Split payment churn from intent churn, recover the payments first, then treat intent churn by cause with uplift-ranked actions and a permanent holdout.
To reduce churn, work in this order: split payment-failure churn from intent churn, fix the payment side first because it costs no margin, then attack intent churn by cause with treatments ranked on incremental effect and measured against a preserved holdout. Teams that follow that order usually find half their reported churn was never a retention problem at all.
What follows is the sequence, not a list of tactics. Order matters more than ingenuity here: each step changes the baseline the next step is measured against.
A single churn rate hides three different businesses: customers who cancelled deliberately, customers whose payment failed, and customers who were recovered after a failure. Tag every cancellation at source with its origin and, for declines, the decline code. Until that exists, every retention decision is made blind. The distinction is developed in voluntary vs involuntary churn.
Risk scores rank; causes prescribe. Four causes cover most bases: value not reached (the customer never got to the outcome they bought), value outgrown (the plan no longer fits), friction (a support or product failure), and price sensitivity (an external shock or a competitor offer). Each has a different eligible treatment set, and mixing them is why generic save campaigns underperform.
Cancellation flows are the last line and worth designing well, but by then the decision is largely made and the surviving levers are expensive. The cheap window is weeks earlier, when engagement first decays. Move budget from the exit to the drift, and keep the deflection flow for the residual. Flow design is covered in churn deflection and cancellation flows.
For each at-risk customer, several treatments are eligible: an onboarding nudge, a plan right-size, a service credit, a human call, a discount. Rank them by expected uplift multiplied by the margin at stake, minus the cost of the treatment. This is arbitration, and it is the step that turns retention from a spend line into an investment with a return.
Five to fifteen percent of equally-scored customers receive nothing, per treatment, forever. It feels like leaving money on the table and it is the only way to know whether there was money on the table at all. Read incremental save rate weekly by cohort and cause, and put a kill rule in writing before launch.
Retention treatments buy time; product fixes remove the cause. When a cohort's churn traces to a specific friction point, the durable win is the fix, and the save offer is the interim. Route the cause evidence to the product team with the size of the revenue at stake attached, or it will not be prioritised.
Do not launch a base-wide discount to protect a quarterly number: it trains the base to wait for one. Do not treat everyone in the top risk decile: a share of them are made more likely to leave by the contact. Do not report saves without a control group. And do not let the churn model live in a notebook that nobody has permission to act on.
Markin runs this sequence continuously: signal, revenue opportunity, hypothesis, candidate action, experiment. See Retention decisioning or the telco view.
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
