Why is our SaaS churn rate spiking despite good NPS scores?

Churn often spikes while NPS stays high because NPS only measures the customers who answer. If 20% of accounts respond, a score of +50 says nothing about the other 80%, and in the worked example below those silent accounts produce 52 of 60 cancellations. Failed payments, departed champions and budget cuts never show up in a survey.

Rules of thumb

• Treat about a third of SaaS churn as a billing problem until your data says otherwise: Recurly’s July 2026 network benchmarks put median annual SaaS churn at 3.22%, of which 1.06 points is involuntary (Recurly churn rate benchmarks); Recurly does not publish the share, and 1.06 of 3.22 is our arithmetic.
• Expect non-respondents to churn more than detractors: Typeform found customers who never answered had a much worse net MRR churn rate than those who scored 0 to 6 (ChartMogul, 2016), and Buffer’s 2018 analysis found non-respondents churned at a higher rate historically, though Buffer noted its survey sampling may explain part of that.
• If more than half of churned accounts never answered the last survey, NPS is not your churn signal (the Silent-Share Test, below). In the worked example the silent share is 87%.
• Saving every detractor in the worked example cuts churn only from 60 to 56 accounts, so a detractor-recovery plan alone cannot reverse a spike that sits among non-respondents.
• Check whether the spike lines up with a renewal date before you blame the product: annual contracts concentrate cancellations into the renewal month.

Why does churn spike when NPS scores look good?

Churn and NPS diverge because NPS is a score from a self-selected minority, while churn is a count across every account, including the ones that never open a survey. The Net Promoter System defines the score as the percentage of promoters (9 or 10) minus the percentage of detractors (0 to 6), with passives (7 or 8) left out (Bain, Net Promoter System). Nothing in that formula accounts for who did not answer.

That gap matters most in B2B SaaS. The person who answers the survey is often the daily user who likes the product. The person who cancels is often a finance lead reviewing spend, a replacement manager who never chose the tool, or nobody at all, because a card expired. A rising NPS and a rising churn rate can therefore both be true in the same quarter, and neither number is wrong.

FAQ

Why is our customer churn rate suddenly spiking despite good NPS scores?

Customer churn often spikes under good NPS scores because the score only reflects the customers who answer the survey. Failed payments, departed champions, budget cuts and low-usage accounts sit mostly among non-respondents, so churn can double while NPS stays flat.

Do customers who skip the NPS survey churn more?

Customers who skip the NPS survey often churn more than detractors. Typeform reported that non-respondents had a much worse net MRR churn rate than customers scoring 0 to 6, as described on ChartMogul’s blog, and Buffer’s 2018 analysis found non-respondents churned at a higher rate historically.

How much SaaS churn is caused by failed payments?

Failed payments account for about a third of SaaS churn on Recurly’s July 2026 network benchmarks, which show a median annual SaaS churn rate of 3.22% with 1.06 points involuntary. Your own share may differ, so check the last invoice status of every churned account.

Is NPS a good predictor of churn?

NPS is a weak churn predictor on its own because it ignores who did not answer. It is useful for respondents: Buffer found detractors had the highest churn of those who filled in its survey. Pair it with usage and payment data.

What is the Silent-Share Test?

The Silent-Share Test divides churned accounts that never answered the last NPS survey by all churned accounts. If the result is above 50%, NPS is not your churn signal, and usage and billing data should lead the review.

The NPS-Churn Gap Diagnostic: five ways NPS and churn diverge

The NPS-Churn Gap Diagnostic lists the five causes that let churn rise under a stable NPS, the data check that confirms each one, and the save play that fits it. Work down the table in order: the first two checks take minutes in your billing system and CRM, and they explain the largest share of spikes.

No public benchmark we could verify gives the share of SaaS churn caused by champion departures, budget cuts or low usage, so the table gives the check rather than a number for those three rows. Your own billing and product data is the only reliable source for them.

Worked example: a +50 NPS from 20% of accounts and a doubled churn rate

A worked example shows how a SaaS company with 1,000 accounts can post an NPS of +50 from 200 respondents while churn doubles from 30 to 60 accounts in a quarter. All inputs below are illustrative assumptions; replace them with your own numbers.

The arithmetic, step by step:

1. NPS: 120 promoters is 60% of 200 respondents, 20 detractors is 10%, so NPS = 60 – 10 = +50.
2. Respondent churn: 8 of 200 = 4%. Non-respondent churn: 52 of 800 = 6.5%.
3. Silent share of churn: 52 of 60 = 87%.
4. Involuntary share: 20 of 60 = 33%, in line with the 1.06-of-3.22 ratio in Recurly’s SaaS figures.
5. Even if every detractor had been saved, churn would have fallen only from 60 to 56, still well above last quarter’s 30.

The lesson of the worked example is that improving the score among respondents cannot fix this spike. The two biggest buckets, failed payments and low-usage accounts, sit entirely outside the survey.

Is a churn spike as urgent as it feels?

A churn spike is urgent when the absolute count is large enough to be more than noise and the extra cancellations are not explained by a renewal date cluster. Before escalating, run three checks:

• Base size. On a few hundred accounts, a handful of cancellations moves the rate a lot. Look at the count, not just the percentage.
• Renewal timing. If a large annual cohort renewed this month, more cancellations is expected. Compare churn to the number of accounts that were up for renewal, not to the whole base.
• Revenue vs logos. Ten small accounts on a free-to-paid plan are a different problem from two enterprise contracts. Check churned MRR alongside churned logos.

If the spike survives all three checks, treat it as real. If it does not, say so to the team; telling people a spike is a renewal cluster is as useful as finding a cause.

What to do in the next 24 hours and the next 7 days

The first 24 hours after spotting a SaaS churn spike are for sorting accounts into the five diagnostic causes, because the fix differs completely by cause.

Next 24 hours

1. Export every account that cancelled or failed to renew this period, with MRR, last invoice status, last survey response (or none) and weekly active users.
2. Tag failed-payment accounts first. These are recoverable now, and the playbook is on our page on failed payment recovery with call and SMS.
3. Calculate the Silent-Share Test: churned accounts with no survey response divided by all churned accounts.
4. List every account still active that shows the same pattern: failed card, usage drop or a changed admin contact.

Next 7 days

1. Call or message the at-risk active accounts, starting with the highest MRR. A conversation with a new owner or a finance lead often saves more revenue than any email sequence.
2. Contact the churned accounts with an offer that matches the cause. The cause-matched offers are covered on our guide to winning back churned SaaS customers and our page on recovering cancelled subscriptions with outbound agents.
3. Add usage and payment status to the churn report so next quarter’s review does not rely on NPS alone.

How to stop NPS hiding churn next quarter

The lasting fix for a churn spike under a good NPS is to report NPS alongside response rate and non-respondent churn, so the score can never be read without its blind spot. Four changes do most of the work:

• Publish the response rate next to the score. A +50 from 20% of accounts and a +50 from 70% are different facts.
• Track churn by survey segment every month: promoters, passives, detractors and non-respondents, as in the worked example.
• Survey the buyer, not only the user. Send relationship surveys to the admin and the budget owner as well as daily users.
• Move annual renewals onto a calendar. Contact owners before the renewal notice, and see our page on moving SaaS customers to annual plans for when longer terms help.

Doing this in-house is a few hours of analyst time a month once the billing, CRM and product data are joined, but the outreach is the expensive part: someone has to call silent accounts, reach new owners and follow up every failed card. Teams that run that outreach with AI calling and SMS often use the same setup described on our AI appointment setting page and our SaaS lead generation guide.

Sources

1. Recurly, Churn rate benchmarks (July 2026 network data; read 2026-11-01).
2. Bain & Company, Measuring your Net Promoter Score (read 2026-11-01).
3. ChartMogul, The one thing worse than a low NPS score (Typeform example, 18 April 2016).
4. Buffer, NPS Scores and Churn (16 July 2018).