Boards still talk about gross revenue retention as if 90% is the floor for a healthy private SaaS book. Aleph × Benchmarkit's 2026 file says the market already left that floor.
Across 342 B2B SaaS and AI-native companies reporting CY-2025 actuals, the GRR cut (N=226) prints a median of 84%. That is 4 points below the same sample's CY-24 median of 88%. Top quartile fell from 95% to 91%. The 25th percentile dropped from 81% to 76%.
This B2Bcentr report shows which sample produced the 84% median, how GRR differs from NRR and logo churn, what ACV and GTM motion do to the number, and why a Rule of 40 rebound that rides on weaker retention is not durable.
Key Takeaways
- Median GRR is 84% (N=226). Boards still quote ~90%.
- YoY drop: 88% to 84%. Top quartile fell 95% to 91%.
- At 84% GRR you lose 16% of existing ARR to churn and contraction.
- Enterprise ACV ($50K-$100K) prints 91%; Sub-$5K ACV prints 80%.
- Sales-led GRR is 88%; PLG is 79%; Hybrid is 80%.
Which sample produced the 84%
Primary source: Benchmarkit 2026 SaaS & AI-Native Metrics (research partnership with Aleph). Overall file: 342 B2B SaaS and AI-native software companies, CY-2025 actuals. The Gross Revenue Retention cut is N=226. Context write-up: Aleph's 2026 SaaS benchmarks takeaways, which calls GRR "the most concerning metric in the report."
Definition used here: Gross Revenue Retention is the dollars you keep from last year's customers after churn and contraction, before expansion. It cannot exceed 100%. At 84% GRR, companies lose 16% of existing ARR each year before any upsell can fill the hole. That is not the same as NRR, and it is not logo churn.
| Cut | Number | Sample |
|---|---|---|
| Median GRR | 84% | Aleph × Benchmarkit 2026; N=226; CY-2025 |
| CY-24 median (same trend cut) | 88% | YoY: -4 pts into CY-25 |
| 75th percentile (CY-25) | 91% | Down from 95% in CY-24 |
| 25th percentile (CY-25) | 76% | Down from 81% in CY-24 |
| Enterprise ACV ($50K-$100K) | 91% median | Same GRR cut by ACV |
| Sub-$5K ACV | 80% median | 11 pts below enterprise band |
| Sales-led / PLG / Hybrid | 88% / 79% / 80% | Same GRR cut by GTM motion |
Why ~90% folklore still shows up in board decks
Older private SaaS surveys and operator folklore still treat 85% to 90% as the strong-retention band. Benchmarkit's own key insight on this cut says median GRR at 84% sits below the 85-90% goal for strong B2B SaaS retention. Boards round that goal up to ~90% and keep using it as a planning floor.
Aleph's takeaways post is blunt: median GRR fell to 84% in 2025 from 90% three years earlier in the longitudinal view the authors cite, and even the 75th percentile fell from 95% to 91%. Top performers were not immune. That is a market-level shift, not a few CS teams missing targets.
Here's why that matters. If your plan still assumes a 90% GRR floor while the private median is 84%, you will underfund retention work and over-credit expansion for "growth" that is really filling a leak. Read this next to B2Bcentr's NRR report (median near 101-102%) so you do not confuse a thin net number with a healthy gross floor.
GRR is not NRR, and it is not logo churn
NRR adds expansion on top of the GRR floor. A company at 84% GRR and 102% NRR is replacing a 16% gross leak with expansion, not "low churn." That distinction is why Aleph flags retention as fragile even while Rule of 40 improved.
Logo churn is a different clock again. B2Bcentr's SaaS churn report separates Recurly's billing-network 3% median, Lighter Capital's 16% private logo median, and SaaS Capital's older bootstrapped GRR/NRR cuts. Use that post for sample hygiene. Use this post for the 2025 Aleph × Benchmarkit GRR median.
At $5M starting ARR, an 84% GRR implies a $800,000 annual gross leak before expansion. A 90% folklore plan would have expected a $500,000 leak. The $300,000 gap is not a rounding error. It is a hiring plan, a CS budget, and a product roadmap argument.
ACV, growth cohorts, and GTM motion
Enterprise ACV companies in the $50K-$100K band print the highest median GRR at 91%. Sub-$5K ACV companies sit at 80%. Multi-stakeholder contracts and deep workflow integration show up in the number. SMB volatility shows up too.
By 2025 growth rate, the 11-20% growth cohort hits the highest median GRR (89%). Low-growth companies under 10% print 79%. Poor retention and slow growth compress together: low GRR forces higher new-logo requirements at higher CAC.
By go-to-market motion, sales-led companies lead at 88% median GRR. PLG sits at 79%. Hybrid sits at 80%. High-touch relationships (sponsors, CSMs, QBRs) still translate into superior gross retention. Pure product-led motions need commercial overlays to close that gap.
Read it next to expansion, growth, and Rule of 40
Expansion ARR is already 40% of total new ARR at the median in the same Benchmarkit file. See B2Bcentr's Expansion ARR report. When GRR is 84%, a rising expansion mix can mean substitution for leaky logos, not compounding.
Median YoY ARR growth is 20%, not the old 40% folklore bar (ARR growth report). Efficiency metrics improved while the revenue line got more fragile. Rule of 40 jumped from 15% to 25% in the same year (Rule of 40 report), largely because S&M and R&D came down, not because retention healed.
Aleph's synthesis: the median company is more profitable and less defensible than twelve months earlier. GRR is the durability check that Rule of 40 alone will not give you.
B2Bcentr's take
Who this does not work for
Brand-new products with almost no installed base cannot manage to an 84% GRR target the same way a $20M ARR book can. Pre-PMF teams should still prioritize land quality and early logo fit. Usage-based pricing also muddies contraction labels inside GRR. Segment your own CRM definitions before copying the median.
This number also breaks if you redefine GRR to include expansion or to exclude downgrades. Stick to dollars kept after churn and contraction, before expansion, the way Benchmarkit reports it.
What to do Monday
- Pull trailing-four-quarter cohort GRR from billing or CRM (no expansion in the numerator).
- Put GRR next to NRR on one board slide. If NRR is fine and GRR is weak, expansion is hiding the leak.
- If GRR is under 80%, audit product depth, pricing fit, and renewal ownership before hiring another SDR pod.
- Split GRR by ACV band and GTM motion. Do not manage an SMB book to an enterprise 91% target.
- Re-check Rule of 40 gains against the GRR trend. Cost-line wins that ride on a falling GRR are one-time.
Next number for the same retention pack: what net retention actually prints when boards still quote 120%.
FAQ
What is a good SaaS GRR in 2026?
Aleph × Benchmarkit's 2026 file prints an 84% median GRR (N=226). Top quartile sits at 91%. Benchmarkit flags 85-90% as the strong-retention goal band, so 84% is already below that goal at the median.
Is 90% GRR still a realistic planning assumption?
Treat ~90% as folklore or a top-quartile stretch, not as the 2025 private median. The same cut shows the median at 84% and the 75th percentile at 91%.
How is GRR different from NRR?
GRR excludes expansion, so it cannot exceed 100%. NRR adds upsells and cross-sells on top of that floor. You can print 102% NRR while still leaking 16% of starting ARR at 84% GRR.
Why did median GRR fall from 88% to 84%?
Aleph × Benchmarkit shows a market-level drop in CY-25: median -4 pts, 25th percentile 81% to 76%, 75th percentile 95% to 91%. Aleph links the shift to harder renewals and AI-era scrutiny, not only CS execution misses.
Does ACV change GRR?
Yes in this sample. Enterprise ACV ($50K-$100K) prints a 91% median GRR. Sub-$5K ACV prints 80%. Match ACV before you copy a peer median.
Should PLG companies expect the same GRR as sales-led?
Not in this cut. Sales-led median GRR is 88%. PLG is 79%. Hybrid is 80%. Product-led motions usually need commercial overlays to close the gap.
Where can I read the full Aleph × Benchmarkit sample?
Start with the Benchmarkit 2026 report hub and Aleph's takeaways post. Guide page: 2026 SaaS benchmarks.
Primary source: Benchmarkit 2026 SaaS & AI-Native Metrics (Aleph partnership; 342 companies; GRR cut N=226; CY-2025). Context: Aleph takeaways and Aleph guide page.