Net Retention Audit (Cohort NRR)
When you need this method
Someone hands you a net revenue retention of 118 percent, in a portfolio review, a data room, or an investor deck, and the number carries a valuation. How it was produced is rarely stated. Calculation methods diverge widely in practice, and the common shortcut from the ARR bridge tends to run high at growing vendors. Skip the derivation and you are arguing with a number that does not measure what it claims to measure.
Approach
- 1Fix the set of customers as of the reporting date twelve months ago and record the recurring revenue they carried then.
- 2Value that same set of customers today and divide today's revenue by the earlier figure. That is the defensible net retention.
- 3Put the reported number beside it and ask how it was derived. If it comes from the ARR bridge, meaning opening base plus net expansion divided by opening base, it measures something else.
- 4Quantify the gap and name its source: an ARR bridge cannot separate expansion within the old cohort from expansion at customers won during the year.
- 5Report gross revenue retention and logo retention alongside it, so that expansion at a few large accounts does not mask shrinkage in the base.
- 6Build the cohort layer cake, stacking annual cohorts on top of each other, and look at how each individual layer develops.
- 7Use only the recomputed figures in valuation, price arguments, or diagnosis, and record the cohort definition, the reporting date, and the segment boundaries for every number.
Typical application
A typical B2B SaaS in the HR space reports net revenue retention of 118 percent in its materials. Recomputed against the customer group as of twelve months ago, 104 percent remains. The gap comes from additional licences at customers won during the current year, who therefore never belonged to the cohort under review. The same pass shows gross revenue retention at 91 percent: the installed base is shrinking, and expansion at a few large accounts had covered it up. Conversations then run on three numbers instead of one, and the argument over valuation shifts from the level of retention to how durable the expansion is.
Limits and counter-indications
The check requires recurring revenue at customer level across twelve months. Where only an ARR bridge exists and customer data is out of reach, for instance early in a review, it cannot be run at all. With usage-based billing, monthly cancellable contracts, and volatile consumption, even the cohort definition is contested, and two clean calculations can land on different answers. Acquisitions, currency effects, and renamed product lines shift the mapping and have to be adjusted for one by one. With a handful of very large accounts, even the correctly computed figure is unstable, because a single contract moves it by percentage points. And a cleanly computed net retention says nothing about whether expansion came from additional usage or from enforced price increases.
How to measure impact
Track cohort NRR, gross revenue retention, and logo retention side by side, and record the reporting date, the cohort definition, and the segment boundaries for each figure.
Related methods
Sources
- 1.Lazy NRR is Not NRR. Accept No Imitations or Subtitutes., Kellblog (Dave Kellogg) (opens in a new tab) · Kellblog · 2022-10-05 · practitioner source · describes the methodCarries the procedure and the distinction: net revenue retention is computed from snapshots and cohorts, that is, today's revenue from the customer group of twelve months ago divided by what that group carried then. The shortcut of opening base plus net expansion divided by opening base measures something else, the source argues, because an ARR bridge cannot separate expansion within the old cohort from expansion at customers won during the period. It additionally recommends the cohort layer cake, since it leaves little room for gaming. Limit: a blog post by an experienced practitioner, with no empirical measurement of how large the deviation typically is.
- 2.Essential SaaS Metrics: Revenue Retention Fundamentals, SaaS Capital (opens in a new tab) · SaaS Capital · 2015 (aktualisiert 2019) · practitioner source · describes the methodIndependently of Kellogg, it describes the same computation: a fixed set of customers, their recurring revenue twelve months ago against today's revenue from that same set. For gross retention, each customer's current amount is capped at their figure from twelve months earlier, so that expansion cannot mask churn. The source also notes that two vendors with identical net retention can be in very different shape, depending on whether they keep customers or offset losses through cross-selling and price increases. Limit: a lender's publication, with no disclosed data basis for this particular point.
- 3.How Public Companies Calculate Net Revenue Retention, Ordway Labs (opens in a new tab) · Ordway Labs · 2024 · investment, consulting and analyst firms, industry bodies and public agencies · provides benchmark figuresEstablishes that calculation methods really do diverge in practice. The edition retrieved covers 135 software, cloud, and fintech companies listed on US markets: roughly 47 percent compute on ARR or MRR, roughly 42 percent on reported revenue, with further differences in comparison periods, trailing twelve month averaging, and the exclusion of individual customer segments and product lines. The source cites a 2019 KeyBanc survey that found 110 different approaches. Limit: it documents the spread of definitions, not the direction or size of the distortion in any single figure, and it comes from a billing software vendor.
- 4.Measuring SaaS Renewal Rates: Way More Than Meets the Eye, Kellblog (Dave Kellogg) (opens in a new tab) · Kellblog · 2013-10-04 · practitioner source · provides the contextWorks through seven different renewal rates from a single example data set, with results far apart depending on whether the base is revenue, seats, customer count, or bookings, and on whether expansion is included. From this follows the recommendation to run several figures side by side rather than relying on one, because expansion can arithmetically hide shrinkage in the base. Limit: a constructed worked example rather than a survey, and the terms gross and net are used there differently from today's common NRR vocabulary.
Origin: Kellogg · Adapted from: Dave Kellogg (Kellblog)