Cohort Analysis for Subscription Businesses
A virtual CFO's guide to cohort analysis: the three cuts that change decisions, reading a decaying versus flattening curve, and the common ways cohorts get…
Most cohort charts are wallpaper: colourful triangles that look analytical and change no decisions. A useful cohort analysis is narrower and sharper. It answers a specific retention question, and three cuts do almost all the work. This is a virtual CFO’s guide to the cohort analysis that actually informs decisions in a subscription business, whether that is SaaS or a consumer subscription.
Published: July 2026
A cohort is a group of customers who started in the same period, usually the same month, tracked together over time. Instead of looking at all customers blended into one average, you follow each month’s intake as its own group and watch what happens to it: how many stay, how much revenue they generate, how their behaviour evolves month by month after they joined.
The power of the cohort view is that it separates the effect of when a customer joined from the effect of time. A blended retention number mixes new and old customers together and hides whether the business is getting better or worse at keeping people. Cohorts pull those apart, so you can see whether customers acquired this year retain better than those acquired last year, which is a question the blended average simply cannot answer. This is the retention companion to the unit economics for ecommerce and SaaS unit economics work, and a common input to a 90-Day Number engagement.
Cohort analysis can be sliced endlessly, which is how it becomes wallpaper. Three cuts do the work that changes decisions.
The first is logo retention by cohort: the percentage of customers from each starting cohort still active over time. This answers whether you are keeping customers at all, and whether newer cohorts stick better or worse than older ones.
The second is revenue retention by cohort: the revenue each cohort generates over time relative to what it started with. This is more informative than logo retention, because it captures expansion and contraction, not just survival. A cohort can lose customers while growing revenue if the survivors expand, which logo retention alone would miss.
The third is payback by acquisition-channel cohort: grouping customers by the channel that acquired them and tracking how long each channel’s cohort takes to repay its acquisition cost. This answers which channels bring customers worth keeping, as opposed to which bring the most customers, and it is often the cut that reshapes the marketing budget.
The single most important skill in cohort analysis is reading the shape of the retention curve, because the shape forces different decisions. A decaying curve keeps falling: customers leave steadily and the cohort never stabilises, which signals a fundamental retention problem, the product or the fit is not holding people, and no amount of acquisition will build a stable base on top of it. A flattening curve falls at first and then levels off: early churn removes the poor-fit customers, and a loyal core remains and persists. A flattening curve is the sign of a healthy subscription business, because it means each cohort settles into a durable base you can build on.
The decision each forces is different. A decaying curve says stop pouring money into acquisition and fix retention first, because you are filling a leaking bucket. A flattening curve says the base is sound, so acquisition compounds, and the lever is either acquiring more or lifting the plateau. Telling the two apart, and acting accordingly, is most of the practical value of doing cohort analysis at all. The slope between the first and third month, the early onboarding period, is usually the most decisive part of the curve.
Take a subscription business tracking monthly cohorts on revenue retention. The January cohort starts at $100,000 of monthly revenue. By month three it has fallen to $82,000 as poor-fit customers churn. By month six it is $79,000, by month nine $78,000, and by month twelve it holds at about $77,000. The curve fell sharply early and then flattened, settling at roughly 77 per cent of its starting revenue: a durable base.
Now compare that to the acquisition channel behind it. Split the cohort by channel and suppose the paid-search customers flattened at 85 per cent while the customers from a discount-driven campaign decayed past month six toward 55 per cent and kept falling. The blended 77 per cent hid two very different realities. The decision is immediate: the discount channel is buying customers who do not stay, so its true cost is far higher than its headline acquisition cost, and the budget should shift toward the channel whose cohort actually holds. That is the kind of decision cohort analysis exists to produce, and none of it is visible in a blended retention number.
Cohort analysis goes wrong in predictable ways. The first is blending annual and monthly plans in the same cohort: annual customers cannot churn until renewal, so mixing them with monthly customers flatters the early curve and hides monthly churn. Keep the billing terms separate. The second is survivor bias in reading the curve: focusing on the customers who remain and forgetting the ones who left, which makes retention look better than it is. The third is quoting the best cohort as if it were typical: picking the month that happened to retain well and presenting it as the business’s retention, which is a vanity move that misleads the founder as much as anyone. The discipline is to read the cohorts, keep the segments clean, and follow every cohort including the bad ones. This links directly to net revenue retention for operators, which builds on the revenue-retention cut, and to the price of churn, which values what a point of retention is worth.
What is a cohort?
A group of customers who started in the same period, usually the same month, tracked together over time. Instead of blending all customers into one average, you follow each month’s intake as its own group, which separates the effect of when a customer joined from the effect of time and reveals whether the business is getting better or worse at retention.
Which cohort cuts actually matter?
Three. Logo retention by cohort (are you keeping customers). Revenue retention by cohort (including expansion and contraction, which logo retention misses). And payback by acquisition-channel cohort (which channels bring customers worth keeping). These three change decisions; most other cuts are wallpaper.
What is the difference between a decaying and a flattening curve?
A decaying curve keeps falling, signalling a fundamental retention problem no amount of acquisition can fix. A flattening curve falls early, as poor-fit customers churn, then levels off as a loyal core persists, which is the mark of a healthy subscription business. Telling them apart forces different decisions: fix retention first, or build on a sound base.
Why is revenue retention better than logo retention?
Because it captures expansion and contraction, not just survival. A cohort can lose customers while growing revenue if the survivors expand their spending, which logo retention alone would miss entirely. Revenue retention shows the full picture of what a cohort is worth over time, which is what drives value.
What are the common mistakes in cohort analysis?
Blending annual and monthly plans (annual customers cannot churn until renewal, flattering the curve); survivor bias (focusing on who stayed and forgetting who left); and quoting the best cohort as typical (a vanity move that misleads). The discipline is clean segments, honest reading, and following every cohort including the poor ones.
How does this connect to LTV and NRR?
Cohort revenue retention is the raw material for both. Net revenue retention is essentially the revenue-retention cut expressed as a single figure, and lifetime value depends directly on the shape of the retention curve. A flattening curve produces durable LTV; a decaying one does not. Cohorts are where those headline metrics come from.
Can a virtual CFO build this for me?
Yes. A cohort analysis on the three cuts, read for curve shape and channel quality, is a defined deliverable and a natural component of a 90-Day Number for a subscription business. The output is a model you own that shows which cohorts and channels build a durable base, so you can direct acquisition and retention accordingly.
Sydney Virtual CFO is a Sydney-based virtual CFO service for founders running $2M to $15M businesses across SaaS, ecommerce, professional services, construction, and other low-volume, high-value industries. We deliver fixed-scope CFO engagements with a named deliverable on day 90: a 13-week cashflow forecast, a fundraise-ready financial model, a unit economics build, or a board reporting pack you can run on your own.
Our front-door product, the 90-Day Number, is fixed scope at $17,850 plus GST. We are one of the few project-based virtual CFOs in Australia, in a market built almost entirely on monthly retainers. No retainers without a deliverable. No 80-page reports. No theatre.
Visit Sydney Virtual CFO | The 90-Day Number | Book a Call
This content is general information only, written for Australian founders running businesses in the $2M to $15M revenue range. It does not constitute tax, financial product, investment, or legal advice and should not be relied on as such. The work referenced is led by a Chartered Accountant (CA ANZ), but Sydney Virtual CFO is not a licensed tax agent, not a licensed financial adviser, and not authorised to provide personal financial advice. Tax obligations, accounting treatments, fundraise terms, and statutory requirements depend on your individual circumstances. For advice specific to your business, contact the team directly or consult a registered tax agent, licensed financial adviser, or qualified lawyer. Information was current at the time of publication and may change without notice. We review and update guides periodically.
