Why a Good NPS and Rising Churn Can Both Be True
The NPS report says customers are happy. The renewal report says they are leaving. Both can be accurate, because they measure different things. NPS records how likely the people who answered a survey are to recommend you. Churn records whether a company kept paying. Between those two sits a lot of room: who answered, who signs, who controls the budget, and whether the invoice was paid.
NPS is not noise. Buffer's analysis of its own users found that people who scored 0 churned the most and people who scored 7 to 10 churned the least (Buffer). But a score that tracks churn at the level of individual users can still miss a spike across your book of business. The nine reasons below explain why. Each one comes with a check you can run on your own data.
This post is about NPS as a misleading signal. For the wider case for continuous, behavior-based risk detection, see why your CSMs miss churn in progress.
1. NPS Was Built to Measure Advocacy, Not to Forecast Renewals
The NPS question asks whether someone would recommend you. It does not ask whether they will renew. When ProfitWell compared NPS with retention across about 2,000 companies, companies in the lower quartile and the middle of their industry's NPS range had "essentially the same retention." Only companies in the upper quartile of their industry showed a clear effect, at 5 to 10% higher retention (ProfitWell).
The same research found a stronger link between NPS and expansion revenue. So a good score may tell you something about growth potential while telling you little about who renews.
Check: Plot last year's renewals against each account's most recent NPS. If renewed and churned accounts have similar score distributions, stop using NPS as a renewal forecast input.
2. Your Score Describes Only the Customers Who Answered
An NPS of +45 from a minority of your customers says nothing about everyone else, and the people who skip surveys are not a random sample. Typeform's customer success team segmented net MRR churn by NPS response in ChartMogul. Customers who never answered churned at a much worse rate than detractors who scored 0 to 6. As Typeform's David Apple put it, "Someone who gives us a low score on the NPS is still less likely to churn than someone who didn't respond at all" (ChartMogul).
Check: Report response rate next to the score every time. Then split last quarter's churned accounts into promoters, passives, detractors, and non-respondents. If non-respondents account for most churn, your score is describing the safest part of your base.
3. The Person Who Answered Is Not the Person Who Signs
NPS surveys usually reach day-to-day users. Renewals are decided by an economic buyer, a finance lead, or procurement, who may never open the product or the survey. A team of enthusiastic users can give you 9s and 10s while the budget owner decides the tool is not worth the line item.
Check: Tag each NPS response with the respondent's role. For accounts renewing in the next two quarters, confirm you have a recent signal from the person who signs, not just from users.
4. The Champion Who Gave You a 9 Has Left
A score belongs to a person, and people change jobs. When the champion who scored you highly leaves, the score stays on the account record and keeps looking green. Their replacement inherits a tool they did not choose and starts by asking what it costs. We cover how to handle this in managing churn risk when the executive sponsor leaves.
Check: For every promoter response older than six months, confirm the respondent still works at the account and still owns the relationship.
5. A Budget Cut or Vendor Consolidation Outranks Satisfaction
When a customer cuts spend, tools are ranked by business impact, redundancy, and cost. Satisfaction helps, but a department can rate a product highly and still be told to cancel it. Layoffs, a missed funding round, or an acquisition can trigger these decisions without any change in how users feel about your product.
Check: Compare your churned accounts with company news from the six months before cancellation: layoffs, leadership changes, mergers, or funding trouble.
6. Payments Are Failing and Nobody Decided to Leave
Involuntary churn happens when a card expires or a payment fails and the subscription lapses. The customer did not choose to leave, so no survey would have warned you. Recurly's network benchmarks (July 2026 data) list median annual SaaS churn at 3.22%, of which 2.16 points is voluntary and 1.06 points is involuntary. In the same data, involuntary churn is 0.18% for subscribers in the $250+ average-revenue-per-customer band, so the share depends heavily on your price point and how customers pay (Recurly).
Check: Look up the last invoice status for every account that churned in the spike. Report failed-payment churn separately from voluntary churn. If your churn rate also looks high against peers, this breakdown of churn above benchmark walks through what to separate first.
7. Friction Lives in Finance, Billing, or Delivery, Not the Product
Some customers leave because dealing with you outside the product is painful. On the Causal Revenue Show, Mark Stouse described a vendor whose satisfaction numbers were improving while churn kept rising. The cause came out when a customer's CEO called: the vendor's accounts receivable team threatened to cut off service whenever a customer paid two or three days late, because their bonus targets rewarded it. A second guest, Paula Skaper, described another company with very high satisfaction scores whose only customer complaint was slow invoices (Klor Consulting).
A user survey about the product will not pick up an adversarial collections process or missed delivery dates.
Check: Add "finance or billing friction" and "delivery or implementation" as root-cause options when tagging churned accounts, and pull collections and delivery history for each one. Our guide to customer churn analysis covers how to run that review.
8. A Blended Score Hides Where the Revenue Sits
NPS counts each respondent equally. A hundred happy users at small accounts can outweigh a few unhappy contacts at your largest customer. The score holds steady while the revenue at risk moves. The reverse also happens: one large account with many respondents can lift a score that most of your customers do not share.
Check: Recalculate NPS by account rather than by respondent, weight it by ARR, and break it out by segment. If the revenue-weighted score is much lower than the headline score, your risk is concentrated in accounts that matter most.
9. The Customers Churning Were Never a Fit
Customers who liked the product can still churn if their size, budget, or use case never supported the deal. On the same show, Stouse put it directly: "If you have an ICP that's too wide, or too inclusive, you're going to have more churn" (Klor Consulting). These customers often rate you well because the product is good. They leave because the economics never worked for them.
Check: Segment the last 12 months of churn by how well each account matched your ideal customer profile at signing. If poor-fit accounts are over-represented, the fix sits in qualification and pricing, not in customer success.
At a Glance
| Reason | What the NPS report shows | What to check instead |
|---|---|---|
| Built for advocacy | A good score | Renewal outcomes by score band |
| Non-respondents | Respondents only | Response rate and non-respondent churn |
| Wrong respondent | User sentiment | A recent signal from the person who signs |
| Champion left | An old promoter score | Whether the respondent still works there |
| Budget cuts | No change | Layoffs, leadership changes, mergers |
| Failed payments | Nothing | Invoice status of churned accounts |
| Finance or delivery friction | Product satisfaction | Collections and delivery history |
| Blended score | An average per respondent | ARR-weighted score by segment |
| Poor fit | Users who like the product | Churn by ICP fit at signing |
How to Diagnose a Spike in One Week
- Pull the churned list. Every account that cancelled or did not renew during the spike, with ARR, segment, and renewal date.
- Remove failed payments first. Check the last invoice status. Anything that lapsed on a failed payment goes into its own bucket.
- Add the NPS context. For each account, note the latest score, who gave it, when, and whether that person still works there. Mark non-respondents.
- Tag a root cause. Use the nine reasons above as your categories. Allow more than one per account.
- Call five former customers. Ask what bothered them most in their last six months, and have someone outside customer success make some of the calls, as the Klor Consulting recap suggests.
- Change what you report. Publish response rate and non-respondent churn next to the score, and stop using NPS alone as a renewal input.
Once you know which reasons drive your churn, decide which behaviors to watch and what to automate. Our guide to behavior-based retention automation covers that next step.
Where Quivly Fits
Quivly scores customer health from connected systems rather than from survey sentiment. A health score combines five categories: Revenue (MRR, outstanding balances, renewal timing), Product Usage, Engagement (call frequency and recency), Support (ticket volume, resolution, and response times), and Market Signals such as funding, hiring, and layoffs. Teams set the weights, and can create separate configurations for different customer segments (Quivly Docs).
Scoring reads evidence across permitted CRM, billing, support, communications, and usage sources, and keeps citations and uncertainty with its conclusions. Missing evidence is not treated as a healthy result (Quivly Docs). That matters for reason 2, where silence is the risk.
For reasons 6 and 7, the Revenue tab lists each invoice's status and due date, whether payment was early, on time, or late, and the payment attempt count when the billing provider sends it (Quivly Docs). For reasons 4 and 5, the Market Signals tab flags layoffs, executive departures, funding difficulties, and acquisitions (Quivly Docs). Agents can be triggered by health score changes, gather context, draft outreach, and pause for a person to approve before anything is sent (Quivly Docs).
Conclusion
A churn spike under a good NPS is usually not a contradiction. It means the score is measuring a narrower group, a different person, or a different question than the one your renewals answer. Keep collecting NPS for what it is good at, and report it with its response rate. For renewal risk, look at who answered, who signs, whether invoices are paid, and whether the account was a fit to begin with.
If you'd like to see how Quivly would flag these risks in your own accounts, book a demo with Quivly.
Sources
- NPS Revenue Correlation: Impact of NPS on Revenue Expansion | ProfitWell (archived) - profitwell.com via web.archive.org
- The one thing worse than a low NPS score | ChartMogul - chartmogul.com
- NPS Scores and Churn | Buffer analysis - bufferapp.github.io
- Churn rate benchmarks | Recurly Research - recurly.com
- Churn Is Not a CS Problem: Who Owns B2B Customer Churn? | Klor Consulting - www.klorconsulting.com
- Health Scores Overview | Quivly Docs - docs.quivly.ai
- Investigate customer health across connected sources | Quivly Docs - docs.quivly.ai
- Revenue Management | Quivly Docs - docs.quivly.ai
- Identifying At-Risk Customers | Quivly Docs - docs.quivly.ai
- Agents | Quivly Docs - docs.quivly.ai
