fine until the cohorts are drawn.
In 2012 a department store replaced its sales with honest everyday prices. It lost a quarter of its sales in a year.
Ron Johnson joined JCPenney as chief executive in late 2011, from building Apple’s retail stores. Early in 2012 he announced a new pricing plan, “fair and square”: lower everyday prices instead of inflated list prices, a small number of planned monthly values, and an end to the constant coupons and sales the store had run for decades.
It was, on paper, more honest. Shoppers would pay about what they’d been paying on sale, without clipping anything. It failed faster than almost anyone expected.
Comparable-store sales fell 25.2% for the year and 31.7% in the fourth quarter, the holiday quarter when a department store makes its year. The company lost $985 million. Johnson was replaced in April 2013 by his predecessor, and the coupons came back.
JCPenney’s shoppers weren’t buying shirts at a price. They were buying the feeling of beating the price. The coupon, the red tag and the struck-through number were part of the product. Remove them and a $20 shirt that used to be “$40, now $20” became just a $20 shirt, with nothing to compare it to and nothing to win.
That’s the decoy idea from chapter 3 at the scale of a whole store. The inflated list price was a decoy. It made the sale price feel like a win. Honest pricing took the decoy away, and with it the reason to come in.
The coupon was part of the product. Remove it and the product changed.
A DTC brand doesn’t run a department store’s coupon calendar, but the mechanism is the same at any size. Every offer teaches a customer what your price means. A welcome discount teaches “never pay full price the first time.” A monthly sitewide sale teaches “never pay full price.” Anderson and Simester measured the second lesson in chapter 4: deeper discounts made established customers buy less afterwards.
Once taught, the lesson is expensive to unteach. JCPenney tried to do it in one year with its whole customer base, and paid about $4.3 billion in lost annual sales for the attempt (derived from the two results above). A brand that wants to wean customers off discounts has three safer options:
Your best seller is not always your best first product. The door is the product whose first-time buyers come back.
Every catalog follows a lopsided rule: a few products carry most of the revenue. Most teams stop there and put the best seller in the ads. The better question is narrower: which product do strangers buy first, and which of those first purchases turn into customers?
Pull every customer’s first order for the last twelve months and rank products by how many first orders contain them. Do the same by units and by revenue. Three things usually jump out:
Before you trust the ranking, check whether it’s circular. If a product leads first orders only because it’s the only one you’ve ever advertised, the ranking measures your media plan, not demand. Look at first orders that came from organic search, direct traffic and email signups. If the same products lead there, the demand is real.
Now take the same customers and, for each first product, measure what share ordered again within 180 days, and what they spent in those 180 days. This is the table that decides your first offer.
| First product | First orders | Reordered in 180 days | 180-day revenue per customer |
|---|---|---|---|
| Picture: the starter set | 3,100 | 31% | $118 |
| The best seller | 2,400 | 19% | $104 |
| The refillable | 900 | 44% | $141 |
| The gift box | 1,500 | 9% | $71 |
Made-up numbers, but a common shape. The best seller brings volume and leaks. The refillable brings fewer strangers and keeps almost half. The gift box brings buyers who were shopping for someone else and never meant to come back, which is fine if you know it and price it that way.
Plot each first product on two axes: how many first orders it wins, and how often those customers come back. Four kinds of door fall out.
| Customers come back | Customers don’t | |
|---|---|---|
| Wins many first orders | The door. Scale it. Build the first offer here. | The leaky door. Keep it, and fix the next step: a refill, a pairing, a reason to return. |
| Wins few first orders | The hidden door. Test it as a first offer before anything else. | The wall. Take it out of acquisition. Sell it to existing customers. |
Scale the door. Fix the leaky door. Test the hidden door. Stop advertising the wall.
The hidden door is where most of the upside hides. A product with a small first-order count and a high return rate often just hasn’t been shown to strangers. Give it a real test, with enough creative behind it that the creative can’t be blamed (chapter 20).
Sometimes the product strangers want most is the one they leave after. That tension is a business decision, not a data problem, so surface it rather than settling it quietly. Put both numbers in front of the founder: the leaky door acquires customers at a lower cost; the hidden door acquires fewer customers worth more. Then price both with the tool in chapter 9, which includes the reorders.
If the door is expensive, the first offer may need to be its affordable sibling: a smaller size, a single instead of a set, a starter kit. The job of the first offer isn’t to sell the flagship. It’s to get a stranger through the door that leads to the flagship. Chapter 14 covers trials and samples, and the one-dollar door that built a razor company.
Average lifetime value hides the most important fact about most DTC brands: most customers never order twice.
A founder asks for lifetime value at 90 days, 180 days, a year, and all time. Someone runs a report, four numbers come back, and the room nods. Often all four are wrong, and even the right ones answer the wrong question.
I’ve seen the same mistake on more than one client account. A store’s reporting tool is asked for “90-day, 180-day, 365-day and lifetime LTV.” What comes back is four calendar windows stacked back from today: revenue from customers active in the last 90 days, the last 180, and so on. Customer counts rise as the windows get longer.
That’s the tell. In a real cohort table, the count should fall as you require more age: fewer customers have had 365 days since their first order than have had 90. If the counts go up, you’re looking at windows, not cohorts, and the averages mix new customers with old ones.
The fix is to rebuild from a customer export that has each customer’s first-order date, order count and lifetime spend. Group by first-order month. Then you can see the real curves. Appendix A has the fields.
Once the data is clean, split customers by how many orders they’ve placed. On most DTC brands, the shape is the same: a large majority ordered once, and a minority of repeat buyers carries a disproportionate share of revenue.
| Picture | Customers | Share of customers | Revenue | Share of revenue |
|---|---|---|---|---|
| Ordered once | 8,400 | 70% | $620,000 | 34% |
| Ordered twice or more | 3,600 | 30% | $1,180,000 | 66% |
Made-up numbers, typical shape. The average customer here is worth $150. Nobody is worth $150. The one-timers averaged about $74; the repeat buyers about $328.
Manage the one-and-done rate. The average is a blend of two different businesses.
The one-and-done rate is partly a retention problem, which The Second Order covers. It’s also an acquisition problem, because the first offer chose who is in that 70%. A first offer that brings buyers who were never going to return raises the rate no matter how good your flows are. The door table from the last chapter tells you which first products feed it.
It’s also the clearest way to price the second offer. Every one-time buyer you turn into a two-time buyer is worth roughly the difference between the two groups’ averages, before margin. That turns “improve retention” into a number per customer, and it tells you how much a second offer can afford to give away (chapter 15).
From my work: the same split usually shows a second fact. Revenue among repeat buyers is lopsided too. A small top slice of customers, often around a tenth, carries a large share of all revenue, and the very top few hundred people can matter more than whole acquisition months. That produces two retention motions with different economics, and they shouldn’t share a calendar, a creative team or a scorecard.
Mixing them is how brands end up sending their best customers the winback code meant for people who left. It’s also how the one-timer motion gets starved: the top customers produce most of the revenue, so they get most of the attention, and the largest group in the file gets a monthly newsletter.
The first offer feeds one of these motions more than the other. A first offer that brings collectors feeds the top; a first offer that brings deal-seekers feeds the one-timers. Knowing which one your offer feeds is what lets the retention team plan for the customers acquisition is actually sending them.
| One-timers who order again | Customers | Second-order revenue | Contribution |
|---|
If fewer than half your customers have ordered only once, or one-time buyers bring more than half your revenue, you’re a repeat-purchase business already. Your offer work should move to the second and third orders, starting with Make the Next Order the Default.
An offer can win on ROAS and lose money on every customer it brings. Do the arithmetic on paper, where it’s cheapest.
Platform ROAS divides revenue by ad spend. It doesn’t know what the product cost, what shipping cost, how many orders came back, or whether the customer ever ordered again. An offer judged on ROAS alone will reliably pick the one that gives the most away.
The number to price an offer on is what the first order leaves after every variable cost, per new customer:
first-order contribution =
price paid after discount
− product cost
− pick, pack and shipping
− payment fees
− expected refunds and return costs
Put your acquisition cost next to it. If first-order contribution covers acquisition cost, the offer pays for itself on day one. If it doesn’t, the gap has to come back in reorders, and you need to know how many and how soon.
ROAS tells you the ad worked. Contribution tells you the offer did.
Divide nothing; subtract. The most you can pay to acquire a customer on this offer and break even on the first order is the first-order contribution. The most you can pay and break even over twelve months adds the contribution from expected reorders. The gap between the two is what retention is worth to your media buyer, and it should be written into the targets they’re given.
Targets have to move when the offer moves. Switch the core offer from a $30 single to a $70 set and the right acquisition cost roughly doubles. Judge the new offer against the old target and you’ll kill it. Every offer change should trigger a target review before anyone reads the results.
Rank your last three offers by first-order ROAS, then by 180-day contribution per new customer after acquisition cost. If the two rankings match, ROAS is a good enough proxy in your business and you can judge offers faster. In accounts IIn accounts I’ve worked on, they often don’t.rsquo;ve worked on, they often donIn accounts I’ve worked on, they often don’t.rsquo;t.
Most people who don’t buy aren’t waiting for a lower price. Find what they’re actually waiting for, and answer that.
A discount answers one question: is this worth the money? It’s the right answer when that’s the question. When the question is “will it fit,” “is this company real” or “can I be bothered,” a discount is a bribe to ignore the doubt, and the doubt comes back as a return or a customer who never reorders.
You don’t have to guess. Customers write their objections in five places, and almost nobody reads them together:
From my work: when I’ve built voice-of-customer guides for brands, the most useful output was never the list of what customers love. It was the objection map: each doubt, what’s underneath it, and what proof answers it. Price almost always turned out to sit on top of something else, usually doubt about quality or fit that a lower price didn’t fix.
| What they say | What’s underneath | Mechanic that answers it |
|---|---|---|
| “Will it work for me? Will it fit?” | Risk of being wrong | Guarantee, free exchanges, try before you pay, fit tools |
| “Is this company real?” | Trust | Reviews from people like them, visible policies, a founder, a phone number |
| “I don’t know if I’ll like it” (taste, scent, shade) | Unknown experience | Samples, a variety kit, a small size |
| “Not worth that much” | Value doesn’t clear price | Bundle, gift with purchase, clearer proof of quality |
| “Too expensive” | Can’t justify the spend right now | Introductory price, payment plan, a smaller entry product |
| “I’ll do it later” | No reason to act now | A real deadline: a launch window, a season, a limited run |
| “Too much hassle” | Effort | Free or fast shipping, easy returns, fewer steps, a pre-built set |
This is Fogg’s model from chapter 3 put to work: the top three rows lower risk, the bottom row raises ability, and only the middle rows are about price. Most brands answer every row with the middle ones.
A discount is a bribe to ignore a doubt. The doubt comes back as a return.
A guarantee usually sells more than it costs. The evidence says so. Warby Parker, Zappos and a Houston furniture store show when to retire one, and how to pay for one.
Loss aversion, from chapter 3, says a possible loss weighs about twice an equal gain. A stranger looking at your product is weighing the chance it disappoints: wrong size, wrong shade, not as pictured, a hassle to send back. Risk reversal moves that loss from the buyer to you.
Narayan Janakiraman, Holly Syrdal and Ryan Freling pooled 21 studies of return policies in a 2016 meta-analysis in the Journal of Retailing. Overall, more lenient return policies increased purchases more than they increased returns. The detail is more useful than the headline, because they split leniency into five parts:
| Kind of leniency | Example | Effect found |
|---|---|---|
| Money | Full refund, no restocking fee | More purchases |
| Effort | Prepaid label, no questions | More purchases |
| Time | A longer return window | Fewer returns |
| Exchange | Easy swaps for size or color | Fewer returns |
| Scope | Accepting worn, used or sale items | More returns |
The surprising row is time. A longer window reduced returns, probably because a customer with no deadline stops thinking about sending it back. The expensive row is scope. So the cheapest strong guarantee is usually: full refund, easy label, a long window, easy exchanges, and a clear line on what counts as returnable.
A long return window lowers returns. A loose one on what can come back raises them.
For years Warby Parker’s first offer was Home Try-On: pick five frames, try them at home, send them back, pay nothing. It answered the one objection that kept people from buying glasses online, which is not knowing how they look on your face. The door was the try-on; the next step was a prescription order.
In August 2025 the company said it would end Home Try-On by the end of the year. Its reasons, as reported by Retail Dive: most people using it now lived within 30 minutes of one of its roughly 300 stores, and virtual try-on had improved. The objection hadn’t gone away. The answer to it had moved into a store down the road and a camera on the phone. It took $2.5 million of inventory write-downs and $1.3 million of restructuring costs to close the program in the quarter.
That’s the rule for any risk-reversal mechanic: it’s tied to an objection. When something else answers the objection better or cheaper, retire the mechanic, even if it built the brand.
Zappos built its name on free shipping both ways and a 365-day return window. As of September 2026, its returns page says merchandise must be returned within 60 days, for a refund to the original payment or store credit, with store credit possible beyond that in limited cases. An archive comparison reported the change happening in July 2025. The long window was a first-order promise from an era when buying shoes online was itself the risk. That risk is smaller now, and returns are expensive.
Note what the meta-analysis would predict: shortening the window could raise returns, not lower them, as customers act before the deadline. A brand shortening its window should measure return rates before and after, not assume the savings.
Gallery Furniture in Houston, where I led retention, has run a different kind of risk reversal for years. In 2022, as Forbes described it, customers who spent at least $3,000 on qualifying mattresses and made-in-America furniture got their money back if the Houston Astros won the World Series, with double money back on purchases made in the early summer. The owner, Jim McIngvale, hedged the promise with futures bets on the Astros at long odds, placed across several sportsbooks, sized so that a Houston win paid enough to cover the refunds. The Astros won. The customers got their money back, and the bets paid it.
Three things make it worth studying beyond the headlines:
You don’t need a sportsbook to borrow the structure. A guarantee tied to an outcome the customer cares about, priced in advance, with a deadline and a reason to talk about it, is available to any brand. Have counsel read anything that ties a refund to a chance event: prize, chance and purchase together can make a lottery.
| Guarantee | Answers | Watch for |
|---|---|---|
| Money back, no questions | “Will it work for me?” | Scope creep: define returnable condition |
| Free exchange | “Will it fit?” | Keeps the sale; cheaper than refunds |
| Try before you pay | “How will it look on me?” | Logistics cost; retire when stores or tools answer it |
| Results guarantee | “Will it actually do what it says?” | Define the result and the time; claims law |
| Event-contingent refund | “Why now?” and “is it worth it?” | Hedge the cost; lottery law |
The best bundle is already in your order data. And a buy-one-get-one never costs what it looks like it costs.
A bundle is the cheapest way to raise value without cutting price, when it’s the right bundle. The wrong bundle is a clearance box with a bow on it, and customers can tell.
When one unit can’t clear your economics, put quantity into the offer: 1, 2 or 3 units at falling unit prices, with 2 selected by default. It’s the decoy from chapter 3 again: the 3-pack makes the 2-pack look sensible. The page has to match. Sending a multi-unit offer to a single-unit product page makes the customer do the math you should have done.
A buy-one-get-one sounds like 50% off. It isn’t. The second unit adds its product cost and a pick fee; it doesn’t add a second shipment or a second payment fee. Put the options side by side before you pick one:
| Offer | Customer pays | Left per order | Really costs you |
|---|
With these example numbers, buy-one-get-one leaves more per order than 30% off a single unit, while looking like a bigger deal to the customer. That’s the bundle’s advantage when product cost is low against price. It flips when product cost is high: a $40 item that costs $25 to make can’t afford to give a unit away. Run your own numbers.
Customers are very sensitive to shipping charges. Free shipping sells. It can also cost more than it sells.
To a customer, shipping isn’t a separate line. It’s part of the price, discovered late. That’s why the cart-to-checkout step loses so many people: the price just went up.
Michael Lewis, Vishal Singh and Scott Fay studied an online retailer that had tried many different shipping-fee schedules, and published the results in Marketing Science in 2006. Their summary has both halves of the lesson in it. Customers were very sensitive to shipping charges, and fees changed both whether people ordered and how much they put in the basket. Free shipping, and free shipping above an order threshold, were very effective at generating additional sales. But the shipping revenue given up, and the fact that several customer segments barely responded, were large enough to make those promotions unprofitable for that retailer.
Free shipping sells. Whether it pays is a separate question.
The second half is the one most brands never check. A free shipping promotion is paid on every order, including orders from customers who would have paid for shipping. The segments that don’t respond still collect the benefit.
A free-shipping threshold is an offer in its own right: it tells the customer how much to buy. Three rules from my work on thresholds:
If you can afford only one, look at your objection map from chapter 10. If the top objection is fit or quality, free returns and exchanges answer it; free shipping doesn’t. If the top objection is hassle or cost at checkout, free shipping above a threshold answers it. Many apparel brands pay for free shipping and charge for returns, which is the reverse of what their customers are worried about.
The retailer in their study learned because it varied the schedule and watched order rates and basket sizes. You can do the same with a holdout: show a threshold offer to a random share of new visitors, keep the rest on the old schedule, and compare contribution per visitor, not conversion rate. Chapter 20 has the evidence bar.
A trial lowers the price of finding out. It also selects the people who like to try things. Build the path from the trial to the full price before you launch it.
In 2012 a company called Dollar Shave Club launched with a video and an offer that was mostly its name: razor blades delivered for a dollar a month, plus shipping. Four years later Unilever bought it for about $1 billion in cash.
The dollar wasn’t a discount on a razor. It was the price of a club, and the offer’s five decisions (chapter 2) all pointed the same way:
It fits chapter 4’s split. The buyer’s uncertainty wasn’t about the product; everyone knows what a razor does. It was about whether this was a better deal than the drugstore. A low price answered exactly that question, which is Anderson and Simester’s catalog, not Lewis’s newspaper.
In 2023 Unilever sold a majority of Dollar Shave Club to Nexus Capital Management and kept a 35% stake, as Retail Dive reported. By then the brand was also sold through Walmart and Target, well beyond the subscription it launched with. A subscription door gets a customer in; it doesn’t guarantee that every future customer wants to come in that way.
| Trial | Answers | Selects for | Next step to build |
|---|---|---|---|
| Sample or small size, paid | “Will I like it?” | People curious enough to pay a little | A credit for the sample’s price toward the full size |
| Free plus shipping | “Is it worth trying at all?” | Collectors of free things | A reason to pay for the second order, set up in the first box |
| Low first month of a subscription | “Is this a better deal?” | Deal-sensitive buyers in known categories | A second-month price that isn’t a shock |
| Try at home, pay later | “How will it look on me?” | Serious buyers with a fit problem | The purchase itself; retire when stores or tools answer the fit |
Every trial selects someone. Decide who before you pick the trial.
The most common failure is the second invoice. A customer who joined at a dollar meets the real price in month two, and a large share leave. That cliff is the first offer’s fault, not the retention team’s. Two ways to soften it:
Anything that renews automatically falls under auto-renewal and negative-option laws that differ by state and change often. Show the renewal terms clearly next to the price, make cancelling easy, and have counsel read the checkout before launch.
The second offer is where most first offers get paid for or written off. Plan it before the first one runs, and price it for the people who would have reordered anyway.
A customer who came in at 25% off meets your full price for the first time when they go to reorder. If nobody planned that moment, the retention team writes a campaign with another code in it, and the customer learns that every order comes with one.
The right moment is when a customer is about to run out, lose interest or forget you, whichever comes first. You can read it from your own data:
Chapter 7’s door table gives you the split. The refillable door has a short, reliable clock. The gift box door may have no clock at all, and its second offer is a different product for a different person.
A second-order discount goes to everyone who uses it, including the customers who were going to reorder at full price. If 20 of every 100 customers reorder without an offer, and a 15% offer lifts that to 26, you’ve bought 6 extra orders and given 15% off 26. Whether that pays depends on numbers most teams never put side by side. The tool does.
A reorder discount is paid on every reorder, including the ones you already had.
With the example numbers, a 15% offer that lifts reorders from 20% to 26% loses money: it gives away more on the 20 reorders you already had than it earns on the 6 new ones. The same offer aimed only at customers unlikely to reorder on their own, past the point where the curve flattens, looks very different. Target the discount at the lapsing, not the loyal.
The strongest second offer isn’t an offer at all. It’s an arrangement where reordering is what happens unless the customer decides otherwise.
Two of the most durable retail businesses in America sell the next order in advance. One charges a fee for the right to shop. The other gives a discount for scheduling the next delivery. Both turn the second order from a decision into a default.
A Costco membership is a first offer with the next step built in: pay the fee, and every visit afterward is cheaper than it would be elsewhere. The renewal is the second offer, and the customer has already been paid, visit by visit, to accept it. In the fourth quarter of fiscal 2025, as reported from the company’s results, 92.3% of members in the U.S. and Canada renewed, and 89.8% worldwide. Membership fee income in that quarter was $1.72 billion, up 14% on the year.
The fee does something a discount can’t. It makes the customer want to use what they paid for. Every trip to Costco is partly a trip to get the fee’s worth.
Chewy’s Autoship offer is a discount for scheduling repeat delivery of pet food and supplies the pet was going to need anyway. In fiscal 2025, Chewy reported net sales of $12.60 billion, and $10.50 billion of it, 83.3%, came from Autoship customers. A year earlier the share was 79.2%, as Subscription Insider reported from the filing.
Pet food is the ideal category for a default: the need is certain, the timing is predictable, and the cost of running out is a hungry animal. The offer makes the obvious next order effortless. Fogg’s ability term, from chapter 3, is doing almost all the work.
A default beats a discount because the customer doesn’t have to decide again.
| Default | Fits when | Watch for |
|---|---|---|
| Subscribe and save | The customer runs out on a predictable clock | The schedule must match real use, or boxes pile up and they cancel |
| Membership fee | Customers buy often enough to feel the fee paid back | Members who don’t use it resent it at renewal |
| Replenishment reminder | Use varies too much to schedule | A reminder is a prompt, not a default; keep it one tap to reorder |
| Collection or set | Customers build over time, like jewelry, figures, books | Show what they own and what’s missing, never just the newest item |
An existing customer who gets a discount learns to wait. One who gets currency, access or recognition learns to come back.
Anderson and Simester’s established customers bought less after deeper discounts. Dholakia’s self-starters reacted badly to reminder coupons and well to being recognized. Put together, the research says the people who already chose you should be rewarded in a different currency from the people you’re trying to win.
| Reward | What it costs you | What it teaches |
|---|---|---|
| Percent off | Margin on every order that uses it, including orders already coming | “The real price is lower. Wait for the code.” |
| Points or store credit | Product at your cost, only when redeemed, and only for what’s redeemed | “Coming back is how you collect.” |
| Early or exclusive access | Almost nothing, if the product sells anyway | “Being a customer gets you in first.” |
| A gift at a milestone | One product at cost, at a moment you choose | “They noticed.” |
Points and credit are the same idea Close the Loop calls a currency you can print. A $10 credit costs you the product cost of whatever it buys, and nothing at all when it isn’t redeemed. A 10% code costs 10% of revenue on every order it touches.
A discount spends margin. Points spend a currency you print.
From my work: in a retention plan I built for a jewelry brand in 2026, every lifecycle email carried one value module, and it had two states. Most customers saw their loyalty points balance, framed as something to spend (“you’re sitting on 340 points; they’d look better as jewelry”). Only customers the platform scored as high churn risk, and the last touch of a sunset flow, saw a discount code instead. One switch on the profile decided which state rendered.
The design point is the switch. Discounts don’t disappear; they become a tool for the customers who need one, instead of a habit for everyone.
The most common mistake in loyalty programs is paying the best customers the most margin. Top customers were buying at full price before the program, which makes them Anderson and Simester’s established customers in their purest form. Give them access, recognition, service and first look. Give the discount to the customers who haven’t decided yet.
A lapsed customer isn’t waiting for 20% off. They’re waiting for a reason. Show what changed before you show what’s off.
The standard winback flow is three emails at 60, 90 and 120 days with escalating discounts. It recovers some customers, gives the largest discount to the least loyal, and teaches the whole file that the way to get the best price is to disappear for four months.
From my work, the same jewelry plan had to wake up customers who had lapsed as far as a year back. Waking a whole file at once is how a sender reputation gets damaged, so the plan sent in waves, warmest first: recently active on the site and still opening email; then active but not opening; then older buyers; then older non-buyers who had engaged. Each wave kept a random tenth as a holdout that received nothing, so the result could be measured instead of claimed.
Between waves the plan had stop rules, written in advance: spam complaints under 0.1%, bounces under 1.5%, unsubscribes under 0.6%. Breach one, and the next wave pauses and shrinks. Google’s bulk sender guidance asks senders to keep spam complaints below 0.1% and never reach 0.3%, which is why the first threshold sits where it does.
A winback without a holdout is a guess about who would have come back anyway.
More than half of one big retailer’s promotions lost money. A sale is an offer to your whole file at once, so plan it like one.
Kusum Ailawadi and colleagues studied about 36 million promotions at CVS, across 189 categories and 3,808 stores, and published the results in the Journal of Marketing Research in 2006. More than half of the promotions weren’t profitable, because the promotional margin was so much lower than the regular margin. Only about 45% of the extra sales a promotion produced were truly incremental. The rest were sales that would have happened anyway, moved into the discount.
There was a bright spot: promotions had a significant positive halo on sales in other categories. People who came in for the deal bought other things. For a DTC brand, that halo is the second product in the cart and the new customer who arrives during the sale. It’s the reason to run sales deliberately, not the reason to run them constantly.
From my work, this is the sequence I use for a promotion that runs about three weeks, including a holiday. Each send goes to a narrower, more interested audience than the one before, so the most interested people hear the most and the least interested hear the least.
| Send | Who gets it | The job |
|---|---|---|
| 1 | Everyone engaged, split by what they bought before | Launch. Lead each group with the offer closest to what they already buy. |
| 2 | People who didn’t open send 1 | Same email, new subject line and preview, sent at the opposite time of day. Never change the offer. |
| 3 | Opened and clicked, didn’t buy | Restate the value in one sentence, remove the friction, add soft urgency. |
| 4 | Non-buyers, split by which offer they looked at | Separate the offers. One email per offer, each with its own argument. |
| 5 | Anyone who opened in the last 30 days | A timing email before the holiday: set it up now, not a hard sell. |
| 6 | High intent only: clicked twice, viewed checkout, added to cart | Firm and calm. The window is closing; standard pricing returns. |
| 7 | All non-buyers | One week left. One offer per email, a clear deadline. |
| 8 | Opened any send, didn’t buy | Forty-eight hours left. Short. No storytelling. |
| 9 | Highly engaged only, once | Closes tonight. One call to action. |
Count what each person receives: the least engaged get three or four emails over the event, the warm get five or six, the high-intent six or seven. Buyers drop out of the ladder the moment they buy. The ladder is a first offer, too: new subscribers who arrive during the sale see it, and the ones who buy are a cohort worth tagging, because chapter 4 says they may behave differently.
The most interested hear the most. The least interested hear the least.
A creative test asks which way to say it. An offer test asks what to say. Mixing them teaches you nothing about either.
Two ads run side by side. One says 20% off the starter set with a sunrise video. The other says free shipping on the bestseller with a founder talking to camera. One wins. What did you learn? Nothing you can use, because the product, the mechanic, the promise and the creative all changed at once.
One ad per offer is a creative test wearing an offer test’s name.
Run offer tests in a separate lane from your scaling campaign, with its own budget per offer, so the platform can’t pour the whole budget into whichever offer got lucky in the first two days. Run them on the audiences you plan to scale into. An offer that wins with past visitors and existing customers tells you little about strangers, and existing customers shouldn’t see new-customer offers at all (chapter 19).
On-site, the cleanest test is a holdout: a random share of new visitors sees the new offer, the rest see the old one, and you compare contribution per visitor. Randomization does what a platform comparison can’t: it makes the two groups alike before the offer touches them.
Ad platforms don’t split traffic randomly; they chase early winners. Treat platform comparisons as structured evidence, not experiments. Before calling a winner, want something like 30 to 50 new customers per offer at the very least, and a gap that holds up across a full week. Below that, the honest verdicts are “extend,” “stop for business reasons,” or “decide on judgment and say so.” Judgment is fine. Judgment dressed up as a significant result isn’t.
Every offer test gets five lines, written before launch: what changes, which of the five decisions it is, what it has to beat, what you expect and why, and what you’ll do in each outcome. If a result wouldn’t change your plan, don’t run the test. Appendix B has the template.
An offer isn’t finished when the first order lands. Judge it again at 90 and 180 days, on new customers only, with your own data.
The campaign report closes when the attribution window does. The customers the offer brought are just starting. Every offer in this book has been judged by what its customers did afterwards, because that’s where the first offer’s real price shows up.
| When | What you read | What it can tell you |
|---|---|---|
| Week 2 | New customers, first-order contribution, acquisition cost | Whether to keep running it while you wait |
| Day 90 | Second-order rate, return rate, contribution after acquisition cost | Whether it brings customers or buyers |
| Day 180 | Revenue and contribution per customer, one-and-done rate | Whether it earned its place as the core offer |
The first order tells you the offer converted. The 180th day tells you what it bought.
One row per first offer, updated monthly: start date, new customers, first-order contribution, acquisition cost, 90-day second-order rate, 180-day contribution per customer after acquisition cost, and existing customers who used it. After two quarters, this table is the most valuable document your growth team owns. It’s how you know which study in chapter 4 your brand lives in.
Five first offers, each half-funded, teach you nothing and bring five different kinds of customer. Back one, test beside it, and let it earn its replacement.
The pull toward breadth is constant. The catalog has thirty products, every launch wants its own campaign, the sale runs next to the evergreen offer, and the account fills with offers that each get a little budget and a couple of ads. It feels thorough. It’s the fastest way to learn nothing.
One first offer brings one kind of customer, and one kind of customer can be given a second offer that fits.
Six gates every first offer passes before it gets the budget, and the two sentences that come out the other side.
By now you have a door table, an objection map, a payback number and a second offer. The scorecard puts them in one place. An offer that fails a gate doesn’t get the core budget; it goes back to the chapter that gate came from.
The output of the scorecard is two sentences anyone on the account can repeat: what the first offer is on all five decisions, and why you believe it wins. If the media buyer, the retention lead and the founder can’t all say it, the offer isn’t settled, and the ads, the landing page and the welcome flow will each say something different. That’s where chapter 1 started.
If three people can’t say the offer in two sentences, there isn’t one.
Four weeks from pulling data to a live test with a holdout, and a date on the calendar to judge it.
This is the order that works: data before design, design before creative, a test before a rollout, and a judgment date before launch. Each week ends with something written down.
The test ends in four weeks. The judgment ends at day 180.
Then leave it alone long enough to learn. The core offer earns its keep in the cohort, and the cohort takes months to read.
Eight categories, each with the first offer that usually fits, the next step it should set up, and the mistake most brands in it make.
Every category has a stranger with a typical doubt and a customer with a typical clock. The first offer should answer the doubt; the next step should match the clock. These are starting points from the patterns in this book, not rules. Your door table and objection map outrank them.
Coffee, supplements, pet food, household refills. The stranger knows what the product does and wonders whether yours is better or better value. That’s the catalog in chapter 4: a strong introductory price can build a good customer here, as long as the reorder price is no surprise.
Shade, scent, skin type, results. The stranger’s doubt is about the product on them, which is Lewis territory: a discount admits people who find out after that it doesn’t suit them, and they leave, sometimes with a return.
Fit and quality against the photo. Returns are the category’s tax. The Janakiraman meta-analysis in chapter 11 points to free exchanges and a long window as the cheapest strong guarantee: they raise purchases and don’t raise returns the way a loose scope does.
Meaning, style and whether it will look like the picture. The best customers are collectors, and the first piece is the first entry in a collection.
Mattresses, furniture, luggage, appliances. The first order may be the only order for years. Price the offer on the first order alone; any reorder is a bonus.
The buyer isn’t the user. The giver worries about delivery date, packaging, and whether the recipient will like it. The recipient never saw your ad.
The stranger fears being stuck. The first offer and the next step are the same thing, which is why the second invoice is where the offer is judged.
Fashion, collectibles, sneakers, small-batch goods. The drop is the offer, and a real deadline answers “I’ll do it later” better than any discount.
Answer the category’s doubt with the first offer. Match its clock with the next step.
What whoever owns the first offer needs on the first day.
Whoever owns the first offer, a hire, a promotion or you, needs five things on day one.
The books and papers this one leans on, and what to take from each.
And the research: Lewis (2006) on acquisition discounts and customer value; Anderson and Simester (2004) on promotion depth for new and established customers; Dholakia (2006) on how customers arrive; Ailawadi and colleagues (2006) on promotion profitability at CVS; Lewis, Singh and Fay (2006) on shipping fees; Janakiraman, Syrdal and Freling (2016) on return policies. Full references are in Appendix C.
What most offer writing leaves out is what this book tried to add: judging the offer by the customer it brings, and planning the second offer before the first one runs.
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, and Close the Loop, on getting customers to bring the next customer.
As Senior Director of Growth and Retention Marketing at Gallery Furniture, he rebuilt the customer journey and the sales playbooks together, and worked on the promotions described in chapter 11. 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 marked “from my work” come from 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 building a first offer, including those looking for someone to own it. Write to andrew@growthlegend.com or message him on LinkedIn.
The fields an offer needs, and five pulls that build every table in this book.
Every pull below runs on an orders table, an order-lines table and a customers table. Resolve identity first: customer ID, then normalized email, then phone. A customer who checked out as a guest and later made an account is one customer, and counting them twice turns a repeat buyer into two one-timers.
| Field | On | Written when | Rule |
|---|---|---|---|
first_order_at | Customer | First paid order | Set once, never overwritten |
first_offer | Customer | First paid order | The offer that won them: code, landing page or campaign, mapped to your offer list |
first_products | Customer | First paid order | Product IDs in the first order |
first_discount_pct | Customer | First paid order | Discount as a share of the full-price subtotal |
holdout_digit | Customer or visitor | Profile creation | Random 0 to 9, set once, never recomputed |
existing_customer_at_order | Order | Every order | True if the customer had a prior paid order; used to count cannibalized codes |
-- first products and what their buyers did next
WITH firsts AS (
SELECT customer_id, MIN(ordered_at) AS first_at
FROM orders WHERE status = 'paid'
GROUP BY customer_id
HAVING MIN(ordered_at) <= CURRENT_DATE - 180
),
first_lines AS (
SELECT f.customer_id, l.product_id
FROM firsts f
JOIN orders o ON o.customer_id = f.customer_id AND o.ordered_at = f.first_at
JOIN order_lines l ON l.order_id = o.order_id
),
after AS (
SELECT f.customer_id,
COUNT(o.order_id) FILTER (WHERE o.ordered_at > f.first_at
AND o.ordered_at <= f.first_at + 180) AS reorders,
SUM(o.net_revenue) FILTER (WHERE o.ordered_at <= f.first_at + 180) AS rev_180
FROM firsts f JOIN orders o ON o.customer_id = f.customer_id
WHERE o.status = 'paid'
GROUP BY f.customer_id
)
SELECT fl.product_id,
COUNT(DISTINCT fl.customer_id) AS first_orders,
AVG((a.reorders > 0)::int) AS reorder_rate_180,
AVG(a.rev_180) AS revenue_180
FROM first_lines fl JOIN after a USING (customer_id)
GROUP BY fl.product_id
ORDER BY first_orders DESC;
A customer whose first order contains two products counts toward both. That’s intended: you want to know what each door leads to. For a clean split, rerun it on single-product first orders only and compare.
An offer thesis, a test brief, a guarantee and a second-offer email to start from. Edit them to your voice.
OFFER THESIS Our first offer is [product], sold as [door], at [price and mechanic], promising [promise]. It wins because [evidence], and it leads to [next step]. FIVE DECISIONS Product: [what they buy first] Door: [brand / collection / hero product / use] Mechanic: [price, discount, bundle, gift, trial, guarantee] Promise: [in customer words, from reviews] Next step: [what happens at day __, and how] NUMBERS First-order contribution: $__ Break-even acquisition cost: $__ first order, $__ 12 months Payback: __ months Top objection it answers: [from the objection map] JUDGMENT DATES Week 2: __/__ Day 90: __/__ Day 180: __/__ Replace it when: [the agreed rule]
Change: [the one decision that differs: product, door,
mechanic, promise or next step]
Beat: [the current core offer, on new customers and
first-order contribution]
Bet: [what you expect to win, and why, in one line]
Evidence: [30+ new customers per offer, a full week, 5+
creatives per offer]
Then: [if it wins / if it loses / if it's unclear]
Try it for [60] days. If it isn't right, send it back for a full refund or swap it for a different [size / shade] free. We'll email you a prepaid label. No questions, no restocking fee. Returns must be [unworn / at least half full / in original packaging]. Final-sale items are marked on the product page.
Subject: Time For Your Next [Product]? Preview: most people reorder around now [First name], you ordered [product] [N] days ago. Most people who use it every day run low about now. Your refill is one tap: [link] It ships [date] and arrives by [date]. You also have [points] points. That's [$ value] toward it, already applied if you use the link above.
Every external source, by chapter. Web sources were read in September 2026.