Domains
SaaS Metrics Every Business Analyst Should Understand: MRR, Churn and Retention
Part of our guide to business analyst roles and responsibilities.
The core SaaS metrics in plain words
Most software-as-a-service (SaaS) companies sell subscriptions, so revenue repeats month after month instead of arriving as one sale. Leaders want to know three things: how much repeating revenue do we have, how much are we losing, and how much do we keep from existing customers. Answering those questions with clear definitions and reliable data is a classic part of what a business analyst does in a product or SaaS team.
MRR: monthly recurring revenue
MRR is the normalised monthly value of all active, paying subscriptions on a given date. Normalised means an annual plan is divided by twelve and a quarterly plan by three. One-off items such as setup fees, training days or overage invoices are usually left out because they do not repeat. ARR (annual recurring revenue) is simply MRR multiplied by twelve.
MRR changes for five reasons, and most teams track each one separately:
- New MRR: subscriptions from customers who were not paying before.
- Expansion MRR: existing customers paying more, for example after an upgrade or adding seats.
- Contraction MRR: existing customers paying less, but still staying.
- Churned MRR: the full value lost when customers cancel or lapse.
- Reactivation MRR: former customers who come back. Agree whether this counts as new or as its own line.
Churn: what you lose
Churn always needs a qualifier. Customer churn (often called logo churn) counts customers who left. Revenue churn counts MRR lost from the customers you started with, and it comes in two forms. Gross revenue churn adds up cancellations and downgrades only. Net revenue churn takes those same losses and subtracts expansion from the same starting customers, so it can fall below zero when upgrades outweigh losses.
A company can lose few customers but a lot of revenue if the leavers were large accounts, so customer churn and revenue churn can tell very different stories. That is why every chart must name its measure.
Retention: what you keep
Retention looks at the customers you had at the start of a period and asks what is left at the end. Customer retention counts customers. Gross revenue retention (GRR) counts revenue kept and ignores upgrades, so it can never go above 100%. Net revenue retention (NRR) adds upgrades back in, so it can go above 100% when upgrades outweigh losses. Retention and churn are two sides of one coin: GRR equals 100% minus gross revenue churn, and NRR equals 100% minus net revenue churn.
Churn and retention measures compared
| Aspect | Customer churn | Gross revenue retention | Net revenue retention |
|---|---|---|---|
| Question it answers | How many customers left? | How much starting revenue did we keep, before upgrades? | How much starting revenue did we keep, after upgrades? |
| Unit counted | Customers or accounts | MRR from the starting customers | MRR from the starting customers |
| Includes downgrades | No, only full exits | Yes, as a loss | Yes, as a loss |
| Includes upgrades | No | No | Yes, as a gain |
| Includes new customers | Not in the starting count | No | No |
| Matching churn measure | Customer retention is its opposite | Gross revenue churn | Net revenue churn |
How to calculate MRR, churn and retention
The formulas are simple. The hard part is agreeing what goes into each term. Here they are in plain text, using one period such as a calendar month.
TEXT
MRR = sum of the monthly value of every active, paying subscription
(annual plans divided by 12, after discounts, before tax)
ARR = MRR x 12
Ending MRR = Starting MRR + New + Expansion + Reactivation
- Contraction - Churned
Customer churn rate = customers lost / customers at start
Customer retention rate = starting customers still active at end
/ customers at start
Gross revenue churn = (Contraction + Churned) / Starting MRR
Net revenue churn = (Contraction + Churned - Expansion) / Starting MRR
Gross revenue retention = (Starting MRR - Contraction - Churned)
/ Starting MRR
Net revenue retention = (Starting MRR + Expansion - Contraction - Churned)
/ Starting MRR
Multiply any rate by 100 to show it as a percentage.Three rules sit behind these formulas. First, the denominator is the count or MRR at the start of the period, so customers who joined during the month do not dilute churn. Second, retention follows a fixed group (a cohort): only customers who were paying on day one. Third, every term must use the same snapshot date and the same exclusions, or the bridge from starting to ending MRR will not add up.
Most teams calculate these from a monthly snapshot of subscriptions. If you are comfortable with SQL for business analysts, a simple MRR query looks like this:
SELECT
SUM(CASE
WHEN billing_interval = 'year' THEN net_amount_ex_tax / 12.0
WHEN billing_interval = 'quarter' THEN net_amount_ex_tax / 3.0
ELSE net_amount_ex_tax
END) AS mrr
FROM subscription_snapshots
WHERE snapshot_date = '2025-03-31'
AND status = 'active'
AND is_paying = 1
AND is_test_account = 0;Notice how many business decisions hide in that short query: which statuses count as active, how test accounts are flagged, whether discounts are already removed, and which date is the snapshot. Each one belongs in a written definition, not only in the code.
How to turn SaaS metrics into dashboard requirements
A dashboard requirement is not a picture of charts. It is a set of agreed definitions, data rules and checks that let anyone rebuild the same number. Use these steps.
- Start with the decision. Ask who reads the dashboard and what they will decide from it. A sales leader watching upgrades needs different views from a finance team closing the month.
- Write one definition card per metric. Include the name, formula, what is included and excluded, the timing rule, the data source and the owner. Keep the cards in a shared place, such as a data dictionary.
- Map every term to data. Name the table, field and status values behind each term. For example, which subscription statuses count as active, and where discounts are stored.
- Fix the grain and timing. State the snapshot date, the time zone, and which event date counts: the cancellation request, or the end of the paid term.
- Describe the views and filters. Typical views are an MRR bridge (starting MRR, each movement, ending MRR), a monthly trend and filters by plan or region. If you build in Power BI, each formula usually becomes a measure; see Power BI for business analysts.
- Write acceptance criteria you can test. For example: ending MRR for a closed month matches the billing export for the same date, and the bridge movements sum exactly to ending MRR.
- Agree refresh, ownership and change control. Say how often data refreshes, who signs off a definition change and how past months are treated when a rule changes.
Diagram in words
- 1. Agree the decision: Who reads the dashboard, how often, and which decision each metric supports.
- 2. Write definition cards: Formula, inclusions, exclusions, timing rule, source and owner for every metric.
- 3. Map terms to data: Tables, fields and status values behind each term, with known data quality gaps.
- 4. Specify views and filters: MRR bridge, monthly trend, plan and region filters, plus drill-down rules.
- 5. Set acceptance tests: Reconcile with billing for a closed month and check the bridge adds up exactly.
- 6. Agree refresh and ownership: Refresh schedule, sign-off owner and how rule changes affect past months.
Worked example: March metrics at Lumen Desk
Lumen Desk is a fictional company that sells helpdesk software on three plans: Starter at $50, Team at $100 and Pro at $200 per month. The Team plan can also be paid annually at $1,200 a year, which is $100 a month. Nadia, the business analyst, is defining the metrics for a new leadership dashboard. Here is the March data after the team agreed the rules. All amounts are per customer per month, after discounts and before tax.
- Starting MRR on 1 March: $10,000 from 100 paying customers (20 Starter = $1,000, 70 Team = $7,000, 10 Pro = $2,000).
- New: 10 customers joined the Team plan. 9 pay $100 monthly and 1 paid $1,200 upfront for a year, which is $100 a month. New MRR is 9 x $100 + $100 = $1,000.
- Expansion: 5 Team customers upgraded to Pro. Each now pays $200 instead of $100, adding 5 x $100 = $500.
- Contraction: 2 of the original 10 Pro customers downgraded to Team. Each now pays $100 instead of $200, losing 2 x $100 = $200.
- Churned: 5 other Team customers reached the end of their paid term in March, losing 5 x $100 = $500.
Ending MRR is $10,000 + $1,000 + $500 - $200 - $500 = $10,800, from 105 customers (100 + 10 - 5). You can check this against the ending plan mix. Starter stays at 20. Team is 70 + 10 new - 5 upgraded + 2 downgraded - 5 churned = 72. Pro is 10 + 5 upgraded - 2 downgraded = 13. That gives 20 + 72 + 13 = 105 customers and 20 x $50 + 72 x $100 + 13 x $200 = $1,000 + $7,200 + $2,600 = $10,800. Both routes agree.
From this, the rates are:
- Customer churn: 5 / 100 = 5%. Customer retention: 95 / 100 = 95%.
- Gross revenue churn: ($200 downgrades + $500 cancellations) / $10,000 = $700 / $10,000 = 7%.
- Net revenue churn: ($200 + $500 - $500 expansion) / $10,000 = $200 / $10,000 = 2%.
- Gross revenue retention: ($10,000 - $200 - $500) / $10,000 = 93%, which is 100% minus 7%.
- Net revenue retention: ($10,000 + $500 - $200 - $500) / $10,000 = 98%, which is 100% minus 2%.
The first draft was wrong. The billing export counted the annual customer's full $1,200 invoice as March MRR instead of $100. So new MRR showed as $2,100 ($900 from the 9 monthly customers + $1,200), and ending MRR as $10,000 + $2,100 + $500 - $200 - $500 = $11,900, which is $1,100 too high. Normalising the annual plan to $100 a month brings new MRR back to $1,000 and ending MRR back to $10,800. The churn and retention rates did not change, because new customers are not part of the starting cohort.
There was also a timing question about one more Team customer, separate from the 5 who churned. This customer clicked cancel on 28 March, but their paid term runs until 15 April. The team agreed to count churn when the paid term ends. So this customer is one of the 72 Team customers at the end of March, still adds $100 to March ending MRR, and will count in April's churn. If the team had counted the cancel click instead, March customer churn would have been 6 / 100 = 6% and churned MRR $600.
Why two churn numbers were both right
Lumen Desk, a fictional helpdesk SaaS company: the March metrics review before the leadership dashboard goes live.
- Helena Project sponsor
The board pack shows churn of 7% for March, but the customer success report says 5%. Which one is right?
- Nadia Business analyst
Both, because they measure different things. 5% is customer churn: 5 of our 100 starting customers left. 7% is gross revenue churn: $500 lost to cancellations plus $200 lost to downgrades, so $700 of $10,000.
- Helena Project sponsor
Then a tile that just says churn is not good enough.
- Nadia Business analyst
Agreed. Each tile will name the measure, and the definition card will be one click away.
- Mei Lead developer
One more thing. The export counted the annual customer's $1,200 invoice as March MRR. New MRR showed $2,100 instead of $1,000, and ending MRR $11,900 instead of $10,800.
- Nadia Business analyst
I will add a rule to the MRR card: annual plans are divided by 12, after discounts and before tax. Mei, can the data load apply that before the snapshot is saved?
- Mei Lead developer
Yes. I will also add a test that the bridge movements sum exactly to ending MRR.
The people and the company in this scene are fictional.
Metric definition card for net revenue retention
Metric card: Net revenue retention (NRR)
Name: Net revenue retention (NRR), monthly1
Formula: (Starting MRR + Expansion - Contraction - Churned) / Starting MRR2
Cohort: customers paying on the 1st of the month3
Excludes: new customers, one-off fees, tax, test accounts4
Churn timing: counted when the paid term ends, not on the cancel request5
March check: ($10,000 + $500 - $200 - $500) / $10,000 = 98%6
Source: subscription_snapshots, refreshed daily
Owner: Helena (project sponsor) approves changes
- Say the period. Many teams report NRR over twelve months. Lumen Desk chose a monthly view first so each month can be checked against billing. The name says which one it is.
- Same terms as the MRR bridge. Expansion, contraction and churned use the exact rules from the MRR card, so NRR and the bridge never disagree.
- A fixed starting group. Only the 100 customers paying on 1 March are in the cohort. The 10 customers who joined during March are left out.
- Exclusions in writing. Leaving out new customers, setup fees, tax and test accounts is a choice. Writing it down lets testers and finance check the same thing.
- Timing rule. The customer who cancelled on 28 March with a term ending 15 April counts in April. This keeps churn aligned with the month revenue is actually lost.
- A worked check. A hand calculation for one closed month gives testers an exact expected result: 98%, which matches 100% minus the 2% net revenue churn.
Data pitfalls that quietly break SaaS metrics
Most wrong SaaS numbers come from data handling, not from the formula. Watch for these.
- Invoices treated as MRR. Annual and quarterly invoices must be normalised to a monthly value. Otherwise MRR jumps in the month a large invoice is raised.
- One-off charges included. Setup fees, services and usage overages do not repeat. Decide whether usage revenue counts, and say so.
- Tax and discounts mixed in. Agree whether MRR is before or after discounts, and make sure tax is removed. Different systems often store these differently.
- Trials and free plans counted. A free user is not a paying customer. Track them in a separate funnel metric, not in MRR or churn.
- Wrong churn date. Counting churn on the day someone clicks cancel, instead of when the paid term ends, moves losses into the wrong month.
- Gross and net mixed up. A chart labelled revenue churn might subtract expansion or might not. Name gross or net on every tile.
- Customer, account or subscription? One customer can hold several subscriptions. Decide the unit for customer churn, or a customer who drops one of two products may be counted as lost.
- Test, internal and duplicate accounts. Staff accounts and test tenants inflate counts. Ask for a reliable flag, not a list someone remembers.
- Currency conversion. If you bill in several currencies, fix one reporting currency and one exchange rate rule, such as the rate on the snapshot date. Otherwise MRR moves when no customer did.
- Restating history silently. When a rule changes, past months change too. Record the change in a decision log and show which months were recalculated.
Profiling the subscription data before you write the requirements catches many of these early. Count records by status, look for zero-value subscriptions and compare the totals with a billing report for one closed month.
Checklist before you sign off a SaaS metrics dashboard
Use this list in a review with finance, the product owner and the developer who builds the data model. If any item is unclear, the dashboard is not ready, however good it looks.
SaaS metrics dashboard sign-off checklist
- Every metric has a definition card with formula, inclusions, exclusions, timing rule, source and owner.
- Churn tiles say which churn they show: customer churn, gross revenue churn or net revenue churn.
- Annual and quarterly plans are normalised to a monthly value.
- Tax, one-off fees, trials, free plans and test accounts are excluded or clearly labelled.
- The MRR bridge adds up exactly from starting to ending MRR for every month.
- Ending MRR for a closed month matches the billing export for the same snapshot date.
- Refresh time, time zone and the rule for restating past months are written down.
Frequently asked questions
Is MRR the same as the revenue in the company accounts?
No. MRR is an operational measure of repeating subscription value at a point in time. Accounting revenue follows recognition rules that depend on the country and the accounting standard used. The two should be explainable against each other, but they rarely match exactly. Ask the finance team how they reconcile them, and do not treat MRR as financial reporting.
Should churn be measured monthly or annually?
Both are common. Monthly figures show problems quickly and suit companies with mostly monthly plans. Annual figures suit companies where most customers sign yearly contracts, because few renewals fall in any one month. Pick the period that matches how customers buy, label it clearly, and do not convert a monthly rate to an annual one without agreeing the method.
Can net revenue retention be higher than 100%?
Yes. If upgrades and extra seats from your starting customers add more revenue than downgrades and cancellations remove, net revenue retention goes above 100% and net revenue churn drops below zero. Gross revenue retention cannot, because it ignores upgrades. Showing both on a dashboard gives a fairer picture.
Do I need SQL to work with SaaS metrics as a business analyst?
You do not need to be a data engineer, but basic SQL helps a lot. Being able to count subscriptions by status, sum MRR for a snapshot date and find test accounts lets you check definitions yourself and spot data problems before they reach a dashboard.
What is the difference between churn and contraction?
Churn means a customer stops paying altogether. Contraction means an existing customer still pays, but less, for example after a downgrade or removing seats. Customer churn ignores contraction, while gross and net revenue churn both include it as a loss, so always say which measure you mean.
This article was written by an AI system and published automatically after automated accuracy, safety and quality checks. All companies, people and numbers in the examples are fictional. BA Mentorship does not issue professional certifications and cannot guarantee any job or interview outcome. Spot an error? Email our support team and we will check and correct it.