Half your subscribers are gone by the 4th charge

Happy Wednesday!

I opened a CPG subscription brand's account last week and ran a survival curve nobody on the team had ever modeled. Of every 100 subscribers, about half are gone by the fourth charge. The discount everyone reaches for to save them answers the wrong question entirely.

Here is what I mean by the curve. Most operators watch the cancel rate, the single percentage that shows up on the dashboard each month, and almost nobody charts where people actually drop off charge by charge. So I built it out: of every 100 subscribers, roughly 83 make it to the second charge, 66 to the third, and 53 to the fourth.

That last number is the one that stops people. Just over half survive to the fourth charge, which means almost half are gone before it. When I looked at where inside that span the bleeding happened, about 80% of all the churn landed before the third charge.

Where they actually leave

The instinct is to blame month one, the trial period, the first impression. But the worst single window wasn't month one. It was the stretch of days right before the third charge, in the back half of month two, when the next box was about to ship and the renewal decision was quietly getting made.

I went to read why people were leaving, and the reasons were sitting right there in the cancellation event data, one tagged reason per cancel. They weren't a mystery. The brand had been collecting them the whole time and nobody had added them up.

About 52% came down to consumption and cadence, some version of "this is more than I need" or "I'm not using it anymore." Another 19% were accidental signups, people who never meant to be on a recurring plan. Price was only about 14%, and 4% was involuntary, the failed payments.

That top reason changes the whole fix. The single biggest cause of churn was that the boxes piled up faster than people could use them, not that anyone disliked the product. Almost nobody was leaving because the product was bad, they were buried in it.

The discount reflex

So watch what most CPG brands do the moment a cancellation comes in. The reflex win-back fires and blasts everyone the same offer, 20% off to come back, as if price were the thing pushing people out the door.

But that answers the wrong problem. Half the cancels just said "I have too much already," and a discount on a too-much-product cancel only ships them more of it, cheaper. The brand is reacting to the cancellation with the one offer that doesn't fit the reason the customer gave.

Read the reason, rescue the pain

The fix is to read the reason and answer the specific pain instead of defaulting to a coupon. The reason is sitting in the cancellation event itself, so you trigger the recovery flow off that event and split the flow on the reason they gave. Too much product isn't a price problem, so the rescue is a skip, a stretch, or a pause, not a discount.

Better still, you don't wait for the cancel at all. If someone is buried in product, you remind them to skip the next cycle before the charge ever hits.

Here is where most brands flinch. They go dark in the days right before a subscription bill because they are afraid that emailing will only remind the customer to cancel. That fear is the mistake.

Silence right before the charge isn't safety, it's just being absent at the exact moment the decision gets made. A quiet charge the customer didn't want is how you lose them, and a skip reminder is how you keep them. Industry data puts it at roughly a quarter of would-be cancelers taking a pause when somebody actually offers them one.

The 30/60/90 build

One pre-charge save flow is the idea, and the build runs that same logic across the whole lifecycle so churn comes down on every cycle, not just one. It maps cleanly onto the first ninety days.

The first cycle, days zero to thirty, is activation. You teach people how and when to actually use the product so the box never piles up, hand them cadence control, and catch the accidental signups early. That pre-empts the consumption churn at its root, before it ever has a chance to start.

The second cycle, days thirty to sixty, covers that worst window. It fires in the days before the third charge with a simple skip-or-stretch choice instead of just letting another box ship, meeting the renewal decision head-on at the exact point the curve says people leave.

The third cycle, sixty to ninety and beyond, runs the same pre-charge mechanic from the fourth charge onward, where price starts to surface. It runs automatically for every later charge, so the save isn't a one-time gesture, it's a standing part of the lifecycle.

Underneath those three sit the supporting flows. Dunning recovers the failed-payment churn at close to full margin, a cancel-save fires within the hour to offer a pause instead of a goodbye or a one-tap reactivation, and a paced win-back runs as a real sequence rather than the single email that reaches almost nobody.

The proof loop

The part that makes all of this defensible is the proof loop. Every save flow runs against a holdout, a slice of subscribers who don't get it, so when survival improves we can say the flow caused the lift instead of just taking credit for people who would have stayed anyway. Nobody can honestly claim a save without that control, and most teams skip it.

Then the cohort survival model re-runs every month as a scorecard. Build against the data, measure against a holdout, feed the result back into the next cycle. That's the whole loop, and it's the difference between guessing at retention and actually running it.

The discount was never going to fix this, because the discount was answering a question almost nobody was asking. The curve told us where people leave, the cancel data told us why, and the build just meets them at both.

Talk soon,

- Raymond