NRR and GRR Benchmarks (2026)

Retention gets better as your customers get bigger, and worse as your company does.

Those are different cuts of the word size and they point in opposite directions. Sorted by contract value, gross retention runs 70-80% for SMB customers and above 90% for enterprise ones (Bessemer State of the Cloud, 2019). Sorted by the size of the company doing the retaining, it goes the other way: 92% under $1M ARR against 88% above $50M (High Alpha SaaS Benchmarks, 2025). The previous edition put the same comparison at 92% against 83% (High Alpha SaaS Benchmarks, 2024).

Net retention conceals the second pattern almost completely. Over that same company-size range it moves from 100% to 101% (High Alpha SaaS Benchmarks, 2025). Expansion is doing the work, and it's covering a widening gap in what gets kept.

So read both, and read them against the same cut.

Both start from the revenue of a cohort at the beginning of a period and ask what it is worth at the end. They differ in what they let count.

Gross revenue retention counts only the losses. Start with the cohort's revenue, subtract downgrades and cancellations, ignore every upgrade, and divide by where you started. It cannot exceed 100%, because nothing in the formula can add. A GRR of 88% means 12% of the revenue you had walked out, whatever you sold to the customers who stayed.

Net revenue retention counts the gains too. Same cohort, same period, but expansion is added back before dividing. It can and often does exceed 100%, which is what makes it the metric investors ask for and the metric that flatters. A business can post 110% net retention while losing a fifth of its revenue base each year, provided the customers who remain keep buying more.

Neither includes revenue from customers acquired during the period. That's the whole point: both are asking what happened to the book you already had. The NRR calculator will work yours out with expansion and contraction separated, which is the split the rest of this page is cut on.

Published retention figures usually differ by more than measurement error, because the panels behind them are different businesses. Operator surveys sample private B2B SaaS companies that chose to answer. Public comps cover companies large enough to be listed and to disclose the metric. A billing panel covers whoever runs their subscriptions on that platform, which skews smaller and more self-serve.

The spread below is roughly sixty points, and almost none of it is disagreement between the publishers. Read the population line before the number.

This is the cut where gross and net retention come apart. Gross dollar retention falls through the middle of the range and recovers most of the way at the top, while net dollar retention rises steadily (OpenView SaaS Benchmarks, 2022).

High Alpha's series covers the same span and shows the same shape: gross retention runs 92, 92, 88, 90 and 88 across its five bands while net retention runs 100, 104, 103, 103 and 101 (High Alpha SaaS Benchmarks, 2025). The previous edition put the gross comparison at 92% under $1M against 83% above $50M (High Alpha SaaS Benchmarks, 2024). Two publishers, three series, and gross retention declines with company size in all of them, but all three are operator surveys. Measured billing data runs the other way at every point in the distribution bar the top decile, which the SaaS churn rate benchmarks page sets out.

Sorted by what a customer is worth rather than by what the vendor is worth, the pattern inverts. Larger contracts tend to churn less and expand more, on both metrics, and the ranges do not overlap much (Bessemer State of the Cloud, 2019).

This is the cut to use when you're deciding whether your own number is a problem. A business selling $8K contracts and holding 80% gross retention is at the top of its band. The same 80% on $60K contracts is well below it.

How you charge moves net retention about ten points, which is more than growing fifty times larger does. Seat-based pricing reports 98% net retention against 108% for usage-based (Benchmarkit SaaS and AI-Native Metrics, 2026). Bessemer put the same difference at around ten percentage points (Bessemer State of the Cloud, 2021), and High Alpha reported hybrid models, which combine a subscription with a variable component, ahead of both (High Alpha SaaS Benchmarks, 2025), though it published the ranking without a figure behind it.

The mechanism isn't subtle. Usage-based revenue typically expands when customers do more, without anyone signing anything. Seat-based revenue expands when a customer decides to buy more seats, which is a decision someone has to make and can decline to make.

Gross retention tells a different and more cautionary story, so check both before concluding that a pricing model has fixed your retention.

The ten disclosures below all report a retention metric. They don't disclose the same one.

The name varies: net revenue retention, dollar-based net retention, net ARR expansion rate and renewal rate all appear, for the same family of measurement. The denominator varies: ARR for some, recognized revenue for others, annualized MRR for others again. The window varies, from a quarter measured against the year-ago quarter to Snowflake's trailing two-year cohort (Snowflake 10-Q, 2026). What is excluded varies: CrowdStrike's rate leaves out its incident response business (CrowdStrike 10-K, 2026). And one company changed its method mid-series, HubSpot restating net revenue retention so that 103.5% on the current basis sits beside 103.9% on the previous one for an overlapping period (HubSpot 10-K, 2026).

The same problem appears within a single publisher across editions, which is the clearest warning available. Benchmarkit's 2024 edition reported usage-based pricing with gross retention three points lower than other models (Benchmarkit B2B SaaS Metrics, 2024). The following edition reported usage-based at 92% against 88% for subscription and hybrid, four points higher (Benchmarkit B2B SaaS Metrics, 2025). Same publisher, consecutive years, opposite sign. Neither figure is wrong; they are different panels answering in different years, and the report that flipped said so.

So a benchmark is usable once you know its population, its denominator and its window. Take one without those and you're comparing your number to a number, not to a benchmark.

These are the primary disclosures, read off each company's own reporting. Nine are taken from the filing itself; Atlassian does not put the metric in its 10-K or 10-Q and reports it in quarterly shareholder letters instead. They're the most checkable retention figures that exist, and they're also the least representative: a company has to be large enough to be listed and to have chosen to disclose the metric.

Worth noting where the aggregators and the filings disagree. Meritech's per-company view, as of 26 August 2026, puts HubSpot at 102% and Datadog at 120% (Meritech SaaS Comps, 2026). HubSpot's own 10-K states 103.5% on its current method (HubSpot 10-K, 2026), and Datadog discloses a band instead of a point, low-120%'s (Datadog 10-Q, 2026). Neither is an error, exactly. They are different as-of dates and, in Datadog's case, a different degree of precision than the company itself offers.

The number that circulates as the bar is 120%, which is where the best tier starts rather than where the middle sits (Bessemer State of the Cloud, 2023). The median across public software companies reporting the metric is 109%, as of 26 August 2026 (Meritech SaaS Comps, 2026), computed across the 87 companies that disclose one.

Among private B2B SaaS operators the central figure is lower still, at 102% (Benchmarkit SaaS and AI-Native Metrics, 2026) and 101% the year before (Benchmarkit B2B SaaS Metrics, 2025). For a venture-backed enterprise business, the tiers below are a workable target set, and they pair each net retention level with the logo retention that tends to come with it (Bessemer State of the Cloud, 2023).

What retention is worth is easier to state than what yours should be. Moving up one net retention band is associated with about five percentage points of additional growth, and the highest band carries a 173% growth premium over the population median (SaaS Capital Research Brief 34, 2026). High Alpha put the same relationship at a 2.5x growth-rate premium for high-NRR companies over low-NRR ones (High Alpha SaaS Benchmarks, 2024).

If you want a target rather than a benchmark: hold gross retention as high as your customer size band allows, and let net retention be whatever your pricing model and expansion motion produce. Gross retention is the one you control directly.