Cohort Analysis for E-Commerce: What Your Aggregate Metrics Are Hiding

Your ROAS is stable. Your CAC looks fine. Your revenue is growing. And yet, quietly, your business is deteriorating — you just can't see it yet.

Aggregate metrics are seductive because they're simple. A single dashboard number feels like clarity. But averages mask the thing that matters most in e-commerce: whether the customers you're acquiring today are worth more or less than the customers you acquired six months ago. Cohort analysis is how you find out — before the damage shows up in your P&L.

What Cohort Analysis Actually Tells You

A cohort is simply a group of customers who share a common characteristic — usually their acquisition date. Cohort analysis tracks how those groups behave over time, separately, instead of blending everyone together.

The typical e-commerce dashboard shows you total revenue, total orders, average order value. Those numbers don't tell you whether your January customers are still buying in June, or whether they bought once and disappeared. Cohort analysis does. Specifically, it reveals:

• Retention by acquisition period — Are newer cohorts sticking around as long as older ones?
• Revenue per cohort over time — Which months generated customers with the highest long-term value?
• CAC payback period — How long until a cohort's cumulative revenue covers what you spent to acquire them?
• Channel quality comparison — Do customers acquired through affiliate content retain better than those from paid social?

The Silent Decay Problem

Here's a pattern that shows up constantly in e-commerce brands experiencing flat-to-growing top-line revenue but shrinking margins: cohort quality is declining, but it's being masked by acquisition volume.

If you acquired 10,000 customers in Q1 at a 60% D30 retention rate, and 15,000 customers in Q3 at a 40% D30 retention rate, your aggregate retention metric looks fine — maybe even improved — because you have more customers. But your Q3 cohort is burning marketing budget and churning fast.

The brands that catch cohort decay early have one thing in common: they build retention curves into their weekly reporting, not their quarterly reviews.

By the time declining cohort quality shows up in monthly revenue trend lines, you've usually spent two or three more quarters acquiring the wrong customers at the wrong price.

How to Build a Cohort Report That's Actually Useful

A functional cohort analysis doesn't require a data science team. Most e-commerce platforms and BI tools — Looker, Metabase, even a well-structured spreadsheet — can produce one with the right query structure. The core steps:

• Define your cohort window: Week or month of first purchase is the most actionable for e-commerce — granular enough to catch shifts quickly, broad enough to produce statistically meaningful groups.
• Pick your metric: Start with repeat purchase rate at 30, 60, and 90 days, then layer in cumulative revenue per cohort to understand compounding value.
• Segment by acquisition channel: This is where the real insight lives. Don't look at all customers blended together — break out paid search, affiliate content, paid social, and email acquisition separately.
• Set a CAC payback benchmark: If your target is 90-day payback, flag any cohort that's tracking below the expected curve at day 60 — while you still have time to adjust spend.

The goal isn't to produce a beautiful heatmap. It's to surface a specific, actionable question: which channels are producing cohorts that compound over time, and which are producing one-and-done buyers you're over-investing in?

Connecting Cohort Data to Channel Allocation

Cohort analysis only creates value when it changes how you spend. The most direct application: use it to stress-test your channel mix.

If affiliate-driven content customers have a 90-day repeat rate of 45% and paid social customers repeat at 22%, that's not just a retention stat — it's a signal that your current ROAS comparison dramatically undervalues the affiliate channel. A customer who buys twice is worth far more than a customer who converts once, but last-click attribution credits both the same.

Similarly, if a specific affiliate network consistently produces higher-LTV cohorts — say, AWIN content partners outperforming cashback publishers on repeat purchase rate — that's where you increase commission rates and deepen publisher recruitment, not where you cut costs to chase short-term ROAS efficiency.

Real revenue growth isn't about acquiring more customers. It's about acquiring customers who stay, buy again, and cost less to retain than to replace. Cohort analysis is the diagnostic that tells you whether you're building that kind of business — or just papering over a slow leak with more acquisition spend.