Net Revenue Retention (NRR)

NRR is calculated from a fixed cohort: take the recurring revenue that cohort generated at the start of a period, then track only that same group forward. The formula is: NRR = (Starting MRR + Expansion − Contraction − Churn) ÷ Starting MRR, usually expressed as a percentage and most often measured over a trailing 12 months.

Three movements lift or drag the number. Expansion is more revenue from the same customers — additional seats, higher usage tiers, or new modules adopted. Contraction is existing customers paying less without leaving (downgrades, seat reductions). Churn is revenue lost when a customer cancels entirely. New-customer revenue is deliberately excluded; mixing it in produces a vanity number that hides whether the base itself is healthy.

Worked example: a cohort starts a year at $100k MRR. Over twelve months it gains $25k in expansion, loses $8k to downgrades, and $12k to cancellations. NRR = (100 + 25 − 8 − 12) ÷ 100 = 105%. That cohort grew 5% with zero new logos.

NRR above 100% is the dividing line between a leaky bucket and a self-reinforcing one. Below 100%, the existing base shrinks each period and new sales must first refill the loss before the company grows — you are running to stand still. Above 100%, the base compounds on its own, so every new customer adds to a growing foundation rather than patching a draining one.

This is why investors and operators treat NRR as a durability signal. It captures product-market fit, pricing structure, and customer success in a single ratio: a number that holds above 100% over time usually means the product gets more valuable as customers use it, and that expansion outruns the churn no business fully escapes.

As a general rule of thumb, healthy SMB-focused SaaS often lands somewhere near 100%, while strong enterprise products with usage- or seat-based expansion frequently report meaningfully above that. Benchmarks vary widely by segment, contract model, and how a company defines the cohort window, so treat any single target with caution and compare like with like.

Three failure modes distort NRR. First, contaminating the cohort with new-customer revenue — that is gross new-business growth, not retention. Confusing it with Gross Revenue Retention (GRR), which caps at 100% because it ignores expansion, is the second: GRR shows pure leakage, NRR shows net of leakage plus growth, and they answer different questions. Third, reading NRR without segmenting — a blended 110% can hide a healthy enterprise tier masking a hemorrhaging SMB tier.

NRR is only trustworthy when subscription revenue, customer identity, and product behaviour line up on the same records. In stitched-together stacks, billing lives in one tool and usage in another, so reconciling expansion versus contraction by cohort turns into a manual, error-prone export job — and the segmentation that makes NRR actionable rarely gets done.

A product operating system built on a shared spine keeps revenue (MRR/ARR), the customer record, and product analytics joined as one object, so a cohort's expansion and contraction can be read directly against the accounts driving it. With customers, revenue, and behaviour on the same record, you can move past a single blended NRR figure toward seeing which accounts and segments are expanding, contracting, or at risk.