Retention

NRR vs. GRR: two retention numbers that tell different stories

By Assaf Schwartz · 10 September 2026 · 9 min read

Gross revenue retention answers one question — how much of last year's revenue is still here? Net revenue retention answers a different one, and answers it with expansion revenue mixed in. Reported alone, the second can stay comfortably above 100% for years while the thing it is measuring quietly empties out.

Two numbers, two questions

Both are measured on a fixed set of customers — the ones you had at the start of the period — and both exclude anything you sold to a new logo. That common ground is what makes them comparable and what makes the gap between them meaningful.

Gross revenue retention is (starting revenue − churn − downgrades) ÷ starting revenue. It counts only losses, so it is capped at 100% by construction. It is a statement about whether the product holds its customers.

Net revenue retention is (starting revenue − churn − downgrades + expansion) ÷ starting revenue. It adds upsells, seat growth and usage increases back in, so it has no ceiling. It is a statement about whether the installed base is a growth engine on its own.

The useful reading is always the pair. A business at 90% GRR and 120% NRR keeps almost all of what it has and grows the rest — durable. A business at 60% GRR and 120% NRR is expanding rapidly inside a small group of accounts while the majority walks out. The headline number is the same. The companies are not, and the second one is far more expensive to run, because every point of gross loss has to be re-bought at full acquisition cost.

What the model reports

The simulator does not model individual customers or cohorts. It moves recurring revenue with two proportional monthly rates: Organic MRR growth adds a share of the existing base each month, Monthly revenue churn removes one. Both are applied to the revenue already on the books, so the reported figure is exactly NRR = 1 + organic growth − churn, per month.

On the default inputs that is 101.2% a month — 4.0% added against 2.8% lost — which compounds to 115.4% over twelve months. Gross retention is not reported directly, but it falls straight out of the same input: annual GRR = (1 − monthly churn)¹², or 71.1%. The gap between those two, 44.3 points of the starting base, is what the growth control is standing in for.

Be clear about what that stand-in covers. Because the growth rate multiplies the existing base, it behaves mathematically like expansion revenue: bigger base, more growth. Paid acquisition is handled separately in the engine — new revenue bought with CAC out of re-invested profit — and is correctly excluded from the retention figure. So the model's NRR is a net retention number in the strict sense, not a growth rate with new logos folded in.

Monthly rates, annual retention

Retention is quoted annually and modelled monthly, and the conversion trips people up in the same direction every time. Multiplying a monthly churn rate by twelve overstates the annual loss, because each month's churn applies to a base that the previous month already reduced.

The error grows with the rate. At 1.0% monthly, multiplying by twelve overstates the annual loss by 0.6 points — survivable. At 8.5% it overstates it by 36.4 points and produces a figure above 100%, which is nonsense: you cannot lose 102% of a base that only ever contained 100%. The compounded answer is 65.6%, and half the revenue base is gone in 8 months.

Net retention compounds the same way, in the other direction. annual NRR = (1 + growth − churn)¹², so a monthly net rate barely above parity turns into a substantial annual figure: the default company's 101.2% monthly becomes 115.4% annual. This is the arithmetic that makes best-in-class NRR possible without heroic upselling — a business that expands one and a half points more than it loses each month reports 119.6% a year.

How NRR hides a hollowing base

Here is the case that matters. Four versions of the default company, each with a different mix of expansion and churn, all landing on the same 101.2% monthly net retention. Run each of them through the full sixty months.

The revenue path is identical to the dollar. All four end at $3.4M of ARR with $1.3M in the bank, because the projection only ever sees the net rate. What changes is everything underneath it: annual gross retention falls from 94.2% to 47.6%, lifetime value from $65,520 to $5,460, and the LTV:CAC verdict from 18.20x to 1.52x.

Think of it as a treadmill. In the first row, 29% of gross additions go to replacing revenue that left; in the last, 83% do. The fourth company has to generate 4.2 times the gross new revenue every month to stand in the same place, and it is one soft quarter of expansion away from shrinking. Nothing in the net retention figure tells you which of the four you are looking at.

Benchmarks by segment

Retention benchmarks are only meaningful within a segment. An SMB self-serve product and an enterprise platform are not measured against the same bar, and the difference between them is structural rather than a matter of execution quality.

Two rules of thumb worth holding. GRR under 80% in an enterprise business is a product or fit problem, not a customer success staffing problem. And NRR above 130% almost always means usage-based pricing rather than exceptional account management — which is a pricing decision available to you, not an execution gap.

The saved scenarios illustrate the spread. The seed-stage scenario runs 169.6% annual NRR on 57.5% gross retention — a strong headline over a weak base, exactly the SMB pattern. The churn crisis scenario runs 83.4% NRR against 34.4% GRR: expansion is real and it is nowhere near enough.

Setting the two controls

To make the model represent a retention profile you have actually measured, work backwards from the annual pair. Churn comes from GRR alone: monthly churn = 1 − GRR^(1/12). Growth then has to cover that churn and deliver the net figure on top: monthly growth = NRR^(1/12) − 1 + monthly churn.

The last row is the instructive one. A business at 80% GRR and 130% NRR — the profile a lot of usage-based companies report — needs 4.05% monthly expansion to carry 1.84% monthly churn. Compare it to the 95% GRR row, which lands 10 points below it on less than half the monthly expansion, because it barely loses anything. Retention is cheaper than expansion, every time.

One caution when you do this: if your organic growth genuinely comes from new logos rather than existing accounts, splitting it between the growth control and paid acquisition matters for the cash path. Revenue routed through Organic MRR growth arrives free; revenue routed through CAC is paid for out of profit. Putting new-logo growth in the wrong control flatters both the retention figure and the burn.

What to track, and how often

• Report GRR and NRR together, always, revenue-weighted. Either one alone is a half-answer. The gap between them is the expansion contribution, and it is worth naming as its own line so a fall in expansion cannot be papered over by a fall in churn.
• Add logo retention as a third line. Revenue-weighted retention is dominated by large accounts. If NRR is healthy and logo retention is falling, you are consolidating into a handful of customers, which is a concentration risk long before it is a revenue problem.
• Measure annually, on cohorts, not monthly on a rate. Monthly retention is too noisy to manage and annual contracts make it meaningless anyway. Take the accounts you held twelve months ago and compare their revenue today.
• Split expansion from contraction. Net expansion of ten points can be twelve points of upsell against two of downgrade, or thirty against twenty. The second business has a serious problem that its NRR is concealing.
• Segment it. Blended retention across segments hides the one that is failing. Cut it by acquisition channel, plan and cohort size, and the churn almost always turns out to be concentrated somewhere specific.

For what compounding churn does to the size of company you can build, see churn, compounded. For the metric that prices the cash cost of replacing everything you lose, see burn multiple — retention improvements land in its denominator, which is why they are the cheapest efficiency work available.

Educational content, not financial advice. Figures here are illustrative and exclude taxes, financing and one-off items. Model your own numbers in the simulator and check them with a qualified accountant before acting.