Diagnostics · Answered by Golden Section from more than 400 B2B software companies observed
Stop treating new sales as the fix and spend the next quarter on retention. Below 90%, the company loses more than a tenth of its recurring revenue every year after expansion, so every new logo first refills the hole. Split the number before acting: gross revenue retention tells you how much walks out, expansion tells you how much the survivors grow, and the two need different work. Then find where the loss concentrates by cohort, segment, contract type and reason, and separate seasonal pauses and unproven accounts from real churn so you are fixing the right thing. Most of the answer usually sits in onboarding and adoption. Next step: rebuild the ARR schedule with churn, contraction and expansion broken out by customer for eight quarters.
Retention is fixed before growth is funded. Below 90% NRR, capital raised or borrowed mostly finances the leak, and every serious source of it will price that or decline.
| Metric | Value | What it means | Source |
|---|---|---|---|
| Net revenue retention floor | 85% (equity) / 90% (lending) | recurring revenue from a starting cohort 12 months later, including expansion, ÷ its starting recurring revenueGS equity passes below 85%; GS lending requires 90%+ for revenue-based financing and 95%+ for term loans | Golden Section, publishedGolden Section Equity and Lending pages |
| Net revenue retention target | 105%+ | as aboveWhat great looks like for capital-efficient vertical SaaS on The Balanced Path | Golden Section, publishedThe Balanced Path |
| Private B2B SaaS medians, 2025 | NRR 101%, GRR 84% | net and gross dollar retention, private B2B SaaSBenchmarkit annual benchmarks (n ≈ 340–580), as cited in GS's 2026 addendum; gross retention fell from 88% at every quartile | External benchmarkBenchmarkit via Investing in Software addendum (Sept 2026) |
| Margin cost of churn | 1.3–2.0 points per point | steady-state cash-flow margin lost per point of gross retention below 95%GS's re-test of its 2020 valuation framework | Golden Section, publishedInvesting in Software addendum |
Retention is the variable that sets what the company is worth. It decides whether growth compounds or refills, and our own re-test of software economics found that gross retention sets the achievable margin directly. A board, a bank and an acquirer all read the same number and price the whole book off it, which is why a blended figure that hides a weak cohort costs you twice.
The work runs in a fixed order. First get the measurement right on the ARR schedule: committed recurring fees only, every change on its own line. Then take out what is not churn. A seasonal customer who pauses every winter belongs in its own ledger with a return date, and an account you could not underwrite at signature belongs in a provisional line outside core retention. What remains is real loss, and the churn identification process finds its predictors. The fixes usually land in account management: a customer who never reached the value promised in the sale leaves at renewal, and one who reached it without anyone pitching expansion never grows.
A company at $2.5M ARR reports 87% NRR. Split apart, gross retention is 80% and expansion adds 7 points. The churn review finds two things. About a third of lost ARR is a seasonal segment that cancels every November and re-signs in March, now tagged and tracked separately. Most of the rest is small accounts sold through a partner and never fully onboarded, lost in their first year. The founder tightens qualification for that channel, rebuilds the onboarding handoff and puts an account manager on a weekly customer-meeting quota. Real gross retention, measured the same way for four quarters, is the number that goes to the board. All figures are invented for illustration.
A company deliberately exiting a legacy product or a segment it no longer serves can run below 90% for a planned period; say so in the board materials and report the continuing book separately. Very early companies with a handful of customers should treat the percentage as noise and read each loss individually.
From the Golden Section mistakes list, each paired with the play that prevents it.
One blended retention number lets the weakest cohort set the price of the whole book for the board, the bank and the buyer.
In seasonal verticals, logging a pause as churn can push reported NRR below 90% when the core business is healthy.
Low NRR is often an expansion problem as much as a churn problem, and expansion does not happen without a pitch.
Some of the churn is customers who should have been let go earlier, and holding onto them distorts the product and the number.
In the order we would run them. Each is on its own page, most with a free Excel template.
Gets the measurement right, with every expansion, contraction and churn event on its own dated line.
Separates seasonal pauses from real churn with a graduation rule, so the reported number is the true one.
Keeps unproven customers out of core retention until they earn their way in.
Finds the predictors of real churn and turns them into an early-warning process.
Fixes the onboarding, adoption and renewal stages where most retention is won or lost.
Customer plays Retention is built across onboarding, adoption, account management and the ARR schedule, and all of them live here.
Work both halves separately. Reduce what leaves by fixing onboarding and adoption so customers reach the value promised in the sale, and grow what stays by running a deliberate expansion motion, because customers do not expand without a pitch. Measure each half every month on the ARR schedule.
Find the predictors first: analyze churned accounts against hypotheses about fit, use, support and satisfaction, then build a process that flags at-risk accounts before renewal. Most gross churn in vertical software traces back to a weak onboarding handoff or a customer who should not have been sold, so qualification and onboarding are usually where the fix lives.
Fix retention first. Capital raised against leaking revenue funds the leak, and diligence will find the number anyway. A quarter spent on retention usually changes the terms more than a quarter spent pitching.
Below 90% NRR this is a fix-first situation for us on both sides: our equity passes below 85% and our lending starts at 90%. Once retention is repaired, the same work that fixed it is what makes the company financeable.
Growth capital lending →Reviewed by Dougal Cameron, CEO & Co-Founder on 2026-09-23. Golden Section observations are labeled separately from external benchmarks and illustrative arithmetic.