Cohort Analysis for DTC Brands How To Read the Retention Curve That Decides Your LTV

Blended LTV divides total revenue by total customers. That single operation averages a customer you acquired in early 2024, who has now bought eleven times, against a customer you acquired last month, who has bought once. The 2024 customer carries the number. The recent customer disappears into it. You can lose 40% of your retention quality over a year and watch the blended figure barely move.

Cohort analysis removes the averaging. You group customers by the month they first bought, then track each group forward on its own clock. The January cohort at month 6 sits next to the June cohort at month 6. Same age, different intake. Now you can see whether the brand is getting better at keeping people or quietly getting worse.

What a cohort actually is

A cohort is a group of customers who share a starting event. In DTC the default is the month of first purchase. Every customer whose first order landed in January 2026 belongs to the January cohort for the rest of their life with your brand. You then measure what percentage of that group came back in month 1, month 2, month 3, and so on, and how much cumulative margin each remaining customer has generated.

The output is a triangle. Rows are cohorts, columns are months since first purchase. Older cohorts have more columns filled in. Reading down a column tells you whether newer intakes are behaving better or worse than older ones at the same age. Reading across a row tells you the shape of that cohort's decay.

Most operators glance at the triangle, note that the numbers get smaller as you go right, and close the tab. The value is in the two comparisons, not in the absolute numbers.

The shape of the curve matters more than the endpoint

Two brands can both land at 22% retention at month 12 and have completely different businesses. One drops to 30% by month 3 and then flattens, holding almost everyone it has left. The other holds 55% at month 3 and then bleeds steadily every month after. The first brand has an acquisition and onboarding problem it can fix with a better first 90 days. The second has a product or lifecycle problem that gets more expensive the longer it runs.

The month 0 to month 3 section is where the leverage sits. Most DTC churn happens inside the first 90 days. Once a customer has bought twice, typical retention from that point runs 85 to 90%. That single fact should reorganise your priorities: the second purchase is not a retention metric, it is an acquisition metric. Until someone buys twice, you have not really acquired them.

The four cohort cuts worth building

First-purchase month is the default cut, and it is the least interesting one. It tells you the trend. The other three tell you what to do about it.

1. Cohort by acquisition channel

This is the cut that changes budget decisions. Group customers by where their first order came from, then compare 12-month contribution margin per customer against the CAC you paid for that channel. The pattern I see repeatedly: the channel producing the cheapest first order produces the worst repeat behaviour. Discount-led paid social brings in price-sensitive buyers who never return at full price. Organic search and referral bring in fewer customers who are worth two to three times more over a year. If you are scaling on first-order CAC alone, you are systematically buying your worst customers.

2. Cohort by first product purchased

Not every SKU makes an equally good front door. Some products create a habit. Others get used once and forgotten. Group your cohorts by the first item in the first order and you will usually find a two to threefold spread in 12-month value between your best entry product and your worst. That tells you exactly which SKU should carry your acquisition spend, which should be a bundle component rather than a hero, and which should never be the thing a cold audience sees first.

3. Cohort by discount on first order

Split customers into those who used a code on their first order and those who paid full price. Then compare retention curves. Most brands find the discounted cohort retains materially worse, which means the welcome offer is not just costing you the discount, it is buying a lower-quality customer base. This does not mean kill the offer. It means price it against cohort value, not against first-order conversion rate.

4. Cohort by first-order AOV band

Bucket first orders into ranges and track each band forward. In most consumable brands, customers who bought two units on the first order retain dramatically better than those who bought one, because they got far enough into the product to feel the result before the reorder decision arrived. That finding turns a bundle from an AOV tactic into a retention tactic, and it justifies spending margin on getting the first order bigger.

How to build it without buying another tool

Shopify has a Customer Cohort Analysis report in the admin. It groups customers by first purchase month and shows retention and cumulative spend over time. Start there. It will give you the trend line in an afternoon and it costs nothing.

Its limits are real. It will not break cohorts down by acquisition channel, it will not cohort by first product or discount status, and it reports revenue rather than contribution margin. For the cuts that actually change decisions, export orders and customers to a sheet, join first-order attribution from your post-purchase survey or ad platform, subtract COGS, shipping, payment fees, and returns per order, and rebuild the triangle on margin. It is a day of work the first time and an hour a month after that.

Calculate on contribution margin, not revenue. Revenue LTV routinely overstates the real figure by two to three times once COGS, shipping, returns, and support cost come out. A 200 GBP revenue LTV customer can easily be a 60 GBP contribution customer, and you cannot set a CAC ceiling against a number that is three times too big.

The benchmarks to hold yourself against

Compare against your category, not the overall average. The average DTC repeat purchase rate sits at 25 to 30%. Underneath that, apparel typically runs 10 to 18% with strong brands at 18 to 25%, beauty typically 15 to 25% with strong brands reaching 25 to 35%, and supplements typically 18 to 30% with strong brands hitting 30 to 40%. Consumables at the very top end reach 40 to 55%.

On the value side, the median DTC subscription LTV to CAC ratio reached 4.1 to 1 in 2026, with replenishment categories matching SaaS-level ratios for the first time. If your cohorts are producing below 3 to 1 on contribution margin, the problem is rarely the ad account. It is the shape of the curve in the first 90 days.

One more number worth acting on: for subscription brands, 12-month retention runs around 28% on annual billing against 11% on monthly. If you sell a consumable and you have never tested a prepay option, that gap is sitting on the table.

What this looks like in practice

A wellness brand I worked with was holding a blended LTV that had barely moved in eight months. On that basis they kept scaling paid social, because the CAC still cleared the blended number. When we rebuilt the cohorts on contribution margin and split them by acquisition channel, the picture inverted. Their paid social cohorts were retaining at roughly half the rate of their organic and referral cohorts, and the gap had been widening every month since they turned on an aggressive first-order discount.

Two changes followed. They cut the welcome discount and replaced it with a two-unit starter bundle at full unit price, because the AOV-band cohorts showed two-unit buyers retaining far better. They also moved spend toward the channels whose cohorts were producing real 12-month margin rather than the cheapest first order.

First-order conversion rate dipped. New customer count dipped with it. Ninety days later the month 3 retention point on new cohorts had moved up by a third, and contribution profit was up despite acquiring fewer people. The average had been telling them to do the opposite for the best part of a year.

Inside the system

How we build this for brands

We run profit and cash-flow dashboards built from live Shopify and ad platform data, with a reporting agent that rebuilds the cohort triangle on contribution margin every week and surfaces the specific cohort that has started to decay before it shows up in the blended number. The same model carries acquisition channel, first product, and discount status on every customer, so the four cuts in this article are one view rather than four exports.

What the cohorts find then feeds the lifecycle work. When the curve says the second purchase is the bottleneck, the fix is a replenishment or post-purchase flow built and deployed in Klaviyo, timed to the actual consumption window rather than a guessed delay. When the curve says the entry product is wrong, the fix runs through the VOC engine into new positioning and ad creative. Part of this runs live for portfolio brands today; the full system is what we deploy when we take a brand on.

Frequently asked questions