The reflex that costs you money
Someone is about to leave, so you send them 20% off. It feels like action. It usually isn’t. A discount treats the symptom — a customer heading for the door — while ignoring the cause that put them there. Sometimes the cause is price. Far more often it’s a broken onboarding, a feature they never found, a support ticket that went cold, or a need that has simply moved on.
And there’s a second, slower cost. Every time you rescue a wobble with a coupon, you teach your best customers a lesson: hesitate, and you get paid. Do that at scale and you’ve built a machine that manufactures the exact behaviour you’re trying to kill. Indiscriminate discounting is a margin crime — and the customer learns the loophole faster than your finance team can model it.
Start with why people actually leave
Churn is not one thing. It’s a folder of very different stories wearing the same label. The customer who outgrew you leaves for different reasons than the one who never activated, who leaves for different reasons than the one genuinely shopping on price. Pour the same discount on all three and you overpay two of them to do something they’d have done anyway.
So the first move isn’t an offer — it’s a question. Segment by reason for leaving, not by tenure or spend. Once you can name the why, the right lever becomes obvious, and most of the time it costs nothing.
Every coupon you throw at a wobble teaches your best customers that hesitating pays.
Predict the risk, then reach for the right lever
Treat retention the way a data team treats any prediction problem. Score who is drifting before they announce it — declining frequency, slipping engagement, the quiet signals that precede a cancellation by weeks. Then act early, while you still have options that don’t involve your P&L.
Most of those options aren’t price at all. A nudge to the feature that drives stickiness. A proactive fix to the problem they were about to complain about. A check-in from a human. A reminder of value already delivered. These are the first three levers you should pull, in order, before money ever enters the conversation.
The 2×2 I actually use
Once you can score both churn risk and customer value, the right play falls out of a simple grid. Risk on one axis, value (LTV) on the other — four cells, four very different moves:
High value · High risk
Save at all costs
Your best customers, heading out. Full-court press: a human reaches out, the real problem gets fixed, and price is the last resort — not the first.
High value · Low risk
Reward, don’t discount
Loyal and valuable. Reward them — never discount them. They’d have stayed anyway, so a coupon here is pure margin given away.
Low value · High risk
Price match — or let go
Price-driven and low value. A measured, time-bound offer — or a graceful goodbye. Don’t spend more than they’re worth.
Low value · Low risk
Nurture
Stable but small. The job here isn’t retention — it’s growing the value before churn is even the question.
Two rules keep it honest. Don’t incentivize anyone already above roughly 80% likely to buy — they convert without the gift. And cap retention spend at a share of predicted LTV, agreed with Finance, or you slide into the death spiral of over-discounting: bigger offers, thinner margin. The worst move is a discount to a valuable customer who left over service, not price — it burns margin and insults them. A double failure.
Reserve price for where it changes behaviour
This isn’t anti-discount. Sometimes price is the genuine objection, and a targeted, time-bound offer to a customer who is truly price-sensitive is exactly right. The discipline is in the word targeted. A price move should be the scalpel you use on the one segment it actually moves — not the bucket of water you throw at everyone whose graph dipped.
Protect contribution margin like it’s the point, because it is. Retention that costs more than the customer is worth isn’t retention — it’s a subsidy you forgot to expense. Predict the risk, diagnose the cause, intervene with the cheapest lever that works, and spend money only where it changes what the customer does. That’s the difference between buying loyalty and earning it.