TL;DR
MRR = the sum of every active subscription's recurring charge, normalized to one month (annual plans divided by 12), excluding one-time fees and taxes. MRR moves through new, expansion, reactivation, contraction and churn. NRR = (starting MRR + expansion - contraction - churn) / starting MRR, and it leaves new customers out. Both are easy to compute and easy to compute differently. Define them once on your warehouse subscription tables so finance, product and the board read the same number.
Monthly recurring revenue (MRR) is the sum of the recurring charges of all active subscriptions, normalized to one month. Net revenue retention (NRR) is how much of the recurring revenue you had at the start of a period you still have from those same customers at the end, after upgrades, downgrades and cancellations. MRR tells you the size of the recurring business; NRR tells you whether your existing customers grow it or shrink it.
- MRR = sum of (recurring charge / billing interval in months) over all active subscriptions
- NRR = (starting MRR + expansion - contraction - churn) / starting MRR
This guide covers how to calculate both with worked examples, how MRR moves month to month, what NRR benchmarks can and cannot tell you, and how to compute both from the subscription tables in your data warehouse.
What MRR is, and how it relates to ARR?
MRR is the predictable, repeating part of a subscription business's revenue, expressed per month. It counts what active customers are contracted to pay each month and leaves out anything that does not repeat. It is a run rate taken at a point in time, usually the first or last day of the month, not the cash collected during the month.
Annual recurring revenue (ARR) is the same run rate expressed per year: ARR = MRR × 12. Companies with mostly annual contracts tend to report ARR, and companies with monthly plans tend to report MRR. The rules below apply to both.
How to calculate MRR?
The formula is short. The work is in the rules for what counts. Decide each of these once and write it down:
- Normalize billing periods. Divide annual plans by 12 and quarterly plans by 3. A $2,400 annual plan is $200 of MRR, not $2,400 in the month it was paid.
- Apply recurring discounts. A customer on a permanent 20% discount contributes the discounted amount. Decide whether a first-month coupon reduces MRR for that month or is ignored.
- Exclude one-time charges. Setup fees, onboarding services, hardware and one-off add-ons are revenue but not recurring revenue.
- Exclude taxes. VAT and sales tax are passed through, not earned.
- Decide on usage-based charges. Many teams count only committed minimums as MRR and report overage separately, because usage can swing month to month.
- Exclude free trials until the first paid period starts.
- Time churn consistently. Count a cancellation when the paid period ends, not when the customer clicks cancel.
- Fix a currency rule. Convert foreign-currency plans at a fixed or a monthly rate, and know that exchange rates alone will move MRR.
Worked example. A company has 15 customers on a $120 monthly plan and 5 customers on a $1,200 annual plan. One monthly customer has a permanent $10 discount, and a new customer paid a one-time $500 setup fee.
| Item | Calculation | MRR |
|---|---|---|
| Monthly plans | 15 × $120 | $1,800 |
| Annual plans | 5 × $1,200 / 12 | $500 |
| Recurring discount | 1 × -$10 | -$10 |
| Setup fee | One-time, excluded | $0 |
| Total MRR | $2,290 |
MRR movements: new, expansion, contraction and churn
A single MRR figure hides what happened inside the month. Breaking the change into movements shows whether growth came from new customers or from existing ones, and where revenue leaked:
| Movement | What it is | Example |
|---|---|---|
| New MRR | MRR from customers paying for the first time | A customer starts a $300/month plan |
| Expansion MRR | Increase from existing customers: upgrades, more seats, add-ons | A customer moves from $30 to $50 a month: +$20 |
| Reactivation MRR | MRR from customers who had churned and came back | A customer who left in January returns in May |
| Contraction MRR | Decrease from customers who stay: downgrades, fewer seats | A customer drops from 10 seats to 8 |
| Churned MRR | MRR lost from customers who cancel entirely | Two customers paying $40 each cancel: -$80 |
The movements must reconcile with the change in MRR: ending MRR = starting MRR + new + expansion + reactivation - contraction - churn. Net new MRR is everything after starting MRR. If your movements do not add up to the change in MRR, some change is not being classified, which usually means a plan change or a refund rule is missing from the definition.
Worked example. On March 1, 200 customers pay $50,000 of MRR. During March:
| MRR | |
|---|---|
| Starting MRR (March 1) | $50,000 |
| + New (30 new customers) | +$6,000 |
| + Expansion (upgrades and added seats) | +$5,000 |
| - Contraction (downgrades) | -$1,500 |
| - Churn (10 customers cancel) | -$2,500 |
| Ending MRR (April 1) | $57,000 |
Net new MRR is $7,000, so MRR grew 14% in the month ($7,000 / $50,000). Most of that came from new customers, but existing customers added $1,000 net ($5,000 - $1,500 - $2,500), which is what NRR measures.
Net revenue retention (NRR)
NRR, also called net dollar retention (NDR), measures revenue from a fixed group of customers: those who were paying at the start of the period. New customers are left out by definition, so NRR shows whether the existing base grows or shrinks on its own.
- NRR = (starting MRR + expansion - contraction - churn) / starting MRR
- GRR = (starting MRR - contraction - churn) / starting MRR
Gross revenue retention (GRR) is the same calculation without expansion, so it can never be above 100%. It shows how much revenue you keep before upsells hide the losses. Read the two together: a high NRR with a low GRR means strong expansion is covering a churn problem.
Monthly example. Using the March numbers above: NRR = ($50,000 + $5,000 - $1,500 - $2,500) / $50,000 = $51,000 / $50,000 = 102%. GRR = ($50,000 - $1,500 - $2,500) / $50,000 = $46,000 / $50,000 = 92%. The $6,000 of new MRR plays no part in either.
Annual example. Customers who were paying on October 1, 2025 had $100,000 of MRR. By October 1, 2026, those same customers expanded by $20,000, contracted by $3,000 and churned $5,000, so they now pay $112,000. Annual NRR = $112,000 / $100,000 = 112%, and annual GRR = ($100,000 - $3,000 - $5,000) / $100,000 = 92%.
Annual NRR is the figure most often reported, and it should be measured directly on the 12-month cohort. Do not derive it from a monthly figure: compounding 102% for 12 months gives about 127%, but each month's NRR is measured on a different set of customers, so the result does not describe any real cohort.
NRR benchmarks and their caveats
100% is the line that matters most: above it, existing customers grow revenue even if you sign no one new; below it, you have to win new customers just to stand still. Published benchmarks go further and quote typical and top-quartile figures, but they are hard to compare with your own number. Before you do, check:
- The definition. Whether reactivated customers count as retained, whether usage and overage revenue are included, and whether the figure is revenue-based or customer-count-based.
- The period and the cohort. Annual cohort NRR, trailing-twelve-month averages and annualized monthly figures are different numbers.
- The segment. Companies selling to large enterprises with seat or usage expansion typically report higher NRR than those selling monthly plans to small businesses.
- Price changes. A list price increase raises NRR without any change in how customers use the product.
- Cohort size. With a few dozen customers, one large upgrade or cancellation moves NRR by several points.
- Who is reporting. Surveys of public or venture-backed companies over-represent the companies that grew.
Your own trend, measured the same way every month, is usually more useful than a benchmark. So is NRR broken down by plan, segment or acquisition channel, which shows where expansion and churn actually come from.
Computing MRR and NRR on warehouse subscription tables
Subscription data from Stripe, Chargebee, Recurly, Paddle or an in-house billing system usually lands in the warehouse through an ELT connector, then gets modeled in dbt. The sketch below assumes a modeled subscriptions table with one row per subscription version: a plan change closes the old row and opens a new one. Columns: customer_id, recurring_amount and recurring_discount per billing period (decimal, before tax), billing_interval_months (1, 3 or 12), started_at and ended_at (null while active). It also uses a date spine, calendar_months, with one row per month.
-- Postgres-style SQL; adjust date functions for your warehouse.
WITH months AS (
-- one row per month, starting the month before the first subscription
SELECT
month_start,
LAG(month_start) OVER (ORDER BY month_start) AS prev_month_start
FROM calendar_months
),
customer_mrr AS (
-- each customer's MRR on the first day of each month
SELECT
m.month_start,
s.customer_id,
SUM((s.recurring_amount - s.recurring_discount) * 1.0
/ s.billing_interval_months) AS mrr
FROM months AS m
JOIN subscriptions AS s
ON s.started_at <= m.month_start
AND (s.ended_at IS NULL OR s.ended_at > m.month_start)
GROUP BY m.month_start, s.customer_id
),
customers AS (
SELECT customer_id, MIN(month_start) AS first_month
FROM customer_mrr
GROUP BY customer_id
),
paired AS (
-- this month's MRR next to last month's, per customer
SELECT
m.month_start,
c.customer_id,
c.first_month,
COALESCE(prev.mrr, 0) AS prev_mrr,
COALESCE(cur.mrr, 0) AS cur_mrr
FROM months AS m
CROSS JOIN customers AS c
LEFT JOIN customer_mrr AS prev
ON prev.customer_id = c.customer_id
AND prev.month_start = m.prev_month_start
LEFT JOIN customer_mrr AS cur
ON cur.customer_id = c.customer_id
AND cur.month_start = m.month_start
WHERE m.prev_month_start IS NOT NULL
AND (prev.mrr IS NOT NULL OR cur.mrr IS NOT NULL)
),
movements AS (
SELECT
month_start,
SUM(prev_mrr) AS starting_mrr,
SUM(CASE WHEN prev_mrr = 0 AND month_start = first_month
THEN cur_mrr ELSE 0 END) AS new_mrr,
SUM(CASE WHEN prev_mrr = 0 AND month_start > first_month
THEN cur_mrr ELSE 0 END) AS reactivation_mrr,
SUM(CASE WHEN prev_mrr > 0 AND cur_mrr > prev_mrr
THEN cur_mrr - prev_mrr ELSE 0 END) AS expansion_mrr,
SUM(CASE WHEN prev_mrr > 0 AND cur_mrr > 0 AND cur_mrr < prev_mrr
THEN prev_mrr - cur_mrr ELSE 0 END) AS contraction_mrr,
SUM(CASE WHEN prev_mrr > 0 AND cur_mrr = 0
THEN prev_mrr ELSE 0 END) AS churned_mrr,
SUM(cur_mrr) AS ending_mrr
FROM paired
GROUP BY month_start
)
SELECT
month_start,
starting_mrr,
new_mrr,
reactivation_mrr,
expansion_mrr,
contraction_mrr,
churned_mrr,
ending_mrr,
(starting_mrr + expansion_mrr - contraction_mrr - churned_mrr)
/ NULLIF(starting_mrr, 0) AS monthly_nrr
FROM movements
ORDER BY month_start;Each row reconciles: ending_mrr equals starting_mrr plus new, reactivation and expansion, minus contraction and churn. New and reactivation MRR only come from customers with no MRR last month, so monthly_nrr leaves them out, as NRR should. For annual NRR, take the customers with MRR in customer_mrr on a start month, sum the MRR those same customers have 12 months later (zero for anyone who left), and divide by their starting MRR.
Three limits of the sketch: it snapshots MRR on the first day of each month, so a subscription that starts and ends within one month never appears; it assumes one currency; and the CROSS JOIN is fine for thousands of customers but should become an incremental dbt model at larger scale.
Why MRR and NRR should be defined once?
Every rule in this guide is a choice, and every place that computes MRR makes those choices again. Finance computes it in a spreadsheet from billing exports, a subscription analytics tool computes it from the billing system's API, the board deck uses a third version, and an AI assistant asked about MRR writes its own SQL. Each one treats discounts, annual plans, churn timing and reactivation slightly differently, and the numbers stop matching.
Defining MRR, its movements and NRR once, on the subscription tables in the warehouse, with the data team as owner, removes the copies. It also makes the more useful questions possible. MRR by acquisition channel, NRR by the onboarding path a customer took, or expansion by feature adoption all need revenue joined to product and marketing data, and that join only happens where both live: in the warehouse. Single source of truth for analytics covers the setup, and metric trees show how MRR breaks down into the drivers teams can act on.
Where Mitzu fits?
Mitzu is an analytics platform on your data warehouse for BI, product analytics and experiment analysis, where the data team governs the metrics and everyone uses them through dashboards, an AI agent, Slack and their own AI assistants.
Your data team defines revenue metrics once, on the invoice or subscription tables already in your warehouse, and every dashboard, insight, agent answer and scheduled agent reuses them. Because product and marketing events sit in the same warehouse, a revenue chart can be broken down by country, plan or acquisition channel, and followed into the funnels and retention behind it. A scheduled agent can review the revenue dashboard every Monday and email the team when something moves. Queries run in your warehouse, and nothing is copied out. The video below shows MRR measured in Mitzu on an invoice table.
Start free on your warehouse, or book a demo and we will show your MRR on your own data.
FAQ
How do you calculate MRR?
Add up the recurring charge of every active subscription, normalized to one month: monthly plans at their monthly price, annual plans divided by 12, quarterly plans divided by 3. Apply recurring discounts and leave out one-time fees, taxes and free trials. For example, 15 customers at $120 a month and 5 customers at $1,200 a year give $1,800 + $500 = $2,300 of MRR.
How do you calculate net revenue retention?
Take the recurring revenue of the customers who were paying at the start of the period, add their expansion, subtract their contraction and churn, and divide by the starting revenue. New customers are excluded. If customers paying $100,000 of MRR a year ago now pay $112,000, annual NRR is 112%.
What is a good NRR?
Above 100% means existing customers grow revenue on their own, which is the main threshold. What counts as good beyond that depends on who you sell to and how you price: companies with enterprise customers and seat or usage expansion usually report higher NRR than those selling monthly plans to small businesses. Check that a benchmark defines NRR the same way you do before comparing.
What is the difference between NRR and GRR?
Net revenue retention includes expansion from existing customers; gross revenue retention does not. GRR = (starting MRR - contraction - churn) / starting MRR, so it is capped at 100% and shows pure revenue loss. NRR adds expansion back. A company with 92% GRR and 112% NRR loses 8% of revenue a year and more than makes it up through upgrades.
Is NRR the same as net dollar retention?
Yes. Net revenue retention (NRR) and net dollar retention (NDR) are two names for the same metric: recurring revenue kept from an existing customer cohort, including expansion, contraction and churn. Companies still differ in the details, such as whether reactivations or usage revenue count, so read the definition behind any reported figure.
