Snowflake's S-1 named a retention rate it expected to shrink

The record

On 24 August 2020, Snowflake Inc. filed a Form S-1 registration statement with the SEC ahead of listing its Class A common stock on the New York Stock Exchange. The filing states that Snowflake's product revenue was 227.0 million dollars for the six months ended 31 July 2020, up from 100.6 million dollars a year earlier, and that its customer count reached 3,117 as of that date. The document defines and discloses a dollar-based net revenue retention rate of 158 percent as of 31 July 2020, down from 223 percent a year earlier and 169 percent as of 31 January 2020.

What the documents establish

The S-1 is Snowflake's own registration statement, and the net revenue retention figures in it are a company-disclosed measure, not an independently audited count of customer behavior. The document defines the metric precisely: it tracks a cohort of customers under capacity contracts across a trailing two-year period, and any customer that stops using the platform remains in the calculation at zero revenue rather than being removed. Snowflake states in the same filing that it expects the rate to decrease over time as longer-tenured customers become a larger share of both its base and its product revenue. Snowflake's SEC EDGAR filing record confirms the filing date and accession number for this registration statement and its later amendments.

The operating read

For a founder pricing on consumption rather than subscription, the filing's own language is the more useful lesson than the headline percentage. Snowflake counts a churned account at zero rather than dropping it from the cohort, a stricter convention than some subscription businesses apply, and the company states plainly that the number is expected to fall as its base matures. This is an editorial read: a retention rate published without its cohort definition and expected trajectory tells a reader very little on its own. The consumption model itself, billing only for compute, storage and data transfer actually used, is described as a deliberate departure from ratable revenue recognition, one the company says makes its quarterly results harder to forecast than a typical subscription peer's.

What to check before you decide

A founder citing a retention or expansion metric to investors should be able to answer the same questions this filing answers about itself before treating the number as comparable to anyone else's.

• Is the cohort defined by first usage date, by contract signing date, or by another start point, and does an account that churns stay in the denominator at zero?
• Does the disclosed measurement period allow the metric to be compared against a company of a similar reported revenue scale, rather than across periods of different cohort maturity?
• Is a consumption-based revenue figure being read with the same forecasting certainty as a fixed-term subscription contract, when the filing itself says that certainty does not hold?

Snowflake's own document states the figure, defines its method, and predicts its own decline. That combination, a disclosed number paired with a disclosed reason to discount it, is a more complete record than the percentage alone, and a reasonable standard for any other consumption-pricing filing.