SaaS retention: rebuild the original customer cohort
Net revenue retention measures revenue change within a defined existing-customer population. It can exceed 100% while customers leave. Rebuild the cohort before using a retention headline to argue that customer relationships are durable, and separate revenue retention from the proportion of accounts that remain active.
Inspect the company’s population rule
Snowflake explains its net revenue retention calculation using a specified customer cohort and two annual revenue periods, keeping customers with no second-year use in the cohort at zero. That definition concerns product revenue under its business model. An annual recurring revenue retention metric at a subscription company may use a different basis.
Copy the actual cohort entry rule, time windows, reseller treatment and acquisition policy. Check whether the metric counts usage, contracted recurring value or invoiced revenue. Comparing a consumption metric with a contracted subscription metric can confuse reduced activity with canceled contracts. The same acronym does not guarantee the same economic question.
Hypothetical fixed-cohort bridge
Assume one original cohort contains 100 accounts producing 10 million in annual revenue. In the next year, continuing customers add 2.8 through expansion, reduce spending by 0.7 through contraction, and ten departing accounts remove 0.6. Revenue from that original cohort becomes 11.5, giving net retention of 115%.
| Reusable cohort worksheet | Revenue movement | Treatment |
|---|---|---|
| Opening cohort revenue | 10.0 | Freeze original population |
| Expansion by original customers | +2.8 | Separate price and usage if available |
| Contraction by original customers | −0.7 | Retain reduced-spend accounts |
| Revenue lost from departing accounts | −0.6 | Keep zero-revenue rows |
| Closing original-cohort revenue | 11.5 | Numerator for net retention |
| Revenue from new customers | 1.4 | Exclude from this ratio |
An illustrative gross revenue retention measure would be (10 − 0.7 − 0.6) / 10 = 87%, excluding expansion. Account retention would be 90 / 100 = 90%. These are three different observations. Label the gross-retention formula as an analytical convention and check whether the issuer publishes an equivalent definition.
Ask where the expansion came from
Suppose two large customers provide 2.4 of the 2.8 expansion. The 115% headline then depends heavily on those accounts, although the disclosed data may not reveal their names. If a price rise supplies another 0.3, volume expansion is smaller than revenue expansion. Avoid interpreting every retained dollar as broader product adoption.
Add columns for price, seats or usage, cross-selling, acquisition and currency. Use a residual when the public data cannot split those effects; do not manufacture customer-level precision. Tracking the same published metric over time is still useful, provided its cohort rule and measurement basis remain comparable.
Test cohort survival honestly
Never delete departed customers from the starting denominator. Doing so changes the question from original-customer retention to spending by survivors. Also avoid adding new customers who began during the second year. Either error can inflate apparent retention without improving existing relationships.
For the next review, ask whether expansion remains broad or depends on fewer large accounts, and whether gross losses are growing beneath positive net retention. If the issuer stops disclosing the necessary components, record the observation limit. This worksheet tests a customer-economics mechanism; it does not assign a valuation multiple or imply that any retention level ensures returns.
FAQ
Does 115% retention mean all customers stayed?
No. Expansion from remaining customers can more than offset lost revenue. Check account counts and gross losses separately where available.
Can a rolling cohort be compared with one fixed launch cohort?
Only with clear labels. Their membership and customer ages differ, so matching percentages can describe different underlying customer behavior.