GRR vs NRR: Which Retention Metric to Report

Gross revenue retention measures what you lost. Net revenue retention measures what you lost net of what you grew. Report both — and if a board will only look at one number, give them GRR, because it is the one that cannot be flattered.

How they differ in construction

Both fix a cohort at the start of a period and follow only those customers. GRR subtracts churn and contraction and stops there, so it caps at 100% by construction. NRR then adds expansion, so it can exceed 100%.

That single difference is why they answer different questions. GRR asks how leaky the base is. NRR asks whether the base grows without new customers.

Why the pair matters more than either

Business B is the common case and the dangerous one, because the headline number looks identical to Business A's. It is also the case most likely to deteriorate suddenly: expansion is discretionary spend for the customer, and it contracts first when budgets tighten, at which point the underlying churn becomes visible all at once.

What each metric conceals

• NRR conceals churn when expansion is concentrated. A handful of large accounts growing can offset many small ones leaving.

• GRR conceals whether you have any expansion motion at all. A business at 98% GRR and 99% NRR is retaining well and expanding not at all, which caps growth at the rate of acquisition.

• Both conceal customer count. Revenue-weighted metrics say nothing about how many logos left, which matters if the departing segment is where you plan to scale.

Method: computing both from one cohort

1. Fix the cohort at period start and record its recurring revenue. This denominator serves both metrics.

2. Classify every customer's movement: retained flat, expanded, contracted, churned.

3. GRR = (opening − contraction − churn) ÷ opening. Expansion is excluded entirely.

4. NRR = (opening + expansion − contraction − churn) ÷ opening.

5. Logo retention = customers still active at period end ÷ customers at start.

6. Reconcile: opening + expansion − contraction − churn must equal closing exactly before either ratio is trusted.

Where this framing fails

In usage-based models the distinction blurs, because a customer's consumption falling is contraction that no one decided. Treating natural usage variance as contraction makes GRR look worse than the relationship warrants. Segment committed revenue from variable consumption and compute retention on the committed portion, reporting variance separately.

In businesses with very few, very large customers, both metrics become nearly meaningless as rates — one account movement swings them by tens of points. Report absolute revenue movements instead.