The table that wouldn't fill
We set out to do something that sounded simple: build a table of gross revenue retention for the top 100 public SaaS companies. GRR is the most honest retention number a subscription business has. It ignores upsells and expansion and asks one question: of the revenue you had a year ago, how much did you keep? It is capped at 100%. You cannot dress it up.
Public companies publish hundreds of pages of filings a year, so we assumed the data was out there and just needed collecting. It isn't. After a pass through 10-Ks, S-1s, and earnings materials, our table of 100 had three clean, current, dollar-based GRR figures in it.
Three cells filled in, 97 empty. It took us a while to accept that the empty cells were the more interesting half of the table.
The disclosure gap, in numbers
The most systematic count we found is Ordway's study of 135 NASDAQ and NYSE SaaS companies. In their sample, 72% reported a net retention metric. Only 7%, ten companies out of 135, reported any gross retention metric at all. A quarter reported neither.
The same pattern shows up at IPO, when disclosure incentives are strongest. Blossom Street Ventures found 41 SaaS companies that disclosed net dollar retention in their S-1, and only 6 of those 41 also disclosed the gross number.
So who actually publishes GRR today? From the 2025 and 2026 filings we verified directly:
- Workday: approximately 97% as of January 31, 2026, down from about 98% a year earlier.
- Vertex: 94% at December 31, 2025, down from 95%.
- Blackbaud: approximately 92% for fiscal 2025.
Three numbers, clustered between 92% and 98%. Which sounds reassuring until you remember the selection effect: the companies willing to publish their gross retention are, almost by definition, the companies whose gross retention survives daylight.
Why the honest number stays in the drawer
Net revenue retention has a property that CFOs love: expansion revenue can bury churn. Lose 10% of your base, upsell the survivors by 25%, and you report 115% NRR. The slide looks great. The leak is invisible.
GRR has no such mercy, and the market knows it. That asymmetry explains the 72% versus 7% split better than any accounting argument. Companies disclose the metric that can absorb bad news and keep quiet about the one that can't.
Datadog is our favorite illustration. Its 10-K defines and reports net retention only. But on earnings calls, the CFO regularly describes gross revenue retention as sitting "in the mid- to high 90s." The number exists. Management tracks it. It is just never printed in a filing where it would sit in a table next to last year's.
Even the published numbers come with asterisks
Reading the definitions behind the three disclosed figures was an education in itself.
Workday computes GRR on constant currency with exchange rates frozen at the start of the fiscal year, and excludes certain contracts under one year. Vertex excludes cancellations that are "replaced by new subscriptions associated with customer migrations to a newer version" of the product, so a customer who churned off an old product and onto a new one never counts as churn. Blackbaud's own 10-K says its 92% improved over the prior year "primarily due to our sale of EVERFI." Retention went up because they sold the leaky division, not because customers stayed longer.
Then there is Rimini Street, which reports an 88% "revenue retention rate" that looks like GRR until you read the definition and discover it includes expansion revenue from existing clients. The words "gross" and "net" appear nowhere in the filing.
And on the NRR side, where most of the disclosure lives, Ordway found about half the companies use a non-standard calculation, 29% smooth the number with trailing 12-month averaging, and the sample used more than 30 different metric names and 8 distinct formulas. Zoom's cohort covers enterprise customers only. Snowflake measures over a two-year window that includes only capacity-contract customers. A 2019 KeyBanc study counted 110 different ways companies calculate this one metric.
None of this is fraud. Every definition sits in a footnote, disclosed and auditable. But it means a retention number without its definition is not information. It is decoration.
What this means if you run a private SaaS company
Somewhere in your last board deck, or your next one, there is probably a line comparing your NRR to public comps. Now you know what it is being compared against: numbers chosen because they flatter, calculated 110 different ways, with the honest metric quietly omitted by 93% of the sample.
Three things we would do instead. Compute your own GRR the strict way: no migration carve-outs, no cohort exclusions, no smoothing. Write the definition down in one place, because when definitions live in nobody's head and everybody's dashboards, the numbers drift. And when you benchmark, use a base rate that includes companies with nothing to show off: SaaS Capital's survey of private B2B SaaS companies puts median gross retention at 91%. If your strict GRR clears that, you are doing fine, whatever the public comps slide says.
If you want a second pair of eyes on how your retention metrics are defined, and whether they would survive a diligence read, that is exactly the kind of work we do.