In this guide
A cohort is a group you follow over time. For a subscription business, a useful starting point is customers whose paid recurring service begins in the same month. Tracking those groups separately reveals patterns that a blended revenue total can conceal.
The core idea is to compare customers at a similar age in their relationship with you, rather than comparing a mature group with a group that arrived last week. ChartMogul's cohort analysis documentation explains this elapsed-period structure. The worked examples below are original, fictional illustrations.
What does a SaaS cohort table show?
Each row represents a customer group. Each column represents time since that group's starting period. M0 is its starting month, M1 the next calendar month, and so on. A cell shows a chosen measure, such as customers retained or recurring revenue relative to the starting value.
The following revenue retention example is observed through April. January has reached M3; April has reached only M0.
| Starting cohort | M0 | M1 | M2 | M3 |
|---|---|---|---|---|
| January | 100.00% | 95.00% | 96.00% | 97.50% |
| February | 100.00% | 102.50% | 105.00% | — |
| March | 100.00% | 80.00% | — | — |
| April | 100.00% | — | — | — |
The blank cells are future ages, not zero retention. February's M2 should be compared with January's M2, not January's M3. March's 80.00% at M1 is a prompt to investigate, not proof that the product or acquisition channel caused a problem.
Calendar-month cohorts also contain different exposure lengths within M0: a customer starting on the first day has more initial service time than one starting on the last. Keep that in mind when comparing revenue contributions or payback in the starting month.
Customer retention and revenue retention answer different questions
Customer retention at an age = active customers from the original cohort at that age ÷ original cohort customers × 100.
Revenue retention at an age = that cohort's recurring value at that age ÷ its starting recurring value × 100.
Suppose January starts with 20 customers and £2,000.00 of MRR. At M3, 15 customers remain active and together contribute £1,950.00 of MRR. Customer retention is 75.00%, while revenue retention is 97.50%.
The remaining customers are paying more on average, but one quarter of the original relationships are no longer active. Neither result replaces the other.
Revenue retention can exceed 100% when expansion outweighs losses. ChartMogul's net MRR retention cohort definition describes that effect. A customer retention snapshot can also rise after a previous decline if original customers reactivate; a “never churned” survival measure would treat that history differently.
Define cohort membership before interpreting the colours
Write down the cohort entry rule. Common choices include first paid subscription, first recurring service month or first recorded invoice. These do not always identify the same date.
If your export starts in January and includes long-standing customers, January is their first observed month in that file, not necessarily their acquisition month. Labelling all of them newly acquired would distort retention, CAC and LTV.
Use a stable customer identifier, retain the original cohort after reactivation and define whether customer identity is scoped by source system or entity. When combining exports, check whether two IDs represent one customer or one ID has been reused. Do not merge businesses or currencies simply because their dates match.
Compare equal ages and weight the results correctly
A younger cohort has had less time to earn revenue or lose customers. Compare its M3 result with other cohorts' M3 results, and only include groups that have reached that age.
For aggregated customer retention, weight by the original customer counts. Consider two fully observed cohorts at the same age:
| Cohort | Original customers | Active at selected age | Retention |
|---|---|---|---|
| Small group | 2 | 1 | 50.00% |
| Larger group | 100 | 90 | 90.00% |
| Combined | 102 | 91 | 89.22% |
The unweighted mean of 50.00% and 90.00% is 70.00%, which misrepresents the combined customer population. Revenue retention needs its own starting-revenue weights, not the customer-count weights used here.
Investigate a pattern without claiming causation
If a cohort deteriorates earlier than its predecessors, begin with a specific question: did the export omit renewals, did a discounted annual plan expire, did the acquisition mix change, or did onboarding change?
Customer age, price, product and acquisition source can move together. A weak row does not prove that a particular campaign caused churn. Check sample sizes and comparable terms before proposing a change.
For a practical review, record the affected age, customers or recurring value involved, source coverage and the next evidence needed. A small group of two customers can change by 50 percentage points when one customer leaves. That is an event to understand, not a stable population estimate.
Connect retention to CAC payback and LTV
Retention shows who stays and how their recurring value develops. Acquisition economics adds what it cost to win the group and how much gross profit it has contributed.
An acquisition-cost upload can provide cohort month, original customer count and allocated acquisition spend without requiring a new customer contact list. Matching recurring service history and direct costs are still needed to calculate gross-profit recovery.
Use the CAC payback guide to follow recovery of acquisition spend, and the LTV guide to distinguish observed value to date from a projected lifetime. Do not compare a six-month observed value for one cohort with a two-month value for another and call the difference performance.
Cohorts in Profitcura Revenue Intelligence
Profitcura Revenue Intelligence separates invoice-based cohorts from service-period recurring cohorts. The first observed positive activity in selected history determines current cohort grouping; that date is not independent proof of acquisition.
Revenue and active-customer retention can be explored, with member detail behind the cells and future ages left blank. Reactivated customers remain in their original group. Acquisition economics requires the client to confirm that the supplied new-customer counts and recurring history match the acquisition cohort.
This makes the data boundary part of the review. A billing-only cohort is useful for understanding invoiced activity, but it should not be presented as verified subscription retention.
A monthly cohort review checklist
- Confirm the cohort rule, selected history, currency and completed observation period.
- Separate true acquisition groups from first-observed groups in a partial export.
- Compare the same elapsed ages and keep future cells blank.
- Review both customer and recurring revenue retention.
- Inspect small samples and missing renewal periods before explaining a change.
- Add acquisition spend and cost coverage for economic questions.
- Give each proposed intervention an owner, a review date and a measurable outcome.
Common cohort questions
Why can revenue retention exceed 100%?
Original customers can expand or return, producing more recurring value than the cohort had initially. New customers from other cohorts should not be added to that row.
Does a blank cell mean customers have churned?
No. It commonly means that age has not been observed yet. A verified zero and an unobserved value need different displays.
Can I use cohorts with a small customer base?
Yes, as a structured way to review individual patterns. Show the counts and evidence, and avoid treating a volatile small sample as a reliable market benchmark.
Explore Profitcura's cohort and revenue workspace or request a walkthrough. The product is in active development; pilot access and terms are agreed separately.