Cohort analysis for DTC founders: why your cheapest customers cost the most

A founder shares the influencer sheet on a call. One creator's code drove 400 orders last month at a $30 customer acquisition cost, half of what Meta costs. Black Friday was even better: a $22 CAC and a first-order ROAS close to 3x. The plan is obvious. Double the creator program, go bigger on sales, and cut back the "expensive" channels.

Six months later, the cheap customers have barely come back. The expensive ones, acquired at full price through search and Meta, have paid for themselves twice over. The brand did not have a cheap channel and an expensive one. It had channels that looked cheap on day one and channels that were cheap by month four.

That difference is only visible through cohort analysis, and it is the single most common way we see DTC brands misallocate budget. This guide shows how to build cohorts by channel and offer, why payback has to be measured in contribution rather than revenue, and the checks every discount code needs before you renew it.

What cohort analysis is, in plain English

Shopify has a customer cohort analysis report that groups customers by the month they arrived and shows retention over time. It is a good start. What it does not do is split cohorts by acquisition channel, offer or discount, and that is where the budget decisions live.

The classic view is a cohort triangle: one row per acquisition month, one column per month since first order. Read across a row to see how one cohort ages. Read down a column to see whether newer cohorts behave better or worse than older ones at the same age.

Why payback must be measured in contribution, not revenue

Most cohort dashboards show revenue per customer. That is the wrong unit for a payback decision. A customer who spends $79 with you over six months, mostly at 40% off, may leave you only $18 after product cost, shipping and fees. If you paid $22 to acquire them, they never paid back, no matter how good the revenue line looks.

The case: the supplements brand with the "best" channel

A supplements brand was choosing where to put next quarter's budget. Contribution margin is 50% at full price. Black Friday and influencer codes had the best first-order ROAS by a distance, and the plan was to double both. Here is what the cohorts showed. The figures are an illustrative composite.

Black Friday even had the best six-month revenue ROAS of the four, at 3.59x. But the discount took most of the margin and few customers came back, so the cohort never earned back its cost. Google Search looked worst on day one, with a first-order ROAS of 1.07x, and paid back first. On closer inspection, a third of the influencer code's orders came from existing customers, which made its real cost per new customer far higher than $30.

This is not just a pattern from our own work. A Journal of Marketing Research study of an online grocer and a newspaper found that deeper acquisition discounts led to lower repeat buying in both businesses. In one dataset, a 35% acquisition discount produced customers worth about half as much over the long run as customers acquired at full price.

Three checks before you renew any discount code

Influencer and promo codes are the channel most often credited with sales they did not create. Before renewing a creator deal or repeating a sale, run these three checks.

1. How many orders came from existing customers? People who already buy from you will happily use a code they see. That is a discount on a sale you already had, not a new customer.
2. Did the code leak? Codes get picked up by coupon sites. UGCBloom describes an aggregator driving 200 orders in a week on a code that normally sees 20, which looks exactly like a viral post. HelloPartner points out that shoppers now ask AI assistants for a brand's discount code and get one. The creator gets credit for sales they never touched.
3. Did those customers come back? Code customers were selected by a discount. Many wait for the next one. Compare their 90-day repeat with your full-price cohorts.

Holiday cohorts need their own line

November and December cohorts behave differently from the rest of the year. More gift buyers, deeper discounts, more one-time shoppers. If you average them into your LTV, two things go wrong: your LTV estimate drops and makes good channels look worse, and your first-quarter reporting looks like a retention collapse when it is just the holiday cohort ageing.

Report Black Friday and holiday cohorts on their own, and check them again in March. If they reorder at full price, great: those sales were worth it. If they only come back for the next sale, plan next year's promotion with that cost in mind.

How to build payback cohorts by channel and offer

1. Export 12 months of orders with customer ID, date, discount code, UTM source and refunds.
2. Tag each customer by the channel and offer of their first order. Treat each significant code as its own group.
3. Build cohorts by acquisition month and channel, with at least 100 customers each.
4. Calculate cumulative contribution per customer for each month since the first order.
5. Calculate CAC by channel using Shopify new-customer counts, not platform-reported conversions, and find each cohort's payback month.
6. For every code, check the share of orders from existing customers and any sudden spikes that look like leakage.

The KPIs that matter

The decision

Cohorts become even more useful when you add a second cut: what customers bought first, and whether they used a welcome code. We cover both in customer value by first product and welcome discounts and bargain hunters. Building these cohorts across every channel, code and product is a core part of a Blindspot Snapshot, and usually the part that changes the budget the most.