Churn is the quietest way to lose money, because nothing happens. This guide is for whoever owns retention. It covers what churn means for a business without subscriptions, how the at risk list is built, and how to judge whether a win back campaign is worth running.
Churn without a subscription
For a subscription business churn is a clean event. Somebody cancels, and there is a date.
For most ecommerce brands there is no such event. A customer who has not bought for four months has not resigned. They may simply not need anything yet.
So churn here is defined against your own reorder behaviour rather than against a fixed window. If your typical customer reorders every six weeks, somebody at fourteen weeks is behaving unusually. If your product lasts six months, the same gap means nothing. That is why the churn view sits next to reorder timing, and why the two should be read together rather than separately.
The at risk list
At risk customers are those whose gap since last order has stretched well past what their own history suggests it should be.
The important part is that this is measured per customer rather than against a single business wide threshold. A customer who has always bought monthly and is now at ten weeks is a stronger signal than a customer who has always bought twice a year and is at four months.
Treat the list as a queue rather than a report. Its value is that it identifies people while intervention is still cheap. A customer who lapsed last month can often be recovered with a reminder. A customer who lapsed a year ago usually needs a discount deep enough that recovering them is barely worth it.
Judging a win back campaign
Win back campaigns almost always look successful, because you send them to people who have bought before and some of them were going to buy anyway.
The number that matters is incremental revenue, not attributed revenue. The honest way to find it is to hold out a portion of the audience and send them nothing, then compare. If the holdout group buys at nearly the same rate, the campaign is taking credit for customers who were returning regardless, and the discount you attached to it is a cost with no matching benefit.
Tatheon reports win back performance including cost, so a campaign that recovers revenue at a discount deep enough to lose money shows as what it is. That is worth checking before repeating a campaign that reported well.
Cancellation reasons
Where they are captured, cancellation and churn reasons are the most actionable retention data you have, because they point at a fixable cause rather than a number.
Reasons cluster. Price appearing repeatedly is a positioning problem. Delivery time appearing repeatedly is an operations problem that is being paid for in retention. Product quality appearing repeatedly is worth more attention than any campaign, because no win back sequence survives a product people did not like.
What to do first
Start with reorder timing rather than with the churn number. Knowing your typical gap tells you when an intervention is early enough to be cheap. Then work the at risk queue while it is short. Then measure any win back campaign against a holdout, because without one you are measuring your customers’ own habits and calling it a result.