.png)
Frameworks, core principles and top case studies for SaaS pricing, learnt and refined over 28+ years of SaaS-monetization experience.
Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.
Understanding Customer Churn How to Measure Churn Reasons and Exit Survey Data
A churn dashboard can look healthy while the business is quietly losing its best customers. A company may show 105% net revenue retention because a handful of large accounts expanded, even as smaller accounts leave at a troubling rate. The aggregate figure tells management that revenue held. It does not explain who left, what they lost, or whether pricing had anything to do with it.
That distinction matters more when a company is changing packages, moving from seats to usage, raising prices, or introducing a new product tier. A weak churn measure turns every cancellation into an argument for discounting. A sound measure separates product gaps, failed adoption, budget pressure, buying friction, and genuine price or billing problems.
Monetizely’s position is clear: customer churn should be managed as a loss-accounting system, while exit surveys should be treated as directional evidence that must be checked against account facts. Pricing should change only when gross retention and verified exit evidence show that the package, price point, or billing metric caused the loss.
Churn has to separate lost accounts from lost dollars
“Churn” is often used as one number. In practice, operators need at least four measures because each answers a different question.
The denominator matters. New customers acquired during the period must not enter the starting cohort. Otherwise, a company can conceal a weak retention problem behind strong acquisition.
The following exhibit shows how the measures work for a SaaS company that began the year with 100 customers and $1.0 million in ARR.
The example makes the central point: NRR of 100% can coexist with 8% logo churn and 15% gross revenue churn. Expansion may have covered the loss financially, but it did not remove the product, segment, or pricing problem that caused eight customers to leave.
For pricing decisions, GRR should carry more weight than NRR. NRR is valuable because expansion signals customer value. GRR is more useful when management needs to know whether the existing offer is retaining the revenue it already won.
Public retention figures reward expansion as well as survival
Public SaaS disclosures reinforce why management should not confuse NRR with customer health. Each company below tracks a version of dollar-based retention, yet the measure includes different combinations of expansion, usage, contraction, and attrition.
The published range runs from 108% to 126%, but the strategic lesson is not that every SaaS company should seek a number above 120%. The lesson is that expansion can mask a material loss of customers or contracted revenue, especially when a company has a concentrated enterprise base.
Snowflake’s consumption model makes this especially visible. A customer can remain active while reducing usage. Twilio similarly reports that its expansion rate falls when customers reduce usage, stop using products, or receive lower usage prices. Those are commercially different events, even when they land in the same net-retention calculation.
Management therefore needs a retention scorecard with three views:
- Customer view: logo churn by segment, tenure, and contract size.
- Revenue view: GRR and NRR by the same cuts.
- Cause view: verified churned ARR by root cause, with unknown causes shown separately.
A company that only reports NRR is measuring the net result of several forces. A company that reports all three views can decide which force to address.
An exit survey records what a respondent is willing and able to say at the moment of cancellation. That makes it useful, but incomplete.
A procurement lead may select “too expensive” because the company cut budget. A departing administrator may cite “missing features” because the product never reached the team that would have found it valuable. A champion may choose “other” because the actual reason is a merger, a new corporate standard, or an internal political decision.
Survey research in other service settings has repeatedly found that response rates, survey mode, and wording can affect reported satisfaction. A 2025 peer-reviewed study of B2B digital subscriptions also used archival account data alongside cancellation reasons rather than treating survey labels as self-sufficient proof.
The operating answer is not to abandon exit surveys. It is to connect every response to observable account data.
| Reported exit reason | The question to ask next | Account facts to inspect | Likely owner of the response |
|---|---|---|---|
| “Too expensive” | Was the price high relative to realized value, or did the customer fail to adopt? | Active users, feature use, discount history, support tickets, contract value, budget change | Pricing, customer success, product |
| “We did not use it enough” | Why did adoption stall? | Time to first value, implementation status, champion activity, integrations, training attendance | Product, onboarding, customer success |
| “Missing feature or integration” | Did the gap block a core workflow or only a secondary use case? | Feature requests, usage path, competitor named, segment needs, roadmap status | Product, packaging |
| “Billing was unpredictable” | Did the meter match the customer’s ability to forecast spend? | Usage pattern, invoice variance, overages, contract terms, pre-purchase estimate | Pricing, billing operations |
| “We chose a competitor” | Did the buyer receive a better product, package, commercial term, or procurement fit? | Competitor named, evaluated features, security needs, implementation effort, total contract value | Product marketing, sales, pricing |
| “Business closed or changed strategy” | Is the event outside the company’s reasonable control? | Acquisition, shutdown, headcount cut, business-unit closure, vendor consolidation | Finance, RevOps |
The table means that a reason code should open an investigation, not close one. “Price” is a hypothesis. It becomes a pricing finding only after the commercial record supports it.
A cancellation flow should not become an interrogation. It should collect enough structure to identify patterns, while preserving space for context the company did not anticipate.
The strongest design uses one required classification question, a small number of follow-ups, and an invitation to speak with someone who was not responsible for the original sale. Keep the survey short enough to complete in under two minutes, but make the response fields specific enough to support analysis.
Three rules keep the data credible:
Response rate alone is not sufficient. A company that receives responses from 80% of its smallest accounts may still have no evidence about the losses that matter most. Report response coverage by churned ARR and by strategic segment. If $240,000 of a $500,000 churn cohort has no usable explanation, management should label 48% of lost ARR as unknown rather than spreading the known reasons across the full population.
Monetizely’s 5-Step Pricing Framework puts pricing choices in a disciplined order: goals and segmentation; packaging; pricing metric; price points; and operationalization. The sequence matters because a price is only one part of the customer’s commercial experience. As discussed in Monetizing Agentic AI, a company first decides whom it serves and what it is trying to achieve, then builds offers, selects what it will bill for, sets the rate, and makes the model work in systems and customer-facing processes.
The third step, choosing the pricing metric, is where churn evidence becomes particularly useful. The pricing metric is the thing a customer is charged for: seats, usage, storage, transactions, locations, revenue, or another measurable unit. Churn does not select that metric by itself. Churn shows whether the existing metric is creating a recurring mismatch between customer value, buyer expectations, and the bill.
The decision table below prevents the common mistake of treating every commercial complaint as a reason to lower rates.
| Combined evidence | Likely diagnosis | Pricing action |
|---|---|---|
| “Too expensive” appears in the survey, but adoption is low and few intended users activated | The customer did not reach value | Do not cut the price. Revisit segmentation, onboarding, implementation support, or package fit |
| Customers use the product heavily, receive measurable value, and still lose to lower-priced alternatives | The price point may exceed willingness to pay in that segment | Test a lower rate, a narrower package, or a clearer value boundary |
| Customers cite invoice surprise, difficult forecasting, or a meter unrelated to their own value measure | The pricing metric creates buying or budgeting friction | Redesign the primary billing metric so the customer can predict and govern spend |
| Customers identify a missing core workflow or required integration | The offer does not meet the segment’s job to be done | Fix the product or package before changing the rate |
| The loss follows a shutdown, merger, or mandated vendor consolidation | The cause is largely outside the company’s control | Classify separately and do not use it as evidence for a price change |
The implication is demanding but practical: a company should change its billing metric only when the same type of customer repeatedly shows direct meter-related friction and the account record confirms it. A few complaints about variable invoices do not prove that usage pricing is wrong. Repeated invoice surprises among a defined segment, coupled with high value realization and low renewal rates, do.
That is also why price cuts so often disappoint. A discount can retain an account temporarily while leaving the actual problem untouched: a missing integration, a package built for the wrong buyer, or a billing unit the customer cannot forecast.
Several operator mistakes recur because they make the dashboard easier to produce, not because they make decisions better.
Twilio, Datadog, and monday.com each define retention as a cohort measure that excludes new-customer revenue while incorporating various forms of expansion, contraction, and attrition. Their disclosures underline the discipline that private SaaS companies should adopt: define the cohort before interpreting the rate.
Churn analysis earns its place in pricing strategy when it changes decisions before the next renewal cycle. A monthly dashboard is not enough. The company needs a regular operating review where finance validates the cohort, customer success brings account evidence, product assesses workflow gaps, and pricing determines whether commercial design is actually implicated.
The goal is not to explain every cancellation perfectly. The goal is to identify patterns strong enough to justify action and narrow enough to avoid expensive, broad-based price changes.
Make GRR by segment a formal gate for pricing decisions. Do not approve a broad rate change from aggregate NRR alone; require evidence from the affected customer cohort.
Create one executive owner for the churn taxonomy and data quality. That owner should have authority to reject vague reason codes, preserve definitions over time, and report unknown churned ARR openly.
Review the ten largest losses every quarter as a portfolio, not as isolated account stories. Look for repeated combinations of buyer type, package, usage pattern, and stated cause.
Require a written causal case before changing prices or billing metrics. The case should show the affected segment, GRR trend, exit-survey ARR coverage, account evidence, expected commercial effect, and the measure that will prove whether the change worked.
Track post-change cohorts for at least one full renewal cycle. A pricing change should improve retention in the intended segment without creating new discounting, adoption, or billing problems elsewhere.
Assumptions: The worked calculations use ARR as the recurring-revenue basis and treat churn, contraction, and expansion as mutually identifiable events within a fixed cohort. Public NRR figures differ in timing, revenue basis, customer definitions, and treatment of acquisitions; the disclosed range is therefore a reference point rather than a directly comparable target.
Footnotes
- https://www.amazon.com/Monetizing-Agentic-AI-Handbook-Transformation/dp/B0H7Z13VKJ/
- Journal of the Academy of Marketing Science, “Unintended Consequences of Selling B2B Digital Subscription Add-ons for Customer Onboarding,” 2025. (link.springer.com)
- Snowflake Inc., Form 10-K for the fiscal year ended January 31, 2025. (sec.gov)
- Twilio Inc., Form 10-K for the year ended December 31, 2025. (sec.gov)
- Datadog Inc., Form 10-K for the year ended December 31, 2025; monday.com Ltd., Form 20-F for the year ended December 31, 2025. (sec.gov)