Key Takeaways

• NRR adds expansion to the numerator, which is the only reason it can exceed 100%; gross revenue retention leaves expansion out and caps at 100%.

• On a five-customer book with $400,000 of starting ARR, $65,000 of expansion, $18,000 of contraction, and $45,000 of churn, NRR lands at 100.5% and GRR at 84.25%.

• Leaving one new $75,000 logo in that book's numerator reports 119.25%, a 19-point overstatement caused entirely by a cohort error.

• Snowflake reported net revenue retention of 125% for fiscal 2026, and Twilio reported a Dollar-Based Net Expansion Rate of 116% for Q2 2026 against 108% a year earlier.

• Under consumption pricing, a customer whose workload grows on its own registers as expansion with no upsell behind it, so NRR partly measures your customers' growth rather than your commercial execution.

Why does expansion revenue push NRR above 100% when GRR stops at it?

Expansion sits in the numerator of NRR and nowhere in GRR, which is what separates the two numbers. Both start from the same denominator, the cohort's recurring revenue twelve months ago. NRR adds upsells, seat additions, tier upgrades, and overage growth before subtracting downgrades and cancellations. GRR only subtracts, so its ceiling is a book where nobody downgraded and nobody left, which is exactly 100%.

Here's one customer book run both ways:

NRR = ($400,000 + $65,000 - $18,000 - $45,000) / $400,000 = 100.5% GRR = ($400,000 - $18,000 - $45,000) / $400,000 = 84.25%

That 16-point gap is why you report both. NRR says the book held its dollar value. Gross revenue retention says a sixth of it walked out and two expanding accounts covered the hole. A board seeing only the NRR figure never learns the business is one large renewal away from shrinking.

Which cohort and time window should you measure NRR over?

Fix the cohort at the start of the window and never let a customer join it mid-flight. That's where most reported NRR numbers go wrong, because numerator and denominator have to describe the same accounts.

The choices that change the answer:

• Cohort membership. Only customers already paying at the window's start count. Sign one new customer at $75,000 inside the year in the book above, leave them in the numerator, and the calculation returns 119.25% instead of 100.5%.

• Window length. Trailing twelve months smooths a single bad quarter. A quarter annualized amplifies it. Quarterly reporters usually compare a quarter against the same quarter a year earlier, which is how Twilio frames its metric.

• Reactivations. A customer who churned in month three and returned in month ten is a judgment call. Pick reactivation or expansion and apply it every period.

• Recurring revenue only. Services, setup fees, and hardware pass-through stay out, and a fixed FX rate stops currency swings from masquerading as expansion.

Publish the definition next to the number. An NRR figure without its cohort rule and window stated isn't comparable to anyone else's, including your own from last year. Contracted ARR has the same problem.

What does usage-based revenue do to NRR that seat-based revenue doesn't?

Usage-based revenue moves NRR without anyone selling or cancelling anything, because consumption changes on the customer's side of the contract. Seat-based expansion needs a purchase order and contraction needs a downgrade request, so every movement has a commercial event behind it. A customer who doubles their API traffic produces expansion revenue while your account team does nothing.

How that plays out:

• Inflation during customer growth. Snowflake's 125% net revenue retention for fiscal 2026 reflects a customer base whose data volumes grew, not 125% worth of upsell campaigns.

• Deflation without churn. A quiet workload decline lands in the contraction bucket even though the customer is retained, satisfied, and still under contract.

• Higher volatility. Consumption books swing period to period in a way seat books don't, which makes one quarter's NRR a weak signal.

• Missing attribution. Without tagging each movement as commercially driven or consumption driven, you can't tell whether NRR measures your pricing power or your customers' growth rate.

Split expansion into those two buckets before the number reaches a board deck, at the point where revenue churn and expansion get classified.

Related terms

Five neighbours worth reading next, roughly in the order they show up in a revenue review:

• Gross revenue retention strips expansion out and shows what the book keeps on its own.

• MRR movements breaks a period into the new, expansion, contraction, and churn components NRR compresses into one ratio.

• Revenue churn measures the lost-dollars side that sits in the NRR numerator as a subtraction.

• Annual contract value sets the per-customer figures the cohort math runs on.

• Revenue leakage explains why reported retention and billed retention drift apart.

FAQ

Is net revenue retention the same as net dollar retention?

Yes, NRR and net dollar retention describe the same calculation. Companies also publish it as dollar-based net expansion rate or dollar-based net retention rate. Twilio reports 116% for Q2 2026 against 108% for Q2 2025. Check each company's stated definition before comparing figures.

Does NRR include revenue from new customers?

No. New logos acquired inside the measurement window stay out of both the numerator and the denominator. Including them inflates the result without any retention having occurred, which is the most common NRR reporting error.

Can NRR stay above 100% while customer count falls?

Yes, and it happens constantly. A few expanding accounts can outweigh several departures in dollar terms, so a book can lose a third of its logos and still report expansion. Customer counts and GRR expose that pattern.

How often should finance teams recalculate NRR?

Monthly, on a trailing twelve-month basis. Monthly recalculation catches contraction early, and the trailing window keeps one heavy usage month or one large renewal from distorting the trend. Report the same window every period.