NRR and GRR

How do you calculate NRR and GRR?#

1. Freeze the cohort#

Record the customer list and each customer's recurring revenue at the start date. Nothing enters this list afterwards. Nothing leaves it either — a customer who churns stays in the denominator at their starting value forever.

2. Fix the recurring-revenue basis#

State whether you are using ARR, MRR, annualised current-quarter subscription revenue, or committed minimums, and apply the same basis to both ends of the window. Handle usage, currency, services and delinquency consistently at both dates.

3. Classify each movement once#

4. Compute both, from the same denominator#

gross retained = starting − churn − contraction
net retained   = gross retained + expansion
GRR = gross retained / starting
NRR = net retained  / starting

5. Add logo retention and concentration#

Logo retention (surviving customers ÷ starting customers) shows breadth where revenue retention shows weight. Publish the top-customer and top-ten share of the cohort alongside both.

6. Segment, then age#

Cut by customer size, product, channel, geography and tenure. A twelve-month window on a young segment contains only early adopters and should not be generalised — see Cohort Analysis.

Worked example#

A company starts the year with 100 customers and $100,000 MRR. Over the year: 10 customers churn, removing $8,000; surviving customers contract by $5,000; surviving customers expand by $20,000; and new customers add $30,000.

gross retained = $100,000 − $8,000 − $5,000 = $87,000
net retained   = $87,000 + $20,000          = $107,000

GRR = $87,000  ÷ $100,000 = 87.0%
NRR = $107,000 ÷ $100,000 = 107.0%
logo retention = (100 − 10) ÷ 100 = 90.0%

Ending total MRR is $107,000 + $30,000 = $137,000, so total MRR grew 37%. Three true statements, three different numbers: the company grew 37%, the installed base grew 7%, and it kept 87% of starting recurring revenue before expansion. Only the third one tells you about the product.

The denominator trap, in numbers. Compute the same NRR on survivors only — starting revenue of $92,000 for the 90 customers still present:

survivor-only "NRR" = $107,000 ÷ $92,000 = 116.3%
survivor-only "GRR" = $87,000  ÷ $92,000 =  94.6%

Both are wrong by construction, and both are wrong in the flattering direction: +9.3 points on NRR and +7.6 points on GRR. Any retention figure quoted without its denominator should be assumed to be one of these until proven otherwise.

The concentration diagnostic. Suppose $18,000 of the $20,000 expansion came from a single enterprise account. Strip that one event out:

diagnostic NRR = ($87,000 + $2,000) ÷ $100,000 = 89.0%

Reported NRR remains 107.0% and should stay 107.0% — the expansion is real revenue from a real customer. But the broad installed base is running at 89%, and the 18-point gap is the concentration risk. Publish the diagnostic beside the reported figure; do not substitute it.

The caveat that matters: all of this is a single twelve-month window on one cohort. One window is a data point. Retention is a trend, and the trend is what changes decisions.

What are the common mistakes?#

• Excluding churned logos from the denominator. The single largest source of overstated NRR. Worth 9.3 points in the example above.
• Letting new customers into the cohort. That converts retention into a growth rate with a retention label.
• Reporting NRR without GRR. Expansion pays for churn silently until it stops.
• Ignoring concentration. One enterprise expansion can carry a dollar-weighted metric across a whole base.
• Benchmarking across companies without reading definitions. Cohort scope, revenue basis, customer thresholds and renewal grace periods all differ between issuers.

When do NRR and GRR break?#

In transactional and seasonal businesses. Where customer spending naturally varies, a twelve-month window mistakes demand cycles for churn. Use matched seasons, longer windows and purchase cohorts instead.

In usage-based models. Consumption moves with the customer's own business. Split contracted-minimum retention from variable-usage retention so a macro slowdown is not booked as a product failure. See Usage-Based Pricing.

Through corporate events. Mergers between customers, product migrations, acquired revenue and currency moves all need explicit bridge rules that preserve economic continuity without quietly rewriting the starting base.

When high NRR is doing the wrong work. Expansion driven by contractual price escalators or minimum commitments is not the same asset as expansion driven by adoption. Pair retention with product usage, gross margin, logo retention and customer outcomes.

When the base is too small. On 40 customers, a single churn event moves GRR by multiple points. Report the count alongside the percentage, and resist decimal places the sample cannot support.