How do you measure churn when customers do not subscribe to anything?
You define it by silence. Pick a window based on your natural purchase cycle — often eighteen to twenty-four months for a household trade — and count customers with no job inside that window as lapsed. It is a modeling choice rather than an observed fact, so choose one window, write down why, and keep it stable. A definition that moves makes the trend meaningless.
There is no cancel button, so you have to invent one
Subscription businesses observe churn directly. Service businesses have to infer it. The customer does not tell you they left; they simply stop appearing in your job records, and the only signal is elapsed time.
That makes the lapse window a definition you own. It should come from your own data — look at the distribution of gaps between consecutive jobs for repeat customers and pick a point well out in the tail, past where most genuine repeat purchases occur.
Different trades need different windows
A window that fits one trade misdescribes another, and multi-trade operators need more than one.
If you run several trades under one roof, computing a single company-wide churn number blends these cycles into something that describes none of them. Split by the trade of the customer’s most recent job.
- Frequent, small-ticket work — lawn, pest, pool, cleaning — has short cycles, so a lapse is visible within a season.
- HVAC and plumbing service sit in the middle: annual maintenance sets a natural rhythm, and missing two of them is meaningful.
- Roofing, repipe and electrical panel work may have cycles measured in decades, where lapse is almost meaningless and referral behavior matters more than repeat purchase.
Watch reactivation, not just churn
Because there is no cancellation, lapse is reversible. Customers come back. A churn number that ignores returns overstates loss and hides one of the cheapest revenue sources you have.
Report three quantities together: lapsed this period, reactivated this period, and net active customers. The reactivation figure is the one that tells you whether dormant-list work is paying off, and it is directly connected to the segmentation work that decides who gets contacted at all.
Beware the artifacts
Two things regularly produce fake churn. Duplicate records split one customer’s history in half, so both halves look dormant. And customers who moved out of your service area are counted as churn when they are simply gone.
There is also a reporting trap at the edges: if your lapse window is twenty-four months and your data only goes back thirty, your earliest cohorts have almost no room to demonstrate retention and will look artificially bad. Any analysis of retention should state its window, its data horizon, and what it excluded — otherwise the trend is measuring your data collection rather than your customers.
Topics: churn · retention · lapsed customers · definitions
Have a version of this question about your own business?
The useful answer usually depends on which systems you run and how they're connected. That's a conversation, not a blog post.