What NRR measures and why investors weight it so heavily
Net revenue retention isolates one question: if you sold nothing to anyone new, would your revenue go up or down? Take the customers you had at the start of a period, follow only those customers, and compare what they pay at the end with what they paid at the beginning. Above 100% means the base grows on its own. Below means new sales have to refill the bucket before any growth appears at the top line.
That is a stronger statement than any customer-count metric can make, because it weights each account by what it pays. A business can lose 10% of its customers and grow revenue, if the leavers were small and the survivors upgraded. It can also keep 95% of its customers and shrink, if the 5% who left were its largest. Counting customers cannot distinguish those situations; counting dollars can, which is why customer retention rate and this metric belong on the same dashboard rather than in competition.
The compounding is what makes it so consequential. A cohort at 120% annual net retention is worth 1.23 = 1.728 times its original revenue after three years without a single new customer; a cohort at 85% is worth 0.853 = 0.614 times. The same acquisition effort produces radically different businesses depending on which side of 100% the base sits, and the gap widens every year. This is the arithmetic behind the attention the metric receives in investor diligence.
One caution on definitions before you compute. Companies differ on whether NRR is measured on a trailing-twelve-month cohort or a point-in-time snapshot, whether it includes usage-based overage, and whether it is quoted monthly or annually. The US Securities and Exchange Commission's guidance on key performance indicators asks registrants to define such metrics clearly and to apply them consistently between periods, which is good practice even if you never file anything. Write your definition down and keep it fixed.
Net against gross, and the four moving parts
Expansion is additional revenue from customers already in the cohort — more seats, a higher tier, an added module, usage above a committed level. It is the only term that can push the ratio above 100%.
Contraction is revenue lost from customers who stayed: seats removed, a downgrade to a cheaper plan, a discount granted at renewal. It is often under-measured because it does not generate a cancellation event, and it is the term most likely to be missing from a hand-built calculation.
Churn is revenue lost from customers who left entirely. Note that a customer who reduces their spend to zero without formally cancelling belongs here, not in contraction, if they have effectively gone.
The distinction between net and gross is simply whether expansion is included. Gross revenue retention takes the starting base, removes contraction and churn, and stops. Because both subtracted terms are non-negative, GRR can never exceed 100% and is never higher than NRR — that relation holds for any inputs, which is what makes the gap between them a clean measure of the expansion contribution.
Both figures are worth reporting because they answer different questions. GRR tells you how leaky the bucket is: it is the ceiling on what you keep without selling anything more to anyone. NRR tells you whether the upsell motion is filling the leak faster than it drains. A company at 120% NRR and 78% GRR is losing more than a fifth of its base every year and covering it with aggressive expansion inside the accounts that stay — a real and fragile position, since expansion is usually concentrated in a handful of accounts while churn is spread across many.
Worked example: a $100,000 annual cohort
A cohort of customers paid you $100,000 a year at the start of last year. During the year they bought $20,000 of upgrades, downgraded $5,000, and $10,000 of them cancelled outright.
- Ending cohort revenue. $100,000 + $20,000 − $5,000 − $10,000 = $105,000.
- Net revenue retention. $105,000 ÷ $100,000 × 100 = 105.00%.
- Gross revenue retention. ($100,000 − $5,000 − $10,000) ÷ $100,000 × 100 = $85,000 ÷ $100,000 = 85.00%.
- Expansion contribution. 105% − 85% = 20 points, which is the $20,000 of expansion divided by the $100,000 base, as expected.
- Three-year projection at this net rate. $100,000 × 1.053 = $100,000 × 1.157625 = $115,762.50.
- Three-year projection without expansion. $100,000 × 0.853 = $100,000 × 0.614125 = $61,412.50.
Steps 5 and 6 are the ones to present together. Expansion is worth $115,762.50 − $61,412.50 = $54,350 of cohort revenue over three years on a $100,000 base, which is more than half the original base. It also shows the fragility: the entire difference between a growing cohort and one that has lost 39% of its value depends on the upsell motion continuing to work. If expansion halved to $10,000, NRR would fall to 95%, and the three-year value would be $100,000 × 0.953 = $85,737.50 — below where the cohort started.
Reading the two numbers together
Always read NRR and GRR as a pair, because the same NRR can describe opposite businesses. 110% built on 105% gross retention is a stable base with modest upsell. 110% built on 80% gross retention is a leaky base rescued by concentrated expansion, and it will collapse the moment the expansion motion stalls or the few large accounts driving it reach their ceiling.
Interpret the level relative to your segment. Businesses selling seat-based software into growing companies have a structural expansion tailwind — their customers hire, so the account grows without any sales effort. Businesses selling a fixed-scope product to small companies have no such tailwind and will sit lower on both measures for reasons that have nothing to do with product quality. Comparing your NRR against a figure quoted by a company with a different pricing model is not informative.
Watch the period convention closely. A monthly NRR of 101% and an annual NRR of 101% are wildly different achievements: the monthly figure compounds to 1.0112 = 112.7% a year. When a number is quoted without its period, assume annual, and confirm before you use it in any comparison.
Finally, treat the projection as an illustration rather than a forecast. It applies the current period's rate unchanged, and real cohorts do not behave that way — churn typically falls as a cohort ages while expansion often slows once easy upsells are exhausted. The projection is most useful for making the compounding vivid and for comparing scenarios against each other, not for predicting a specific number three years out. For an operational forecast, model each cohort separately with the subscription revenue forecast calculator.
What an annual retention rate is worth over time
| Annual NRR | After 1 year | After 3 years | After 5 years | After 10 years |
|---|---|---|---|---|
| 80% | $0.800 | $0.512 | $0.328 | $0.107 |
| 90% | $0.900 | $0.729 | $0.590 | $0.349 |
| 100% | $1.000 | $1.000 | $1.000 | $1.000 |
| 110% | $1.100 | $1.331 | $1.611 | $2.594 |
| 120% | $1.200 | $1.728 | $2.488 | $6.192 |
| 130% | $1.300 | $2.197 | $3.713 | $13.786 |
The ten-year column is arithmetic, not a forecast — no cohort sustains 130% for a decade, because accounts eventually saturate. Its purpose is to show why a difference of twenty points is not a twenty percent difference in outcome.
Definition errors that inflate the number
- Including new customers. The most damaging error. Adding new-business revenue turns NRR into a growth rate, and it will read comfortably above 100% while the existing base collapses.
- Omitting contraction. Downgrades produce no cancellation event, so they are easy to miss in a hand-built calculation. Their absence raises both NRR and GRR.
- Mixing periods. A monthly rate compounds twelve times a year. Quoting a monthly figure as though it were annual overstates performance dramatically.
- Counting one-off revenue. Professional services, setup fees and one-time overages are not recurring and do not belong in either term.
- Changing the cohort definition between periods. A trailing-twelve-month cohort and a point-in-time snapshot give different answers; switching between them creates a trend that is purely methodological.
- Reporting NRR without GRR. Net retention alone cannot distinguish a stable base from a leaky one rescued by a few large upsells.
Where NRR fits with the rest of the metrics
NRR is the revenue-weighted counterpart to customer churn and customer retention rate. The count metrics tell you how many relationships you keep; NRR tells you how many dollars. Both matter, and a wide divergence between them is itself diagnostic: customer retention far above revenue retention means your larger accounts are leaving, while the reverse means you are losing small accounts and growing large ones — usually a healthier pattern, though it concentrates your risk.
It also modifies how you should read every lifetime-value figure in the business. The standard construction of lifetime value assumes revenue per account is constant and only churn erodes the cohort. With net retention above 100%, that assumption understates the truth: the surviving accounts pay more each year, so lifetime value is higher and effective CAC payback is shorter than a constant-revenue model implies. If your NRR is comfortably above 100%, say so explicitly when you present an LTV:CAC ratio, because the standard formula does not capture it.
For a growth-and-efficiency summary, NRR pairs naturally with the Rule of 40: one describes the durability of existing revenue, the other the balance between growth and profitability. And when you need the underlying revenue figures themselves, build them in the MRR calculator first — every term in this calculation depends on a consistent, well-defined recurring revenue base, and errors there propagate into every retention figure downstream.