Methodology
How RetentionCheck calculates the 0-100 Churn Health Score and letter grade shown on every analysis, and where the numbers on our churn benchmark pages come from.
The formula
Every analysis starts at 100 points. For each churn insight identified in the cancellation corpus, we deduct points based on severity:
Score is floored at 0. The resulting 0-100 number is mapped to a letter grade.
Grade bands
Sorted from best to worst. Each band includes what it means, and how realistic it is to achieve in practice.
Rare. Reserved for analyses with almost no high or critical severity drivers. A+ typically requires a product with only low-severity churn drivers in its cancellation corpus.
Unusual. Most SaaS products do not score A+ because paying customers always have some material criticism.
Excellent retention health. 1-2 medium severity drivers at most, rest low. Indicates a product with minor fixable issues.
Possible for beloved products with primarily pricing or nice-to-have feature complaints.
Good retention health. Mixed medium severity with one high. Typical for well-run SaaS products at scale.
Normal for mature products with known pain points that are actively worked on.
Middling retention health. Several high severity drivers present. Typical for mature SaaS with unresolved pain points.
Common. Most enterprise SaaS with multiyear roadmaps land here.
Concerning retention health. Multiple high severity drivers or one critical plus mediums. Signals active churn risk.
Common after pricing events, feature removals, or customer success staffing cuts.
Failing retention health. Multiple critical severity drivers. Indicates active mass-churn conditions that need immediate attention.
Rare. Usually triggered by a specific event (acquisition change, major price hike, feature removal, data breach).
Live grade distribution
Out of 649 analyses run through RetentionCheck to date. Refreshed hourly.
Sample size is displayed on every analysis. A D grade backed by 500 cancellations is more reliable than a B grade backed by 20. Read the count before you read the letter.
These are analyses run by real users. Most land in C-D because most cancellation corpora contain real criticism.
Why A+ is rare
Every product with paying customers has real criticism. Even beloved products generate medium or high severity churn drivers in their cancellation data because users who cancel are, by definition, choosing to leave over something.
A+ requires a cancellation corpus where almost every driver is low severity. In practice this happens only when cancellations are largely driven by non-product factors (business closure, role change, one-off budget cut) rather than product pain.
The grade itself is honest. If every SaaS scored A+, the score would not mean anything. Most analyses land in B, C, or D because most products have mixed feedback. That is the point.
What the score does NOT measure
- •Actual churn rate. The score analyzes cancellation reasons, not customer lifecycle data. You can have high churn with a B and low churn with a D.
- •Revenue impact. We do not estimate dollars lost unless you provide revenue data. Our insights are based on customer counts and percentages from the cancellation corpus.
- •Product quality absolute. The score reflects what the corpus says. A small, noisy sample will produce a different score than a large, representative one. Sample size is displayed on every analysis.
How the Churn Index applies this score
The SaaS Churn Index uses the same Churn Health Score methodology (0 to 100, higher = healthier) and the same A+ through F grade bands.
The difference is input: Churn Index scores come from publicly-observable signals (G2, Trustpilot, Hacker News, Reddit, layoffs.fyi, pricing changes) rather than private cancellation feedback. Same score direction across the product.
Where the benchmark numbers come from
RetentionCheck publishes churn benchmarks for roughly 100 industry verticals at /churn-benchmarks. Here is exactly what those numbers are, and are not.
Today, every per-vertical figure on those pages (monthly churn, annual churn, median ARPU) is a RetentionCheck editorial estimate, anchored to published industry research. The figures are not measured from RetentionCheck's own dataset yet, and they are not copied row-by-row from any single study. Each benchmark page is labeled accordingly, directly under its headline numbers. Where published research brackets a vertical's estimate, the page lists those citations inline as published context, directly under the estimate label.
The published research we anchor against:
SaaS Capital: 2023 B2B SaaS Retention Benchmarks (2023)
Annual survey of roughly 1,500 private B2B SaaS companies. Publishes net revenue retention (NRR) and gross revenue retention (GRR) by contract-value band, measured annually: median NRR 102% and median GRR 91% across all companies. No per-industry vertical breakdown.
ChartMogul: SaaS Benchmarks Report 2023 and SaaS Retention Report 2023 (2023) [retention report]
Behavioral data from companies using ChartMogul's subscription analytics. Publishes median monthly customer churn by ARR range (6.5% under $300k ARR, 3.7% at $1M-$3M, 3.1% above $8M) plus NRR and GRR by ARR and ARPA band. Segmented by company size, not industry vertical.
Recurly: 2024 State of Subscriptions (2024)
Data from 2,200+ merchants and 67 million subscribers. The only major primary source that names industry verticals: Education 4.2% voluntary monthly churn, Software 2.2%, Digital media and entertainment 6.9%, Digital publishing 3.9%. All-industry median combined churn: 4.0% monthly.
Antenna: State of Subscriptions (Premium SVOD) (2024)
US consumer panel covering the major streaming services. Publishes gross and net monthly subscriber churn per service and weighted averages: premium SVOD weighted average gross monthly churn of 5.3% in September 2024, net 3.1%. Consumer streaming only.
KeyBanc Capital Markets + Sapphire Ventures: Private SaaS Company Survey (15th annual) (2024)
Survey of 100+ privately held SaaS companies. Publishes aggregate annual retention: median GRR around 90% and median NRR around 101%. Surveys named verticals like fintech and cybersecurity but does not publish per-vertical retention breakdowns.
None of these studies publishes per-vertical SaaS churn for most of the verticals we cover: there is no primary public study for fintech SaaS churn, healthcare SaaS churn, developer tools churn, and most of the rest. That gap is why the per-vertical figures are estimates. We position each vertical inside the ranges the research above supports, using the structural factors that drive retention (contract value, buyer type, switching costs), and where a study does name a vertical, our estimate is anchored to it.
Definitions used on benchmark pages
Monthly logo churn
Customers lost during a month divided by customers at the start of that month. Counts accounts, not dollars. This is the basis of the per-vertical monthly figure on every benchmark page.
Revenue churn (MRR churn)
Recurring revenue lost during a period divided by recurring revenue at the start of it. Diverges from logo churn whenever large and small accounts cancel at different rates. Studies that report NRR and GRR are measuring revenue retention, not logo churn.
Voluntary vs involuntary churn
Voluntary churn is a customer-initiated cancellation. Involuntary churn is a payment failure: declined cards, expired cards, failed renewals. Published vertical figures, including Recurly's, often report voluntary churn only.
Annualization
Annual churn = 1 - (1 - monthly churn)^12. Monthly churn compounds; it does not simply multiply by 12, because each month's losses come out of an already smaller base.
The plan: replace estimates with measured data
Every analysis run through RetentionCheck carries vertical context. As real, anonymized analyses accrue for a vertical, the editorial estimate on that vertical's page gets replaced with the measured figure from our own dataset.
When that happens, the page's label flips from "editorial estimate" to "sourced" and the citations backing the headline numbers render directly under them. No vertical gets marked sourced until its numbers come straight from cited data.
Updates and corrections
Each benchmark page carries its own updated date. We review the source list above when the underlying reports publish new editions, which is annual for most of them.
Think a number is wrong? Email hello@retentioncheck.com with the vertical and the source you would cite instead. If your source is stronger than ours, the page changes.
Ready to score your own cancellation data?
Run a free analysis