ARR is Climbing But it’s Harder to Grow
Is your NRR masking “revenue erosion”?
SaaSCFO’s 2025 benchmarking data shows median gross retention (GRR) across SaaS dropped four points year over year. The 75th percentile fell from 95% to 91%. Retention is getting harder industry-wide, not just in a handful of struggling categories.
That data point introduces a concept worth knowing calculating: revenue erosion.
Revenue erosion = (1 − GRR) × ARR
That’s the revenue hole you refill from scratch every year, before a single new logo or expansion dollar counts toward growth. The number itself isn’t the test — every business has some erosion. The test is whether it grew year over year. If it did, growth is getting structurally harder no matter what the top-line chart shows.
The second test comes from Benchmarkit: Expansion bookings sit at 40% of net new ARR at the median SaaS company. Below that line, expansion amplifies new-logo growth. Above it — and rising — expansion has stopped amplifying and started substituting. It’s covering for a new-logo engine that isn’t working.
Why it matters
This is a Business Model & Capital signal, not a Market Position one. It’s invisible in the metric every board deck leads with — ARR — and mostly invisible in NRR too, since NRR blends retention and expansion into a single number that can look healthy while both underlying trends move the wrong way. A company can post 110%+ NRR and still be getting structurally harder to grow, if the erosion hole is widening and expansion is quietly substituting for a new-logo engine that’s stalled.
Before your next board meeting
Run both tests on your own numbers, same prior-year-close-to-Q2 window, same math:
Revenue erosion = (1 − GRR) × ARR. Did it grow year over year? Not the level — the direction.
Expansion ÷ net new ARR. Where do you sit against the 40% network median — and is this ratio higher than a year ago? If it’s already above 100%, expansion is currently doing more work than your entire net-new number, which usually means new-logo growth and churn are roughly canceling each other out.
It’s a leading indicator of how hard next year’s growth is going to be, months before it shows up in the number everyone’s already watching.
What to do about it
1. Give erosion an owner and a number.
In most companies this metric belongs to nobody. Churn sits with CS, pricing sits with product, new logo sits with sales, and the dollar hole sits in a gap between all three. Name one executive who owns revenue erosion as a number they report monthly, the way the CRO owns pipeline. Then put it on the same page as bookings so the two get read together.
I’d also add expansion ARR as an explicitly reported figure rather than something you back into from the NRR/GRR spread. When I ran this across a set of venture-backed companies, a handful produced an arithmetically impossible new-logo number once you decomposed their reported retention rates — which told me the two rates were being measured on different cohorts or different windows. You can’t manage a number you’re inferring.
2. Run retention as a revenue function, not a support function.
The companies holding GRR above 91% don’t have better products. They have a different operating system around retention. Customer success is measured with the same rigor as sales, at-risk signals are quantified and acted on 60 to 90 days before renewal rather than at the renewal date, and there’s a retention P&L that breaks renewal rate down by segment, cohort and product.
The practical version: build the right side of the Bowtie with the same discipline you built the left. Winning by Design’s Revenue Architecture model treats onboarding, adoption and expansion as stages with conversion rates and time-in-stage metrics, exactly like your sales stages. Most companies I see have four defined stages before the close and none after it. Net revenue retention is either engineered or accidental, and in most of these businesses it’s accidental.
3. Check whether your erosion is a churn problem or a pricing-architecture problem.
This is the one people get wrong. A contract that renews at 85% of last year’s value doesn’t look like churn — logo retention is 100%, nobody canceled, CS reports a clean quarter. But it lands in GRR identically. If your pricing is per-seat and your customers are reducing headcount through AI-assisted work, your revenue follows their headcount down even when you never lost a competitive deal.
So the question is whether you have a defined unit of value beyond the seat. Three tests: your customer’s CFO understands the unit without explanation, your ops team can count it without building new instrumentation, and it grows with the value your product creates rather than with something orthogonal to it. If you can’t name that unit, no amount of customer success work fixes your erosion — you’re defending a meter that’s structurally disconnected from the value you deliver.
The destination for most horizontal SaaS with compressing retention is a hybrid model: seat baseline plus a metered layer. That used to be a billing-infrastructure problem. Since Stripe acquired Metronome in January, it isn’t one anymore, so “our billing system can’t support consumption” is no longer a real reason to wait.
4. Change your renewal calendar before you change anything else.
This is the highest-leverage move you can make in 30 days. Procurement teams now walk into renewals with utilization audits and AI-driven seat-reduction models — there are buyer-side playbooks that train them to do exactly that. A vendor arriving 90 days out with a status-update QBR loses that conversation before it starts.
Move your top 20 accounts to a 6-to-9-month renewal cycle. Replace the next QBR agenda with a pricing-evolution conversation anchored in the customer’s own usage data, so the renewal is a confirmation rather than a negotiation. Where you have to concede a seat reduction, trade it for a multi-year consumption floor instead of just granting it.
5. Fix the new-logo engine, because that’s what expansion dependency is actually telling you.
If your expansion ratio is climbing, the honest read is usually not that expansion got better. It’s that new logo got worse and expansion is covering the gap. Expansion is the easier number to move — it runs through accounts that already trust you, with a team that already knows them — so it’s where organizations naturally drift when acquisition gets hard.
Be specific about which part broke. Top-of-funnel volume, conversion rate, or sales capacity are three different problems with three different fixes, and the temptation is to fund more expansion motion because it produces a faster number. Resist that for one planning cycle and find out what the acquisition engine actually needs.
