Cohort Retention: How to read retention curves and cohort tables

Cohort retention tracks how a defined group of users or customers behaves after a common starting point. Each cohort is followed through successive time intervals to show how many members continue performing the behavior that represents ongoing product value.

Cohorts can be based on acquisition timing or on behavior. For venture investors, cohort analysis is especially useful in consumer, SaaS, marketplace, fintech, and other products where repeated behavior is central to the business model. It helps separate durable engagement from top-line growth driven mainly by continuous acquisition.

How cohort retention is calculated

Retention at period N = Users from the original cohort active at period N ÷ Users in the original cohort × 100

The definition of active must match the product. A retained user might be someone who logs in, completes a transaction, creates content, makes a payment, or performs another action that represents recurring product value.

Retention can be defined as returning exactly in a given period or returning on or after that period. Those approaches can produce different curves, so the methodology should remain consistent when cohorts are compared.

Worked example

Assume 1,000 users sign up for a product in January. This illustration uses exact-period retention: the user must perform the specified return event during each stated month. Of those users, 600 perform the chosen return action in month one, 420 in month two, and 350 in month three.

Month 1 retention = 600 ÷ 1,000 = 60%

Month 2 retention = 420 ÷ 1,000 = 42%

Month 3 retention = 350 ÷ 1,000 = 35%

A February cohort can then be compared at the same lifecycle points. If February month-three retention is 45%, the company has evidence that newer users are retaining better than the January cohort, even if the overall user base is growing at the same rate.

Illustrative cohort table

Cohort (starting users) | Month 1 | Month 2 | Month 3

January (1,000 users) | 60% | 42% | 35%

February (800 users) | 65% | 50% | 45%

March (900 users) | 68% | 52% | Not yet observed

Each figure is the number returning in that exact month divided by the cohort's starting size. The January curve passes through 60%, 42% and 35%. Compare February's 45% month-three result with January's 35% at the same age. March's month-three result is unavailable because that cohort has not matured; it should not be shown as zero.

How to read a retention curve

The early part of the curve shows how quickly users discover enough value to remain active. A sharp initial decline can signal weak activation, poor customer fit, or a product whose expected usage frequency is lower than the selected interval.

The shape later in the curve is often more important than the first few periods. A curve that begins to flatten suggests that a core group of users continues returning. A curve that keeps falling toward zero suggests the business may need constant acquisition to replace users who eventually leave.

Comparing curves across cohorts helps show whether the product is improving. Better onboarding, stronger product-market fit, changes in customer mix, pricing, or seasonality can all appear as differences between cohort curves.

How venture investors use cohort retention

Headline growth can look strong while underlying cohorts deteriorate. Cohort analysis lets investors ask whether each generation of users is becoming more or less durable and whether growth is being supported by improving product behavior.

Investors also use behavioral cohorts to understand what correlates with retention. Users who complete a key workflow, invite collaborators, fund an account, or reach a particular usage threshold may retain differently from the average user. Those relationships can help identify the product's true activation behavior.

For subscription businesses, user or logo cohort analysis complements NRR and GRR, which measure how recurring revenue from an existing customer base changes over time.

Ask whether newer cohorts improve at the same age, using the same return event and comparable customer mix. Keep the cohort definitions and dated source export in Treto's diligence workspace for the investor's review.

What does a good retention curve look like?

There is no universal retention percentage because the natural usage frequency differs widely across products. A strong curve usually shows a meaningful group of users remaining active after the initial drop-off, followed by a clear flattening rather than a continuing decline toward zero.

Investors also look for newer cohorts to retain at least as well as older ones at the same lifecycle point. Stable or improving cohorts suggest that product quality, onboarding, customer selection, or product-market fit is improving. A curve that keeps deteriorating across successive cohorts deserves closer investigation even when top-line user growth remains strong.

For high-frequency consumer products, investors may care about daily or weekly retention. Enterprise software may be better assessed monthly or through account and revenue retention. The interval should match the product's expected usage cadence.

Common mistakes

Using the wrong definition of active

Logging in is not always meaningful retention. The return event should represent the recurring behavior that creates value in the product or business model.

Comparing incomplete cohorts

Recent cohorts have not had enough time to reach later retention intervals. Comparing their incomplete cells with mature cohorts can create misleading apparent improvements.

Choosing intervals that do not match product frequency

Daily retention is useful for high-frequency consumer products and much less useful for software that customers are expected to use monthly. The observation interval should match the natural cadence of the product.

Relying only on aggregate retention

Company-wide averages can hide meaningful differences across acquisition channels, plans, geographies, customer types, or product behaviors. Cohort and segment analysis should be used together.

Limitations

Retention analysis describes behavior without automatically explaining causality. A cohort can retain better because of product improvements, customer mix, pricing, seasonality, channel quality, or other changes occurring at the same time.

It also requires enough users and enough elapsed time to create a stable pattern. Early-stage companies with small cohorts can show large percentage swings from only a few users. Venture investors should examine the absolute cohort sizes, product usage definition, and maturity of each cohort before drawing strong conclusions from the curve.