Customer churn analysis: a step-by-step guide for SaaS

Customer churn analysis is the work of finding out why customers leave and which ones are likely to leave next. A churn rate tells you how many customers you lost. The analysis tells you who they were, what they did before leaving and what you can change. For a subscription business, the most useful version splits churn into two parts from the start, voluntary cancellations and failed payments, because each has a different fix.

TL;DR

• Churn analysis goes beyond one number: define churn, calculate it, segment it, find the behaviors that precede it, then act and measure.
• Track customer churn and revenue churn separately. They can move in opposite directions.
• Never multiply monthly churn by 12. A 5% monthly rate compounds to about 46% a year.
• Separate voluntary churn (cancellations) from involuntary churn (failed payments) before you look for causes.
• Close the loop: every finding should become a retry rule, a cancel-flow offer or a win-back message, and each should be measured.

What does a customer churn analysis cover?

Customer churn is the number of existing customers you lose, for any reason, over a given period. That is the measurement. The analysis is everything you do with it:

1. Define what counts as a churned customer in your business.
2. Calculate churn for a fixed period, for customers and for revenue.
3. Segment the result by plan, tenure, acquisition channel and region.
4. Diagnose the behaviors and events that show up before customers leave.
5. Act on what you found.
6. Measure whether the action worked.

A subscription business has an advantage at steps 1 and 4: its billing data says exactly when a customer left and why the payment stopped. Most teams never join it to their product data.

Step 1: Define churn for your business

Before any formula, write down what “churned” means. The answer is not the same everywhere:

• Monthly plans: a customer who cancels, or whose subscription ends after failed payments, by the end of the billing period.
• Annual plans: a customer who does not renew. Count them in the month the renewal was due, not the month they stopped logging in.
• Usage-based or freemium products: a paying customer who downgrades to free, or whose usage drops to zero for a window you choose.
• Paused subscriptions: decide in advance whether a pause is churn, and apply the rule consistently.

Two definitions matter most. Voluntary churn is a customer choosing to leave. Involuntary churn is a customer who leaves because a payment failed (an expired card, insufficient funds, a decline from the card issuer, the bank that issued the card) and was never recovered. They have different causes and different fixes, so keep them in separate columns from the start.

Decide too when a failed payment becomes churn. Billing tools do not all decide it for you the same way: Stripe’s analytics count past_due subscriptions as active and only count a subscriber as churned once their subscriptions are canceled or unpaid. If your dunning runs for weeks, churn shows up late.

Step 2: Calculate customer churn and revenue churn

Customer churn rate

HubSpot gives the standard formula: customers lost during a period, divided by customers at the start of the period, times 100. With 250 customers at the start of a month and 10 lost, monthly churn is 4%.

A second example you can reuse as a template. You start the quarter with 400 customers and lose 18 by the end of it:

• 18 ÷ 400 = 0.045
• Quarterly customer churn is 4.5%

Count only customers who were present at the start of the period. New customers who joined and left inside the same period belong in a separate early-churn metric. The variations are in how to calculate churn rate, and the companion metric in customer retention rate: formula and examples.

Revenue churn

Customer churn treats a $9 customer and a $900 customer as equal. Revenue churn does not. A company that loses two $5-a-month customers can have a higher customer churn rate than one that loses a single $100-a-month customer, while the second has the higher revenue churn. Gross revenue churn divides the monthly recurring revenue (MRR) lost in a period by the MRR at the start: $100,000 of MRR losing $5,000 in a month is 5%.

Report both numbers every month. If customer churn is rising but revenue churn is flat, your cheapest plan is leaking. If revenue churn spikes while customer churn is flat, a few large accounts left, and the cause is probably account-level, not product-wide.

Monthly versus annual churn

Churn compounds. You keep 95% of customers each month at 5% monthly churn, so after 12 months you keep 0.95 to the 12th power, about 54%, and lose about 46%, not 60%.

1% a month 11.4%

2% a month 21.5%

3% a month 30.6%

5% a month 46%

0%10%20%30%40%50%

Not 12 × the monthly rate

This is why a monthly rate that looks small deserves attention. At 2% monthly churn you lose about 21.5% of a cohort in a year. At 3% you lose about 30.6%.

Step 3: Segment before you interpret

A single blended churn rate hides most of what you need. Churn on a cheap tier can look alarming while revenue churn stays healthy, and a few lost large accounts can do the opposite. Measure by cohort and by segment.

The cuts that usually reveal something:

Segment	What it can show
Plan or price tier	Whether the entry plan is leaking or a mid-tier is mispriced
Tenure (month 1, 3, 6, 12)	Whether the problem is onboarding or long-term value
Acquisition channel	Which sources bring customers who stay
Billing interval	Whether annual customers renew at a different rate than monthly ones
Payment method and card issuer	Where failed payments concentrate
Company size or use case	Which customers get the most value from the product

Cohort analysis is the cleanest version. Group customers by the month they signed up and track what share of each group is still paying after 1, 2, 3 and 12 months. If newer cohorts retain worse than older ones, something changed: pricing, onboarding, the channel mix or the product. Churn rate analysis by segment walks through this in detail.

Step 4: Combine the right data

Churn analysis fails when each team looks at its own tool. Four groups of data, joined on the same customer ID, cover most needs:

• Product usage: logins, feature use, session length, completion of the core action.
• Billing and subscription records: transaction history, payment method updates, dunning attempts (the automated retries and reminder emails sent after a failed payment), plan changes.
• Support interactions: ticket volume, response times, resolution rates, satisfaction scores.
• Marketing engagement: email opens, clicks, in-app interactions.

The practical requirement is a shared customer ID across your billing system, product analytics, help desk and email tool. Without it you can count churn but you cannot explain it.

Survey data belongs here too. Net Promoter Score (NPS) and customer satisfaction (CSAT) answers are early signals; see predicting churn with HubSpot NPS and CSAT surveys.

Step 5: Find what predicts churn

Once the data is joined, you are looking for behaviors that separate customers who renew from customers who leave. The method has three parts: subscription data tells you who churned, event data is summarized into behavioral metrics for each subscriber, and the two are compared to see which behaviors go with renewal and which with churn. The result can become a churn-risk score for each subscriber.

You do not need a data science team to start. A workable first pass:

1. Pull a list of customers who churned in the last 6 to 12 months and a same-size sample of customers who stayed.
2. For each, record usage in the 30 days before the churn date (or the same window for the stayers): logins, core actions, seats added, support tickets.
3. Compare the two groups. The metrics with the biggest gap are your candidate warning signs.
4. Test each candidate on a later period before you trust it.

The process is worked through with an example in how to find what predicts churn with your own customer data, and the modeling side in AI churn prediction: how it works.

Pair the numbers with what customers say. Cancellation survey answers, exit interviews and support conversations tell you why a pattern exists. A drop in usage may be a missing integration, a price objection or a champion who left the customer’s company. The numbers show where to look; the conversations explain it.

Step 6: Analyze failed payments as their own problem

Many churn analyses stop at product behavior and miss a loss that has nothing to do with customer intent. Involuntary churn happens when a card expires, a bank declines a charge or a balance is short, and the subscription ends before anyone decides to cancel.

Treat it as its own dataset. A starting list of payment metrics:

• First-attempt failure rate: failed charges divided by total charges, by month, and by card issuer and country if you can.
• Recovery rate: invoices eventually paid divided by invoices that failed.
• Time to recovery: days between the first failure and the successful payment.
• Decline reason mix: temporary (soft) declines such as insufficient funds versus permanent (hard) declines such as a closed card. Soft declines can often be retried. Hard declines need a new payment method from the customer.
• Involuntary share of total churn: revenue lost to payment failure divided by total revenue lost.

The last metric tells you how much of your churn a better recovery process can address. Measure retries and customer emails separately, so you know what each one recovers. Decline reasons are explained in credit card decline codes.

Step 7: Turn findings into actions

An analysis that ends in a slide deck does not reduce churn. Map each finding to a specific action:

Finding	Action
Customers who skip the core action in week 1 churn at a high rate	Change onboarding to drive that action; trigger an email when it is missed
Customers on one plan cancel for price reasons	Offer a discount, pause or downgrade inside the cancel flow
Failed payments concentrate on expiring cards	Send card-update reminders before expiry and retry on a schedule
Large accounts show falling usage before renewal	Alert the account owner and reach out
Cancelled customers cite a missing feature	Send a win-back message when the feature ships

The levers fall into three groups: payment recovery, cancellation handling and win-back. For product-led companies, NoChurn for product-led SaaS shows how usage signals feed these actions, and customer engagement strategies for SaaS covers the proactive side.

Step 8: Measure whether it worked

Every action needs a before-and-after comparison, otherwise you will credit or blame the wrong change. Keep it simple:

• Pick one metric per action. Recovery rate for payment changes, save rate for cancel-flow offers, reactivation rate for win-back.
• Compare cohorts, not months. Compare customers who saw the change with those who did not, or compare matching cohorts before and after.
• Change one thing at a time when you can, or stagger changes by a few weeks.
• Check both churn numbers. A discount that saves customers but lowers revenue per account can improve customer churn and hurt revenue churn.
• Review on a schedule. Monthly for churn rates, quarterly for cohort retention curves.

Where should you start?

Most churn analysis is done once, presented and forgotten. The useful version is continuous: the same data that explains churn should also trigger the response.

Start with the split between voluntary and involuntary churn. Failed payments are not a decision by the customer, so they are the most recoverable loss a subscription business has: an expired card is a reminder problem, not a product problem. On the voluntary side, the cancel flow is the last place you can learn something and still change the outcome. A good flow asks why, offers something relevant and records the answer, which feeds the next round of analysis.

FAQ

What is a customer churn analysis?

It is the process of measuring how many customers you lose, then explaining why and predicting who is next. It includes calculating churn rates, segmenting customers, finding behaviors that precede cancellation and testing whether your retention actions work.

How do you measure customer churn?

Divide the customers you lost during a period by the customers you had at the start, then multiply by 100. With 10 lost out of 250, monthly churn is 4%. Calculate revenue churn the same way using MRR.

What does 5% churn mean?

It means that, over the period measured, 5 of every 100 starting customers left. If the period is a month, the effect over a year is about 46%, not 60%, because the losses compound.

What data do I need for churn data analysis?

At minimum, billing records and product usage tied to the same customer ID. Support tickets, survey scores and email engagement add detail. Add payment failure and decline data if you want to separate involuntary churn.

How often should I run a churn analysis?

Check churn rates monthly and review cohort retention quarterly. Re-run your predictive signals whenever you change pricing, onboarding or the product in a meaningful way, since those changes can alter which behaviors predict cancellation.

How NoChurn helps

NoChurn runs the actions that come out of the analysis, every day: it recovers failed payments with retries and AI-written emails, runs cancel flows with an offer per reason (and one-click cancel always visible), sends win-back emails at day 30, 60 and 90, and scores each customer’s churn risk from your billing data and the attributes you send by API. It connects to Stripe in one click and works with Chargebee, Recurly, Paddle and other billing tools, or with any billing through the API. Pricing depends only on MRR, free up to $300 of MRR, with no commission on recovered revenue (your own domain for emails and hosted pages as a $14.99/month add-on on any plan). See pricing, or check your numbers first with the free churn tools.

Sources

All accessed 14 November 2026.

• How to calculate churn rate in 5 easy steps, HubSpot (updated 9 September 2025)
• Analytics (MRR, subscriber churn rate), Stripe documentation