What Churn Rate Means, and What Yours Should Be

Churn rate is customers lost over customers at the start. The formula, the benchmarks that actually compare to yours, and three fixes with numbers.

By Dustin W. StoutPublished 05 Nov 202614 min read

Somebody cancelled this morning and you will not know the reason until the first of next month, when the churn rate turns them into a percentage.

By then it is nobody in particular.

Churn rate is the share of your customers who were there at the start of a period and gone by the end of it: customers lost divided by customers at the start, times 100. Four hundred and twelve accounts on the first of the month, eleven of them cancelled by the last, and the churn rate is 2.67 percent.

The arithmetic takes four minutes. What you do with the answer decides whether the business compounds or leaks.

So here is the formula, the benchmarks that are actually comparable to yours, and three fixes ordered by how fast the money comes back.

The churn rate formula, in one minute

Billing systems and buyers use the same definition, so start there. Salesforce's own explainer states it plainly: customers lost divided by customers at the start of the period, multiplied by 100 (Salesforce).

Three decisions hide inside that sentence, and most companies make all three without noticing.

The denominator is the customer count on day one, not the count at the end and not the average. Anyone who joined during the month stays out of it, because a customer you gained cannot have churned.

The numerator counts cancellations among the customers who were there on day one. A refund is a separate event, and so is a failed payment. Folding them in early makes the number impossible to act on.

The period stays the same every month. A number that means 28 days in June and 31 in July is a number you cannot compare to anything, including itself.

Run it like this:

Export the customer list twice: active on the first of the month, active on the last day.
Count the accounts in the first list that are missing from the second, and check each one against its cancel date.
Divide that count by the size of the first list, then multiply by 100.
Write it in a sheet with the date, beside last month's figure, and stop editing the definition.

Worked, on a real shape: 412 accounts on 1 October, 11 of them cancelled by 31 October, and 9 new accounts signed up in between. Churn is 11 divided by 412, which is 2.67 percent. It is not 11 divided by 421, and it is certainly not 11 divided by 403.

That gap widens as growth slows. Add 40 accounts in a month and dividing by the closing count cuts your reported churn by nearly a tenth without a single customer deciding to stay.

Then annualise, because that is the number a buyer or a board will ask for. Multiply by twelve and you overstate it. Compound it and you get the truth:

$$1 - (1 - 0.0267)^{12} = 0.277$$

Just over 27 percent of the base gone in a year at 2.67 percent a month. Do it with 2 percent and you get 21.5 percent, not 24. Small gaps at the monthly level become large gaps in the sheet.

One case breaks the formula honestly. If you sell annual contracts, monthly churn is a fiction: every customer reaches a single renewal date and stays or leaves there. Measure those as a cohort at renewal, and keep the monthly calculation for plans that bill monthly.

What a good churn rate is at your price point

There is no single benchmark. The published ones disagree by a factor of three because they measure different businesses on different bases, so match the sample before you match the number.

Read the second row and then the last one and the disagreement is plain. One dataset says small SaaS loses a third of its customers a year. Another says the median is under 4 percent.

Both are right about their own sample.

SaaS Capital surveys private B2B SaaS companies above $1M ARR, most selling contracts to businesses, and reported median gross revenue retention of 91 percent across its 2025 population (SaaS Capital, September 2025). Gross retention ignores expansion, so 91 percent means about 9 percent of recurring revenue walked out.

Recurly measures subscriptions running through its billing network, a population weighted toward self-serve and monthly card payments, and put median annual churn at 3.22 percent on July 2026 data (Recurly).

For a contracted B2B business, trust SaaS Capital. For a $29 tool billed monthly to anyone with a card, use neither of those and use CRV's band, which puts SMB logo churn near 4 percent a month at that contract size (CRV, 2026).

Two companies with the same $600,000 of ARR show why the band matters more than the number.

The first has 1,200 accounts at $42 a month and loses 3.4 percent of them a month, which is 41 accounts and $1,722 of MRR gone. That is a normal figure for its band. It also means replacing about $20,700 of lost MRR a year before the business grows at all.

The second has 40 accounts at $1,250 and loses one every seven weeks. On a monthly chart it looks lumpy and alarming. It costs $9,300 a year.

The second business is more than twice as good at keeping what it sells, and its dashboard looks worse.

The move is smaller than the argument. Pick the row closest to your price and contract length, compute your own number on the same basis, and compare those two. A 4 percent monthly logo churn on $49 plans and a 4 percent annual revenue churn on $40,000 contracts are not the same disease. The SaaS section holds the neighbouring metrics if you want the rest of the sheet.

A workshop bench holding four glass beakers at different levels, one for each churn rate band, with a brass plate bearing The Board's mark

Logo churn and revenue churn, split apart

Two customers can cancel for the same reason and cost you wildly different amounts.

Take a business with 40 accounts and $8,000 of monthly recurring revenue, so $200 a month on average. In one month it loses four accounts: three on the $49 plan and one paying $1,200.

Logo churn is 4 divided by 40, or 10 percent.

Revenue churn is $1,347 divided by $8,000, or 16.8 percent.

Four cancellations, two stories, and the logo number is the one most dashboards show by default because it is easier to count. Revenue is the one your bank reads.

Add expansion and you get net revenue retention, where a business can look healthy while shrinking. If those same 40 accounts upgraded by $300 in the month, net revenue churn is $1,047 divided by $8,000, or 13.1 percent. Your net retention is 86.9 percent, against a median of 101 percent for private B2B SaaS in 2025 (SaaS Capital, September 2025).

Expansion hides leaks, which is why gross retention belongs beside it. Net retention above 100 percent means the average customer is worth more each year even while others walk out. Gross retention is the floor: at a 91 percent median, the average private company has to replace 9 percent of its recurring revenue with new logos every year before it grows.

Read net against gross in the same month. Ten points of daylight between them means expansion is carrying the headline, and a product change that dents upgrades will show up as a cliff rather than a slope.

The ratio between logo churn and revenue churn is the diagnostic. When revenue churn runs far above logo churn, your biggest accounts are leaving and the small ones are staying. When logo churn is high and revenue churn is modest, you have a volume problem at the bottom of the market, which is usually price and packaging rather than product. The SaaS pricing models breakdown covers how to move a tier without losing the accounts sitting on it.

Accounts lost and MRR lost, with MRR gained from expansion beside them, updated on the same day each month, is the whole engine in three rows.

Find out why people leave before you fix anything

Every fix below depends on something you already have and have never read closely: the cancellation reason.

Read the sentence behind the dropdown.

Pull every cancellation from the last 90 days with plan, MRR, tenure and the reason field, then sort them into four buckets.

Then call ten of them this week. Ten minutes each, and ask the same three things.

What were you trying to get done the week you signed up?

What did you use instead, and what made you pick it?

When did you last open it, and what happened that day?

Write their words down, not your category. "I couldn't get my team to log in" points at a specific screen. A tag that says "product mismatch" points at nothing.

Ten calls is a pattern, not a rate. If six of the ten name the same missing thing, that is your quarter. If no two of them agree, the sample is too small and the next ten will tell you more than any dashboard has.

A month after the cancellation is the right moment to ask. On the day itself, the answer is about money and mood. Thirty days later it is about what they replaced you with.

A black desk telephone with its receiver off the hook, handwritten index cards fanned beside it and a small printed card with The Board's mark propped against the base

Two failure modes. Asking people who already left to fill in a form gets you a 4 percent response rate from the angriest segment. Believing the reason they select gets you the polite answer, which is usually a different problem.

A phone call gets you the real one in about six minutes.

Whichever bucket dominates decides which fix you run first. A bucket holding less than a fifth of the cancellations is not this quarter's problem, however loud it is. If the third one dominates, no onboarding change will save you, and the work is positioning, which the SaaS marketing guide treats as the starting point rather than the campaign.

Fix one: the first fourteen days

Most churn is decided in the first two weeks and reported in the fourth month.

Find the activation event first. Take every account still paying at day 90 and every account that churned before day 90, and compare what each group did in week one. Look for the single action with the widest gap. It might be inviting a second user, connecting a data source, sending the first invoice.

A gap of 30 points or more is the signal. Below that, keep looking, because a weak correlation will have you optimising the wrong screen for a quarter.

Then build the path to it:

Put the three steps leading to that event on the account's home screen, as a checklist that disappears when it is done.
Send one email on day two that names the next step and links straight to it.
Send a second on day seven with a 90 second video of somebody doing it.
Have a person call every account above $100 a month that has not taken the first step by day ten.
Watch the completion rate weekly. Under 40 percent, the obstacle sits in the product and no email will move it.

The arithmetic, at 500 accounts on $89: monthly churn of 2.9 percent is 14.5 accounts gone a month. Move it to 1.9 percent and you keep five of them. At $89 that is $445 of MRR still billing every month, or $5,340 more across a year than the same business without the change, before those accounts' own retention is counted.

A figure in a dark coat at a distance, pushing open a lit door in a dim hallway, a small enamel sign with The Board's mark on the frame

Two things go wrong. Teams call the activation event "use the product", cannot measure it, and ship a checklist nobody finishes. Teams also build the checklist on the marketing site, where the person who is stuck is not looking.

This is the same work as conversion rate optimization, moved to the other side of the payment, and it pays better because the customer has already given you money once.

Fix two: the payments that fail quietly

This is the cheapest churn you own, because the customer never decided to leave.

Recurly's own guidance is to audit involuntary churn first when a SaaS business runs above 5 percent annual churn, since it is often fixable without touching the product (Recurly, July 2026).

The causes are dull. An expired card, a replacement card from a new bank, a limit hit while the customer was travelling, a bank declining a foreign charge.

The fix is a sequence, not a slogan.

Turn on card account updater in your payment processor. It catches replacement cards when the networks issue them, and most processors charge nothing extra for it.
Send a pre-dunning email seven days before a card expires, with a one tap link to update it. Most customers fix it before the charge ever fails.
Retry on day 0, then day 3, then day 5, then day 7. Retrying three times in an hour is how a small decline becomes an account closure.
Keep the account working for seven days after the first failure, with a banner that says the payment needs attention. Cutting access on day one turns a billing problem into a cancellation.
Send the final email from a person's address with a phone number in it. Nobody answers a no-reply.

The arithmetic: 500 accounts with 1.2 percent of charges failing is six failed payments a month. Default recovery of 40 percent loses 3.6 of them. Lift recovery to 70 percent with the steps above and you keep 1.8 accounts a month, which is $160 of MRR at $89, or roughly $1,920 across a year.

Unremarkable money, and the easiest on this list to collect. It needs no product change and no marketing.

A cream envelope pushed halfway back out of a steel mail slot in a dark door, a small printed card with The Board's mark behind the flap

Fix three: the cancel button, priced honestly

The cancel screen is the last conversation you get, and most companies spend it trying to win an argument.

Treat it as a choice between two real alternatives and a way out. Someone who hit a usage limit needs a smaller month. Someone whose project ended needs a pause.

Offer a 60 day pause, keeping their data, with a date they choose to resume.
Offer the tier below, naming the limit that caused the problem on the screen itself.
Let them cancel in one click, keep the reason field optional, and send the confirmation immediately.

The pause is the one that pays.

On 100 cancels a month at $89, if one in five takes a pause and half of those are still paying a year later, ten accounts come back at full price. Ten accounts at 2.5 percent monthly churn carry an average life of 40 months.

That is $35,600 of revenue the cancel button would have taken with it. If only three of them stay, the screen still paid for itself.

Then price the discount honestly. Two months at half price on an $89 plan costs $89. If it saves one account in five and a saved account is worth eighteen more months, the expected value per cancel sits well above the cost. If it saves one in fifty, you have taught fifty people to threaten to cancel.

Hiding the cancel link and requiring a phone call both produce chargebacks, and so does confirming a save the customer never agreed to. Each one costs you the fee and a slice of the processor's trust. That is churn you created in the act of preventing it.

The number to read every Monday

Aggregate churn hides everything. Cohorts show it.

Build a sheet with one row per signup month and five columns: accounts at signup, retained at day 30, retained at day 90, MRR retained at day 90, and the top reason on file.

Read the curve, not the average. If the cohort from six months ago sits at 78 percent retained at day 90 and this month's sits at 64 percent, something changed in onboarding and the aggregate will not tell you for two more months. That lag is why teams feel ambushed by churn they caused in spring.

Then read the money column against the count column. A cohort keeping 80 percent of its accounts and 55 percent of its revenue has a pricing problem at the top of its range. A cohort keeping 60 percent of its accounts and 78 percent of its revenue is losing small accounts fast, and the work belongs on the entry tier.

Spending on acquisition only makes sense against how long the customer stays, which is the relationship CAC is supposed to describe, and nobody computes it honestly at 4 percent monthly churn. Advertising fills the top of the curve and cannot change its shape. The Board is ours: a leaderboard where brands buy their position, every listing a paid placement, rank set by the amount paid and sorted, 26 brands holding a row today. It sells impressions at a price anybody can check, and it will not make a visitor stay. If a paid visit leaves in month two, the placement did its job and the product did not.

Open the billing export, write last month's churn rate on a sticky note, and put it where you will see it on Monday. Then read the reason field on the last twenty cancellations.

One of those two sources is lying to you. By Friday you will know which.