GRR vs NRR — A Field Guide

Gross revenue retention and net revenue retention answer different questions, and conflating them is the single most common mistake I see in SaaS board decks. GRR tells you how much revenue you keep. NRR tells you how much revenue you keep plus grow from your existing base. Both are legitimate metrics; the trouble starts when a company reports only NRR because it is the flattering one and lets a real retention problem hide behind expansion revenue.

The formulas, precisely

Start with a cohort of customers and their monthly recurring revenue at the start of the period. Gross revenue retention is starting MRR, minus churn, minus downgrades, divided by starting MRR — and it is capped at 100%, because GRR never includes expansion:

GRR = (Starting MRR - Churned MRR - Downgrade MRR) / Starting MRR
NRR = (Starting MRR - Churned MRR - Downgrade MRR + Expansion MRR) / Starting MRR

-- GRR is capped at 100% (it only measures loss)
-- NRR has no upper cap (expansion can push it well past 100%)

Net revenue retention adds expansion MRR — upsells, seat growth, add-on modules — back into the numerator. That's the only difference, but it changes what the number can tell you. NRR can exceed 100% if expansion outpaces churn and downgrades; GRR structurally cannot, because it only measures what you lost.

Why you need both, not one

A SaaS company can post 115% NRR while its GRR sits at 82%. That combination means the product is leaking customers or shrinking accounts fast, and the headline number is being rescued by a handful of large accounts expanding. Investors who only see NRR will miss this. GRR is the metric that isolates the health of the core product experience — onboarding, support, whether the thing does what it promised — independent of your expansion motion (pricing tiers, seat-based growth, cross-sell). If GRR is weak, no amount of expansion sales fixes the underlying churn problem; it just buys time.

What good looks like

Public SaaS benchmarks (Bessemer's state-of-the-cloud reports and similar annual surveys) generally put best-in-class NRR in the 120%+ range for enterprise-heavy, expansion-driven products, and "healthy" GRR in the low-to-mid 90s. SMB-focused products with lower price points and higher logo churn typically post lower numbers on both — GRR in the 80s and NRR closer to 100% is common and not alarming for that segment. The right benchmark depends heavily on customer size and contract length; comparing a self-serve SMB tool's GRR to an enterprise platform's GRR is comparing different businesses, not different execution.

Measure by cohort, not blended

A blended monthly NRR across your whole customer base smooths out exactly the signal you need. The useful view is a cohort retention curve: take everyone who started in a given month, and track what fraction of their original MRR remains (or grows) in each subsequent month. Plotted across several cohorts, this shows whether retention is improving release over release, or whether a product change in a particular quarter quietly made churn worse. A single trailing-twelve-month NRR number cannot show you that — it is already an average of effects that may be moving in opposite directions.

Which one to lead with, and when

For internal decision-making, lead with GRR. It tells the product and customer success teams whether the core offering is retaining value on its own merits. For investor updates and board decks, report both, explicitly labeled, and be ready to explain the gap between them if it's large. A wide GRR-to-NRR spread isn't automatically bad — it can mean you have a strong expansion motion — but it deserves an explanation rather than letting the higher number stand in for the whole story.

Wrapping up

GRR and NRR are not competing metrics — they're a diagnostic pair. NRR tells you whether the existing base is growing in revenue terms; GRR tells you whether that growth is masking churn. Report both, cut them by cohort and account size before drawing conclusions, and treat a wide gap between them as a question to answer rather than a footnote to skip.