Recalculating reported net revenue retention on a consistent cohort routinely moves the figure. Cohort definition and the treatment of contracted-but-unbilled expansion explain more of the gap than churn does. A reported NRR can be entirely true and still describe a business that is not compounding the way the number suggests.
What is net revenue retention?
Revenue from an existing customer cohort at period end divided by revenue from the same cohort at period start, including expansion, contraction and churn. New customers are excluded, which is the entire point of the metric.
It is a good metric. The difficulty is that "the same cohort" is defined by the company, and that definition is where the variance lives.
The metric also assumes the cohort is a meaningful unit, which it is only where customers were acquired on comparable terms. A cohort blending enterprise contracts signed after a competitive process with self-serve subscriptions taken out in an afternoon is averaging two behaviors that have nothing to do with each other, and the resulting figure describes neither.
What can inflate a reported NRR?
Cohort choices, principally. Which customers enter the base, how customers who signed mid-period are handled, and whether monthly and annual contracts are pooled can each move the figure by several points.
The treatment of contracted-but-unbilled expansion is the second. Recognizing an upsell at signature rather than at billing pulls expansion forward, which flatters the current period at the expense of the next. Whether that expansion then sticks is a switching question rather than an accounting one.
Contraction handling is a third lever and the least discussed. Whether a customer who reduces seats but stays counts as contraction or as partial churn changes the figure, and both treatments are defensible. It matters most in exactly the businesses where the number is being used to argue durability, because contraction is what durability looks like when it is failing slowly.
How does cohort choice move the number?
By deciding the denominator. Excluding customers who signed mid-period, or rolling contract types together, changes the base against which growth is measured without anyone doing anything improper.
Because conventions differ between companies, a benchmark is only comparable if the cohort definitions match — and published NRR benchmarks almost never state them, which is why a diligence rebuilds the cohort rather than benchmarking it.
The denominator effect also compounds across periods. A convention that flatters one year sets a base that makes the next year harder, which produces a series that looks strong and then inexplicably softens. Where a reported figure drops without an obvious operational cause, the cohort convention is the first place to look rather than the last.
The number was right. What we had not asked was which customers were in the base, and when we rebuilt it the story about the last two years changed shape.
How do you verify NRR with customers?
Reconstruct the cohort from billing data yourself, then test the expansion story on calls with customers who actually expanded. Customers can tell you whether an upsell was a genuine expansion or a re-papered contract.
That second conversation is the one that matters. Contract restructuring frequently appears as expansion in the data and as nothing at all to the customer.
Verification also has a natural stopping point. Once you have rebuilt one period's cohort and reconciled it, you know the convention, and the remaining series can usually be taken as consistent with it. The work is front-loaded rather than proportional to the number of periods, which is what makes it fit a diligence timetable.
What NRR actually tells an investor
Whether the existing base grows without new logos, which is a real and important thing to know. What it does not tell you is whether that growth is repeatable, and cohort mechanics can make a one-off look structural — which is exactly the distinction an investment committee is being asked to make.
Read alongside gross retention and the cohort definition, it is informative. Read alone, it is a headline.