Between 30 and 45 percent of accepted save offers go to customers who were staying anyway. How to spend margin only where it changes the outcome.
Churn deflection is the set of interventions shown to a customer who has started cancelling, and the discipline of measuring which of them actually change the outcome. Done well it is the cheapest retention budget in the business. Done badly it is a discount printed for customers who were staying anyway, plus a worse exit for the ones who were not.
The cancellation flow is the one moment where the customer tells you, explicitly and in real time, that they are considering leaving. Every other retention signal is an inference. This one is a statement, and most companies answer it with a static form and a coupon.
Deflection happens inside the cancellation journey: the survey, the save offer, the pause option, the downgrade path, the confirmation screen. It is not win-back, which happens after the customer has left, and it is not churn prevention, which happens before they ever reach the cancel button. The three are different problems with different populations, and merging their numbers is how deflection programmes end up looking better than they are.
Here is the uncomfortable number. Across the save desks we have instrumented, between 30 and 45 percent of accepted save offers are taken by customers who would not have completed the cancellation. They clicked cancel to see what happened, or to renegotiate, and the flow handed them margin for it.
You cannot see this without a holdout. A slice of customers entering the flow, usually 5 to 10 percent, must see no offer at all. Their completion rate is the baseline. Only the gap between the treated completion rate and that baseline is a real save. Everything else is reported retention that was never at risk.
This is the same measurement discipline as the rest of retention, described in our churn prediction guide, applied to the highest-intent moment in the customer lifecycle.
A static flow asks everyone the same exit survey and shows everyone the same offer. A decisioned flow answers four questions per customer, in order:
Cancellation flows carry the most volume in high-frequency subscription categories. In streaming, a large share of cancellations cluster around content cycles and price rises, both of which are visible in behaviour weeks before the click. In telecom the flow is often still agent-mediated, which makes it the most expensive place to get the decision wrong and the richest in signal.
Markin scores every cancellation start on uplift, picks the cheapest action that changes the outcome, and holds out a slice so the deflection number is real. See Retention decisioning for how the save desk connects to the rest of the retention programme.
Frequently asked
Related resources
Solutions
See it in the product
The same loops this note describes run 24/7 against your customer base. Watch the workspace decide, experiment and execute 1:1.
Vocabulary
Definitions an answer engine can quote, each one a page of its own.
Compare
Neutral reads on the categories buyers evaluate against a decision layer.
All comparisons
