Net Revenue Retention: How Filers Define It

Two Retention Figures Can Share a Label and Still Measure Different Things

Net revenue retention is the figure most software models start from: how much of last year’s recurring base came back, after losses and after expansion. The label is shared. The measurement is not. Before you put two percentages in the same column, read four choices in the filing: which customers enter the cohort, how long the window is, whether the base is ARR or recognised revenue, and which customers are excluded.

A 125% on a two-year capacity cohort is not the same object as an 8% attrition rate or a 98% renewal rate. Treat each as its filer’s own measure.

What the Six Filers Disclose

Three of the six publish a retention-style number; three do not. The table records the measure, the fiscal year, and the definition that sits beside it in the filing.

Snowflake’s 125% is a net figure: expansion is inside it. Salesforce attrition and ServiceNow renewal are gross-style measures: they track only what was lost. Microsoft, SAP and Constellation leave the slot empty; their growth and recurring-share lines still teach the business, but they do not give you a retention input you can paste into a peer table.

How CrowdStrike and Workday Write the Definitions

Outside the six, two filers publish definitions that separate gross from net cleanly. They sit outside the peer set but work as a dictionary.

CrowdStrike (Q4 FY2026 earnings exhibit) starts with Prior Period ARR from all subscription customers twelve months earlier. Dollar-based net retention divides Current Period ARR from those same customers (including expansion, net of contraction and churn, excluding new customers) by Prior Period ARR. Dollar-based gross retention deducts ARR from customers who left, then divides the remainder by Prior Period ARR; expansion never enters.

Workday (SEC correspondence, September 2024) defines gross revenue retention as current period-end ARR of retained customers, excluding add-ons and net expansions, divided by total ARR at the corresponding prior period end. Net revenue retention uses the same retained-customer set but includes add-ons and net expansion in the numerator.

Those four sentences are the working vocabulary. A filer that only prints “net retention” without the cohort and the window has not told you which of them it is running.

What One Point of Retention Does in a Multi-Year Model

In a recurring-revenue roll-forward, opening recurring revenue is multiplied by net retention (or by gross retention plus a separate expansion term), then new-customer revenue is added. Raise net retention by one percentage point and the same opening base produces a larger closing base. The next year starts from that larger base, so the one-point gap compounds. Over a multi-year path the difference between two retention assumptions widens even if new-customer revenue is held fixed.

That is why the definition matters before the number. A point measured on a two-year cohort is not interchangeable with a point measured on trailing twelve-month ARR, and neither is interchangeable with a renewal rate that never includes expansion. Usage-priced platforms make the same point sharper, because consumption can move the retained base without any seat change.

How to Read a Retention Line in a Filing

1. Find the metric name exactly as printed (net revenue retention, dollar-based net retention, renewal rate, attrition).
2. Note the period (trailing twelve months, two-year cohort, multi-year average).
3. Note the base (ARR, ACV, recognised product revenue).
4. Note exclusions (new logos, acquisitions, self-service, services lines).
5. Only then compare the percentage to another filer, or feed it into a reverse DCF as an input assumption.

Snowflake’s profile walks the 125% two-year definition against product revenue share; Salesforce and ServiceNow profiles keep attrition and renewal in their own columns rather than forcing a single peer metric.