Metrics, retention and the organization · Answered by Golden Section from more than 400 B2B software companies observed
Above 100% is good, 105% or better is what we would call great, and anything below 90% is a fix-first problem. For context, the private B2B SaaS median in Benchmarkit's 2025 data was about 101% net and 84% gross. Read NRR together with gross revenue retention, because a strong net figure can hide a shrinking base that expansion keeps refilling. We treat gross retention as the master variable: at 95% or better a software company can sustain very high cash-flow margins, and below 80% we treat it as disqualifying. To push NRR above 100%, fix gross churn first, then sell expansion on purpose. Start by splitting your ARR bridge into churn, contraction and expansion by cohort.
Judge retention on gross first and net second, measured on your core book with provisional and seasonal accounts reported separately. Fix gross churn before investing in expansion or new-logo growth.
| Metric | Value | What it means | Source |
|---|---|---|---|
| Median net dollar retention | 101% (2025) | ARR from existing customers a year later, including expansion, ÷ starting ARRPrivate B2B SaaS, n ≈ 340–580; down from 105% in 2021 | External benchmarkBenchmarkit annual B2B SaaS benchmarks, as cited in GS Investing in Software addendum |
| Median gross revenue retention | 84% (2025) | starting ARR retained after churn and contraction, excluding expansionFell from 88% in one year at every quartile; upper quartile 91% | External benchmarkBenchmarkit annual B2B SaaS benchmarks, as cited in GS Investing in Software addendum |
| NRR, what great looks like | 105%+ | net revenue retentionBalanced Path benchmark | Golden Section, publishedThe Balanced Path |
| GRR to underwrite the margin ceiling | 95%+ | gross revenue retentionBelow 80% is disqualifying; each point below 95% costs 1.3–2.0 points of steady-state margin | Golden Section, publishedInvesting in Software addendum, September 2026 |
| GS investing thresholds | NRR above 100%; pass below 85% | net revenue retentionLending minimum is 90% NRR for revenue-based financing, 95% for term loans | Golden Section, publishedGrowth Equity and Growth Capital Lending pages |
| Retention by pricing model | 108% usage vs 98% seat | net retention by pricing architectureA ten-point structural gap | Golden Section, publishedInvesting in Software addendum, September 2026 |
Retention is the property that makes recurring revenue worth more than other revenue. Our own research found that gross retention sets the achievable cash-flow margin directly: a company keeping 96% of revenue spends five to eight points of revenue a year replacing churn, while one keeping 84% spends twenty-one to thirty-two, before any growth. That is also why a steady net figure can mislead. Expansion now supplies a larger share of net new ARR, which means many top lines rest on selling more to a shrinking base.
Measurement comes first. A clean ARR schedule separates churn, contraction and expansion by customer and cohort. Customers you could not underwrite at signature belong in their own line, as core and provisional segmentation describes, or they drag down the number your valuation rests on.
Then the fixes. Churn identification finds at-risk accounts before renewal. Adoption and account management get customers to the value they were sold, which is the precondition for expansion. And expansion needs a pitch; customers rarely grow on their own.
A $3M company reports 103% NRR and considers it healthy. Split by motion, gross retention is 86% and expansion is 17 points, most of it from price increases on three large accounts. Four customers signed during a pilot program account for almost half the churn. The CEO reports those accounts separately to the board, which shows core gross retention of 91%, opens a provisional revenue line for future pilot customers from their first invoice, and assigns the churn review to the three remaining at-risk core accounts before their renewals. Expansion targets move from price to a packaged services offer. All figures are invented for illustration.
Seasonal verticals where customers pause and return should report those pauses separately; counting them as churn understates retention every year. Early companies with a few dozen customers will see large swings from single accounts and should read retention by cohort, not as one number.
From the Golden Section mistakes list, each paired with the play that prevents it.
Blending unproven customers into core revenue drags the retention figure every stakeholder prices.
Counting seasonal pauses as churn understates retention in vertical markets that close for part of the year.
Expansion does not happen without selling, and NRR above 100% depends on it.
In the order we would run them. Each is on its own page, most with a free Excel template.
Separates churn, contraction and expansion by customer so retention can be measured at all.
Keeps unproven accounts out of the core retention number.
Flags at-risk accounts before renewal so churn can be prevented.
Gets customers to full use, which lowers churn and opens expansion.
Runs onboarding through renewal as a managed process with expansion built in.
Customer plays Retention is built across the customer plays, from the ARR schedule through onboarding, adoption and renewal.
The private median was about 84% in 2025 and the upper quartile about 91%. We look for 95% or better when underwriting a company's long-run margin, and treat anything below 80% as disqualifying.
Reduce gross churn first, because every point of churn has to be won back before expansion counts. Then make expansion a sold motion with an owner, a pitch and a target, and consider whether your pricing unit grows with the customer's activity rather than their headcount.
GRR, for most decisions. NRR can look healthy while the base shrinks and a few accounts expand, and that pattern is fragile. Report both, and never let a good net figure excuse a falling gross one.
Retention decides which capital fits. Our lending starts at 90% NRR, our equity looks for NRR above 100%, and below 85% the honest answer is to fix churn before raising anything.
Talk to Golden Section →Reviewed by Dougal Cameron, CEO & Co-Founder on 2026-09-23. Golden Section observations are labeled separately from external benchmarks and illustrative arithmetic.