Gross vs Net Retention: How to Calculate GRR and NRR

Gross and net retention measure the same thing, the recurring revenue you keep from customers you already had, with one difference: whether upgrades count. Gross revenue retention (GRR) ignores them. Net revenue retention (NRR) adds them. This guide gives both formulas, shows what goes in each term with worked numbers, compares the period and cohort methods, and covers the cases most guides skip: pauses, skips, delivery intervals and one-off upsells in subscription ecommerce.

Gross vs net retention at a glance

Chargebee's definitions put the difference in one line each: GRR is "the percentage of recurring revenue retained from existing customers, excluding expansion revenue"; NRR is the same percentage "including expansion revenue and accounting for churn."4Chargebee Docs, Recurring revenue (MRR movement summary)

The GRR and NRR formulas

Both formulas run on monthly recurring revenue (MRR), or on annual recurring revenue for annual contracts. Pick a period, usually a month, and take only the customers who were paying on its first day.

GRR = (starting MRR − churned MRR − contraction MRR) ÷ starting MRR × 100

NRR = (starting MRR + expansion MRR − churned MRR − contraction MRR) ÷ starting MRR × 100

The four inputs, with Stripe Billing's names, since its MRR movements are defined in its documentation:1Stripe Docs, Billing analytics: metric definitions (MRR, MRR growth, churned revenue, retention by cohort, discounts)

Stripe's own "churned revenue" metric is churned MRR plus contraction MRR, which is exactly what GRR subtracts.1Stripe Docs, Billing analytics: metric definitions (MRR, MRR growth, churned revenue, retention by cohort, discounts) Stripe also counts a subscription as churn only when it is canceled or marked unpaid; while a renewal is past_due, it stays in MRR.1Stripe Docs, Billing analytics: metric definitions (MRR, MRR growth, churned revenue, retention by cohort, discounts) How churn is counted, and why the denominator matters, is covered in detail in how to calculate churn rate.

Total MRR on October 31 is $44,300. Divide it by $40,000 and you get 110.75%, a number that looks like great retention and is mostly acquisition. That is the most common way NRR gets inflated.

The gap between NRR and GRR is always expansion MRR divided by starting MRR: here $1,800 ÷ $40,000 = 4.5 points. That makes the pair easy to read. GRR is your loss rate, the gap is your expansion rate, and NRR is the sum.

What goes in each: expansion, contraction, churn, reactivation

Most disagreements about retention come from events that don't fit neatly in one bucket. Here is how each one counts, using the rule that the base is the set of customers paying on day one.

Reactivation is the one that trips people up. Chargebee's NRR includes only MRR from customers active at the start of the period and excludes customers who activated or joined during it.3Chargebee Docs, What is NRR (Net Revenue Retention) and where to check this metric in RevenueStory A subscriber who canceled in August and returned in October wasn't in the October base, so their return can't raise October retention. Stripe and Chargebee both report reactivation MRR as its own movement, next to new MRR.1Stripe Docs, Billing analytics: metric definitions (MRR, MRR growth, churned revenue, retention by cohort, discounts)4Chargebee Docs, Recurring revenue (MRR movement summary) The churn guide applies the same rule to customer counts: a return is a reactivation, not a correction of an earlier month.

Here is why that matters, with four subscribers:

If you netted all changes at the total level instead, C's $20 upgrade would hide most of B's $10 downgrade and GRR would look better than it is. The customer-level method needs one row per customer with start and end MRR, which most billing tools can export. Stripe's prebuilt reports include MRR per subscriber per month and a log of every customer MRR change.1Stripe Docs, Billing analytics: metric definitions (MRR, MRR growth, churned revenue, retention by cohort, discounts)

Period method vs cohort method

There are two ways to choose the customers you follow, and they answer different questions.

The period method gives one number per period for the whole base. It's what Chargebee's retention metrics compute: MRR from subscriptions active in the previous period, compared with what they paid then.2Chargebee Docs, Retention (Gross and Net Revenue Retention Rate) It mixes subscribers in their first month with subscribers in their third year, so a change in your acquisition mix moves it even if nothing changed for existing customers.

The cohort method follows one starting group over time. Stripe assigns subscribers to cohorts by the month they first generated positive MRR, and its revenue retention table shows the MRR remaining from that cohort at the end of each following month. Stripe notes it can exceed 100% because it counts upgrades, downgrades and cancellations, so it is a net figure.1Stripe Docs, Billing analytics: metric definitions (MRR, MRR growth, churned revenue, retention by cohort, discounts) Recurly's subscriber retention dashboard is also built by cohort of sign-up month, but it counts subscribers, not revenue.5Recurly Docs, Subscriber retention

A cohort table with both lines shows the shape a single rate hides:

Use the period method for a monthly dashboard and the cohort method to see where the losses happen. If a newer cohort falls faster in its first three months than older ones did, look at what changed for it: the offer, the channel, the first-box experience or the first renewal. How to build cohort tables is covered in what is cohort analysis.

Monthly vs annual retention

Retention compounds like churn does. Measured on the same cohort, twelve months at 98% GRR leave you with about 78.5% of starting revenue, not 76%. Twelve months at 99.5% NRR give about 94.2%.

Compounding only holds when every month is measured on the same cohort. A period-method rate resets its base every month and includes customers who joined since, so twelve of them chained together don't equal the annual retention of the customers you had in January. For an annual figure, measure it directly: revenue this month from customers who were paying twelve months ago, divided by what they paid then. That is how public companies do it, as the next section shows.

How public companies define NRR and GRR

Public software companies report NRR in their annual filings, and each one writes its own definition. Two examples from SEC filings show how far apart they can be:

Sources: Instructure 10-K6Instructure Holdings, Form 10-K for fiscal year 2022 (SEC), Key business metrics; Snowflake 10-K.7Snowflake Inc., Form 10-K for fiscal year ended January 31, 2022 (SEC), Net revenue retention rate

Two things stand out. Snowflake measures recognized revenue from usage over two years, so its NRR captures consumption growth that an MRR-based metric wouldn't see. Instructure's NRR counts "changes in quantities of users, changes in pricing, additional applications purchased or applications no longer used," and it publishes GRR next to it.6Instructure Holdings, Form 10-K for fiscal year 2022 (SEC), Key business metrics These figures describe enterprise software sold on contracts. They are examples of disclosure, not targets for a subscription brand.

That sentence is the safeguard to copy: a customer who left stays in the denominator. Drop churned customers from the base and NRR goes up for the wrong reason.

Inside Tagada

Build the funnel once, route every customer to the right offer

Checkout, upsells and subscriptions on one flow, with conditions per country, order or customer.

Start free

GRR and NRR in subscription ecommerce vs B2B SaaS

The formulas come from software, where expansion means more seats or a higher tier. A subscription brand shipping products every month has other levers, and other traps.

The rules that keep subscription ecommerce numbers honest:

Two more points specific to brands that bill cards every month. First, a renewal that fails isn't lost yet: in Stripe it stays in MRR while past_due, and drops out as churn when the subscription is canceled or marked unpaid.1Stripe Docs, Billing analytics: metric definitions (MRR, MRR growth, churned revenue, retention by cohort, discounts) Your most recent month will look better than it will once retries finish, so compare closed periods. Why renewals fail and how retries work are covered in subscription payment failures and what is dunning. Second, discounts change MRR. Stripe lets you choose whether recurring and one-time discounts are subtracted from MRR, and always subtracts permanent recurring discounts.1Stripe Docs, Billing analytics: metric definitions (MRR, MRR growth, churned revenue, retention by cohort, discounts) If a "save" offer gives 30% off to stop a cancellation, that's contraction in GRR. When the discount ends, the price going back up is expansion in NRR. Keep the setting fixed, or your retention moves when you change a toggle.

How to read GRR and NRR together

The pair says more than either number. Without a benchmark from your own history, read the direction and the gap.

Customer churn and revenue retention also tell different stories. If the customers who leave are cheap and the ones who stay upgrade, subscriber churn can be high while NRR is above 100%. Report subscriber retention, GRR and NRR together. Revenue retention also feeds lifetime value: more on that in what is CLTV.

Common mistakes

• Dividing total ending MRR by starting MRR. It includes new and returning customers. Use start minus losses for GRR, and the ending MRR of starting customers only for NRR.
• Dropping churned customers from the base. They have to stay in the denominator at zero, as in Snowflake's definition.7Snowflake Inc., Form 10-K for fiscal year ended January 31, 2022 (SEC), Net revenue retention rate
• Netting expansion against contraction across customers in GRR. Cap each customer at their starting MRR, as Chargebee does.2Chargebee Docs, Retention (Gross and Net Revenue Retention Rate)
• Counting reactivations as retention. A customer who wasn't paying on day one belongs with reactivation MRR.
• Mixing one-off revenue into MRR-based NRR. One-click upsells and single add-ons are real revenue, measured elsewhere.
• Chaining monthly period rates into an annual one. Measure the year on the cohort that was paying twelve months ago.
• Changing definitions silently. Discount settings, pause rules and FX treatment all move the number. Write the definition next to it, the way public companies do in their filings.
• Comparing with a published benchmark. Benchmarks come from someone else's customer base and definitions, mostly B2B software on annual contracts. Compare your own cohorts over time instead.

How to calculate GRR and NRR in a spreadsheet

One row per customer, one tab per period. This is the customer-level method from the table above.

Then: GRR =SUM(E:E)/SUMIF(D:D,TRUE,B:B) and NRR =SUM(F:F)/SUMIF(D:D,TRUE,B:B). Check that starting MRR minus G, minus H, plus I equals the sum of column F. If it doesn't, some customers changed status in a way you haven't classified yet, such as a pause or a currency change. Customers with B = 0 and C > 0 are new or reactivated: they stay out of both rates, and their total should match new plus reactivation MRR in your billing tool.

Where Tagada fits

Whatever tool computes your numbers, the method doesn't change: freeze the base on day one, subtract churn and contraction for GRR, add expansion for NRR, keep new and returning customers out, cap GRR customer by customer, and write down how pauses, skips, intervals and discounts count.