The Science
Why we measure fit from behavior. Not surveys.
Conventional retention metrics mislead in predictable ways. ChurnRX is built to avoid every one of them — survival analysis, the same mathematics used in clinical trials, grounded in a real behavioral benchmark. Here's the method, and why it's defensible.
① Satisfaction isn't fit
The most-tested finding in churn research is also the most ignored.
There is essentially no correlation between how satisfied a customer says they are and whether they stay. Net Promoter Score correlates about 0.01 with customer lifespan — statistical noise. Worse for the survey worldview: customers who log support tickets tend to stay longer, not shorter. Friction is often engagement, not dissatisfaction.
What predicts retention isn't sentiment — it's whether customers achieve results. So ChurnRX measures behavior, not opinions. No surveys, no instrumentation. The signal is in what customers actually do.
Indistinguishable from zero. Satisfaction surveys do not predict who stays.
② Two populations, not one
Customers are a mixture — and one part has bonded for life.
A retention curve isn't a single line decaying at a constant rate. It's a mixture of two populations: customers who will eventually churn, and customers who have bonded — who can no longer envision success without your product (a concept drawn from MIT Sloan's Arnaldo Hax).
ChurnRX fits a mixture cure model to separate the two and estimate the bonded fraction. That fraction — the floor the curve flattens to — is your PMF Score: the cleanest possible measure of product-market fit, derived directly from observed behavior. A curve that flattens at 30% means 30% of customers will effectively never leave. A curve that falls to zero has no real fit, no matter how good this quarter looked.
③ Why churn rate lies
The rate moves with your sales. The half-life doesn't.
Churn rate has a denominator that moves with your sales, so it swings with how fast you're growing — not how well you're keeping customers. Add a big new cohort and your churn rate falls even if retention got worse. It hides the shape of churn, and it's trivially gamed.
ChurnRX uses customer half-life — the time it takes to lose half a cohort, an idea borrowed directly from radioactive-decay half-life. It captures the non-linear shape of retention, it's comparable across cohorts and companies, and sales growth can't distort it. We surface it as Value Duration.
A sales-distorted ratio. Its denominator moves with growth, so the same underlying retention can read wildly differently quarter to quarter.
A property of the survival curve itself. No sales-driven denominator — so it's stable, non-linear-aware, and genuinely comparable.
④ Survival analysis
The same math as clinical trials.
ChurnRX estimates your survival curve with the Kaplan-Meier method — the gold standard for time-to-event data with censoring, the same mathematics used in clinical outcomes research, equipment reliability, and population longevity studies. It correctly handles customers who are still active (censored), which naive averages cannot. We then fit the mixture cure model on top. The output is stable, reproducible, and immune to survey bias. It's rigor you can put in front of your board.
A survival curve that drops to zero says no permanence exists. A curve that flattens at a non-zero floor says a real, bonded fraction has formed. The level of the floor is the answer.
Your PMF Score is a structural fact about your current customer base, not a forecast. It changes as the business changes — and tracking how it changes is how you tell whether you're winning.
A single aggregate rate flattens old and new cohorts into one number that can rise while the business deteriorates. ChurnRX computes a curve and PMF Score for every cohort — PMF Momentum, the earliest honest read on durability.
⑤ A real benchmark
Grounded in 2.5M+ data points. And growing.
The methodology comes from the customer-retention research of Greg Daines, the Churn Doctor, across millions of real customer outcomes — and your results are benchmarked against an anonymized corpus computed the identical way for every company. Not opinion. Not best-practice folklore. Measurement.
A behavioral retention benchmark built from millions of real customer outcomes, and growing. Every analysis ChurnRX produces is anchored to it.
Because we benchmark the shape of the survival curve — PMF Score and half-life — and not a sales-distorted churn rate, the comparison is actually valid across companies.
You see distributions and your position within them — never another company's identity or data. As the corpus grows, the read sharpens.
Why this matters
Other tools measure churn. ChurnRX explains it.
The aggregate retention dashboards every B2B SaaS company already has tell you what happened. ChurnRX tells you why, what your structural ceiling actually is, where you stand against the field, and what to do about the gap. None of that is possible without the math, the benchmark, and the science working together. That combination is the moat.
Find out who your product is really for. In two minutes.
Upload a CSV. See your PMF Score, identify your real ICP, and walk out knowing exactly what to do next.
Run your free benchmark