Retention for ecommerce is repeat purchase—not SaaS churn, MRR, or DAU/MAU. This playbook covers how to read cohort curves, when to fund retention versus acquisition, and the levers that actually produce a second contributing order.
Key takeaways
- Score ecommerce retention with repeat purchase rate, frequency, and cohorts—not MRR unless you are truly subscription-first.
- A cohort curve that dies after month one means LTV is almost first-order revenue; do not scale CAC on that story.
- If payback needs a second order you do not get, acquisition spend is the wrong lever.
- Email/SMS, replenishment, and post-purchase experience beat a points program that funds perpetual discounts.
- Retention is how first-order unit economics become customer-level profit.
Retention for ecommerce is repeat purchase, not churn theatre
Software companies talk about churn, MRR, and DAU/MAU because the product is on all the time. A store is different. The customer does not “log in.” They either place another order or they do not. The native metrics are repeat purchase rate, purchase frequency, time-to-second-order, and the shape of a cohort curve. Refund rate belongs here too: a “retained” buyer who returns the second order was not retained in cash. See refund rate.
If you run a genuine subscription, you may still watch churn. Even then, report it next to skip rate and time-to-second-box, not as a SaaS slide pasted onto a DTC deck. Queries like “growth and retention” and “retention-modelling” on this URL are looking for that ecommerce framing.
Retention sits beside monetization and after the first successful order. It is also the second half of unit economics: first-order contribution answers whether checkout helped; retention answers whether the buyer helped after ads.
How to measure repeat purchase rate
Pick a cohort: every new customer whose first paid order landed in a calendar month. Measure what share placed a second paid order within 30, 60, and 90 days. That is repeat purchase rate. Do this by channel (paid social, branded search, email, organic) or you will average a healthy branded cohort with a cold-prospecting cohort that never returns.
Avoid proxies. “Returning visitors” in analytics includes people who compare prices and leave. “Returning customer” in a platform report sometimes includes anyone with an account and no second order. Stick to orders. If you need a single weekly number, 90-day repeat on the cohort that is now 90 days old is more honest than a blended lifetime returning share.
How to read a retention cohort curve
Retention-modelling for ecommerce is usually a table, not a neural net. Rows are first-order months. Columns are months since that first order (M0, M1, M2…). Cells are the share of the cohort that ordered in that month, or the cumulative orders per buyer, or contribution per buyer. Start with share who have made a second purchase by month N—it is the easiest to explain in a standup.
What you want: M0 is 100% (they all bought once). A useful slice of them come back in M1 or M2, and a smaller loyal set keeps showing up so the curve flattens above zero instead of touching the axis. What you do not want: a cliff after M0, identical for every channel, that you then fund with more top-of-funnel spend. That curve says LTV ≈ first-order revenue, so CAC must fit inside first-order contribution or you are buying a loss.
Read two overlays. First, contribution per cohort member, not revenue: returns and discounting show up here. Second, acquisition source. Paid-social cohorts often decay faster than organic or referral. If prospecting payback assumes the branded-search repeat rate, the model is fiction. Rebuild LTV from the cohort in front of you, then decide bids.
Durables (furniture, a once-a-year jacket) have long, low curves by design. Do not copy a coffee brand’s 90-day repeat target. For durables, retention might mean accessories, care products, or a two-year replacement cycle—and referrals. Name the actual second transaction or you will beat up a team for missing a replenishment KPI that cannot exist.
| Cohort signal | What it usually means | First move |
|---|---|---|
| Cliff after first order | No reason to come back, or the first order disappointed | Post-purchase experience, replenish timing, product QA |
| Second order then silence | You earned a refill, not a habit | Cadence reminders, subscription option, content that uses the product |
| Channel A retains, channel B does not | Blended LTV is lying | Cap bids on B until creative or offer matches A’s intent |
| Repeat up, contribution down | You retained with discounts | Cut the code; keep the reminder |
Retention versus acquisition budget
The budget question is not philosophical. It is payback arithmetic. Payback in orders ≈ CAC ÷ contribution margin per order. If that number is 1.7 and only 30% of new buyers ever place a second order, most of the cohort never pays back. Spending more on acquisition scales a machine that loses money. Spending on retention (flows, packaging that gets unboxed, a refill that is easy to find) is how you change the 30%.
The other side is also true. If 90-day repeat is already 45%, contribution is healthy, and you are under-spending on prospecting, more email flows will not fix a reach problem. Acquisition then has a unit that can accept it. Rehearse the fork in the retention vs acquisition scenario and in LTV endgame. Related reading: LTV, CAC, and payback.
A practical split for a store that is not yet paying back: freeze scaled prospecting at last month’s efficient volume, and put the next testing dollars into time-to-second-order. When 90-day repeat and contribution LTV move, unlock budget. That is growth and retention as one system, not two departments arguing about creative.
Levers that produce a second order
Email and SMS flows
Post-purchase is the retention channel you already paid for. Delivery updates, how-to, replenish-at-the-right-day, and a cross-sell that matches the first SKU outperform a weekly blast with a new code. SMS belongs on shipping and restock, not on every merchandising whim. Measure the flow on second-order rate for that cohort, not on email revenue (which often double-counts people who would have bought anyway).
Replenishment and subscription
If the product is consumed, remind them when it should be empty. That reminder is retention; the optional subscribe-and-save is monetization riding along. Do not hide cancellation. A trapped subscriber inflates “retention” until chargebacks arrive. See the monetization hub for take-rate and skip-rate hygiene.
Loyalty without giving away the unit
Points that fund 20% off every order are a permanent price cut. Useful loyalty is early access, a meaningful reward at a high threshold, or status that does not sit on contribution. If you cannot show that members have higher contribution LTV after the reward cost, you launched a discount club.
Post-purchase experience
Packaging, fit, arrival time, and support are retention mechanics. They also move refund rate, which is a unit-economics line. A two-day delay with no tracking SMS can erase a brilliant welcome flow. Treat 1-star “never again” themes as a retention backlog, not only a CX ticket queue.
How retention feeds LTV and payback
Contribution LTV ≈ contributing orders per acquired customer × contribution margin per order. Each extra order lifts LTV and shortens payback. LTV and payback period are retention readouts as much as they are finance readouts. If you improve repeat but contribution per order falls (because you discounted to get the repeat), LTV may not rise. Always read retention in contribution, the same way you read acquisition in CAC.
A compact example: $24.80 contribution per order, $42 CAC, 1.3 contributing orders per year-one buyer. LTV ≈ $32.24, LTV − CAC is negative, payback never arrives for the average buyer. Lift repeat so contributing orders go to 1.9 without cheapening the basket: LTV ≈ $47, the hole shrinks, and some channels may even clear a modest scale test. That lift is cheaper than finding a new prospecting channel with a $20 CAC you have never demonstrated.
First five actions (retention)
- Build a 90-day repeat table for last quarter’s new customers, split by acquisition channel. This is the scoreboard; do not skip it for a brand survey.
- Fix post-purchase operations in week one: tracking links, delivery ETA honesty, and a how-to that reduces the top refund reason. Retention starts before the lifecycle email.
- Time one replenish or accessory send to actual consumption or first-use, not to “Tuesday newsletter.” Measure second-order rate for that cohort against a holdout if you can.
- Stop using a standing loyalty percent-off as the retention program. If members only convert with the code, you have a price, not a relationship.
- Tie the next acquisition bid change to the cohort table. If payback still needs a second order you do not get, do not raise CAC this month.
None of those actions require a new app install. Most retention software fails because the underlying cadence and product experience were never specified. Specify the second transaction, fund the operations that make it likely, then let tools automate the reminder. If you cannot name the second transaction in one sentence (refill, size 2, spare part, gift for a friend), you do not have a retention model yet—you have hope.
Review the cohort table monthly in the same meeting that reviews CAC. Splitting those reviews is how teams scale ads on a story that only branded search customers actually live. Growth and retention are one P&L. Keep them in one room. If the meeting cannot show a cohort table, it is not a retention meeting—it is a campaign recap.
For the financial definitions that sit under this stage, use LTV, payback period, and unit economics. This hub is how those numbers show up in a store’s calendar, not a second glossary.
A 30 / 60 / 90 day operating cadence
Retention work fails when it is a campaign, not a cadence. Day 0–2 is operational: delivery promise, tracking, how-to, “did it arrive?” That window decides refunds more than it decides loyalty. Day 7–21 is the first replenish or accessory conversation, timed to when the product is in use—not to when the marketing calendar wants a send. Day 30–90 is the second-order push for people who still have not come back: a useful reminder, a bundle that completes the first purchase, or a genuine restock, not a panic code.
Each window has an owner and a metric. Operations owns on-time and “where is my order” volume. Lifecycle owns second-order rate for the cohort now aging through that window. Merchandising owns whether the second SKU is actually in stock when the reminder fires. If the reminder promotes an out-of-stock refill, you spent your one good attention moment on a 404.
Winback after 90 days is a different motion. Those buyers already told you the default cadence failed. A winback can work for seasonal goods; it rarely repairs a product that disappointed. Split winback campaigns from true retention flows so a 12% discount to lapsed buyers cannot masquerade as “our retention program.”
Worked cohort sketch
One thousand new customers in March, $42 CAC, $24.80 contribution per order. By day 90, 280 have placed a second order (28% repeat). Contributing orders per acquired customer in 90 days ≈ 1.28 if almost nobody has ordered a third time yet. 90-day contribution LTV ≈ $31.74. You are still about $10 short of CAC, with some hope in months 4–12 if the category repeats that way.
Now move repeat to 40% in 90 days (400 second orders) without cutting contribution. Orders per buyer ≈ 1.40; LTV ≈ $34.72. Still not payback, but the hole is smaller, and if a third of those repeaters buy a third time by month six you start to see a real tail. The acquisition team’s bid increase should wait until that tail is visible in the table, not in a brand-tracking anecdote.
If instead you “buy” the 40% repeat with a standing 15% loyalty code, contribution per order might fall to $19. Then 1.40 × $19 = $26.60 LTV—worse than the original 28% repeat at full contribution. That is why retention modelling has to carry margin, not only a second-order count. Discount-funded repeat is often just slower acquisition.
Use the Ecommerce Simulator to stress those numbers, then the customer metrics hub and unit economics glossary when a term in the article grid below needs a definition. Retention work that does not move second-order rate or contribution LTV is a newsletter, not a growth stage. If you only remember one line: retention is a second contributing order, on purpose, on a calendar you can show in a meeting—not a points balance and not a SaaS churn dashboard pasted onto a store.