The test, applied
Subscription license fees pass. So does contracted support and maintenance billed at a fixed rate alongside the subscription, along with committed minimums on a consumption contract, up to the committed amount.
Implementation fees fail, because they are charged once per customer. Custom development and training days sold by the session fail for the same reason. Hardware resold at pass-through fails. So does consumption above a commitment, which is genuine revenue with no contractual obligation behind it.
The failing items are not lesser revenue. Services work often carries the customer through onboarding and shows up later as retention. It just does not belong in the number used to measure durability.
The gray zone
Three cases cause most of the arguments.
Annual retainers for advisory or managed services look recurring and often renew. They pass the test only when the fee is fixed and the renewal is contractual rather than a fresh sale each year.
Overage on usage contracts recurs in practice at accounts that consistently exceed their floor. It still fails the contractual half of the test, which is why the standard treatment keeps it in a separate variable line.
Multi-year contracts with contracted step-ups are recurring at the rate in effect for the current period. The scheduled increase is contracted future revenue, not current recurring revenue.
Why the boundary decides other numbers
Every retention metric divides by a base of revenue that was supposed to recur. Put a $200,000 implementation fee into that base and the following year reads as a $200,000 contraction, even though nothing was lost. Net revenue retention breaks the moment the denominator includes revenue that was never going to repeat.
Capacity planning breaks the same way. A team that treats services attach as recurring will build next year's quota on a base that partly evaporates on January 1.
Write the policy down
Publish the rule, name the specific line items on each side of it, and route every non-standard contract through the same reading. Forecasts miss most often because the model is built on assumptions that quietly stopped holding, and an ARR definition that drifts by deal is exactly that kind of assumption. Grounding the revenue forecast on a stable recurring base keeps the variance conversation about the market rather than about the definition.
Frequently Asked Questions
Do implementation and onboarding fees count as recurring revenue?
No. They are charged once per customer and have to be resold with every new logo. They belong in a services line, which is why a quarter heavy on implementation work inflates run rate while ARR stays flat.
Is a one-year contract that auto-renews recurring revenue?
Yes. Auto-renewal is the clearest version of the test, since the revenue continues unless someone acts to stop it. A one-year contract with no renewal clause still counts if the product and the relationship are built to renew.
Does support or maintenance revenue count?
It counts when it is contracted at a fixed rate on a fixed interval and renews alongside the subscription. Ad hoc support billed by the hour does not, because the amount depends on what happened that month.
Why does the boundary matter if the cash is the same?
Retention math and capacity planning both run off recurring revenue. Mixing non-recurring dollars into the base makes retention look better than it is and makes the revenue base look more durable than it is.
Put these metrics to work
ORM builds custom revenue forecast models that turn concepts like what counts as recurring revenue? into prescriptive action for your team.
Schedule a Demo