Net dollar retention benchmarks: what public SaaS companies actually report, and why the numbers do not compare

Net dollar retention is the metric public SaaS companies are most often asked for and least consistently define. This page collects what eight of them actually reported in 2026, with the source, the as-of date, and the definition behind each figure. The definitions are the point. Two companies both reporting “120%” can be measuring materially different things, and a company reporting 97% on its whole base might report 105% on the slice a competitor chooses to show.

The table

Figures as read from the named source on 17 September 2026. Ranges and floors are shown as the company stated them.

Why the numbers do not compare

Who is in the cohort

Figma's 139% is measured on customers that were above $10,000 ARR a year ago. That excludes the long tail of small teams, which is where churn concentrates in every SaaS business. Asana shows what the exclusion is worth: 97% on the whole base, 98% on customers over $5,000, and 98% on customers over $100,000. One point, in Asana's case; it can be much more. When a company reports only its large-customer cohort, the whole-base figure is lower, and usually by more than a point.

What period

Datadog reports a trailing twelve-month figure. Snowflake uses a two-year window and compares year two to year one. DocuSign and Asana report the quarter's figure against the same customers a year earlier. These are different questions with different answers, and none is wrong.

What revenue

Snowflake counts product revenue under capacity contracts; consumption that was contracted but not used is not revenue. DocuSign counts direct customers only, excluding partner-sold. Most others count ARR. A consumption business and a subscription business reporting the same percentage have had different years.

Whether it is disclosed at all

CrowdStrike defines dollar-based net retention in its releases and then describes it qualitatively (“continued strong retention”). Atlassian gives a floor. Datadog gives a range. A company that stops reporting a point figure has usually had a reason, and the benchmarks that still quote its last stated number are quoting history.

What to take from it

The useful reading of this table is not “120% is the bar.” It is that your own NDR is comparable to your own history and only loosely to anyone else's, and that when you report it, the definition matters as much as the number. State the cohort, the period, and what counts as revenue. If you show a large-customer cut, show the whole base beside it. The NRR guide sets out the standard definition and the three ways the figure gets inflated; the NRR calculator runs it on your numbers.

One more thing the public figures hide. Every one of these companies has a billing operation that recovers failed payments before they become churn. At $100M+ of ARR that is a team; at $2M it is usually nobody, and involuntary churn quietly takes two to four points off NDR that a five-email sequence would have kept.

Frequently asked questions

Where these numbers come from

Each row links to the press release, SEC exhibit or earnings-call transcript it was read from, on 17 September 2026. Where a figure was stated on a call rather than in the release, the row says so. Figures are re-checked when the companies report again; if a row is out of date, the linked source is the authority, not this page. Exeechain does not aggregate NDR across its customers and publishes no benchmark of its own.