How do we tell a customer we lost from one who simply is not due yet?
Compare time since last job against that customer's own expected interval rather than against the calendar. A maintenance member is late at thirteen months; a water heater customer is not late at three years. Compute an expected interval per job type from your completed-job history, then flag customers who are past a multiple of it. Any fixed cutoff labels normal customers as churned and buries the real losses.
Interval is a property of the work, not the customer
Different job types have completely different natural rhythms. Seasonal maintenance repeats annually or twice a year. Drain service repeats irregularly. Equipment replacement repeats on a scale of a decade or more.
A single company-wide definition of lapsed averages all of that together and produces a threshold that is too short for half your customers and too long for the other half. The list it generates is mostly noise, people stop trusting it, and it quietly stops being used.
Compute the interval from your own gaps
For each job type, take customers with at least two completed jobs and measure the gap between consecutive jobs. Use the median rather than the mean, because the distribution has a long right tail and the mean will be dragged well past reality by a handful of customers who returned after six years.
You now have a per-job-type expected interval derived from your actual customers in your actual market, which is worth more than any industry rule of thumb because it reflects your climate, your equipment mix and your pricing.
Flag on a percentile, not the median
Half your customers are past the median by definition, so flagging at the median flags half your base. Use an upper percentile of the gap distribution instead, for example the point by which three quarters of returning customers have come back.
Past that point, the customer is genuinely unusual relative to your own history, and that is what makes the flag actionable. Reviewing the threshold once a year keeps it honest as your job mix shifts.
One-job customers are a different problem
A customer with a single completed job has no interval of their own, and applying the job-type average to them is a guess. They are not lapsed; they never became repeat customers in the first place.
Handle them separately. The question for them is why the second job never happened, and the answer is usually visible in what happened on the first: a callback, a pricing conversation that went badly, a technician who did not offer anything else. That analysis needs the job record and the call record joined together, which is the sort of thing customer intelligence exists to do, and it points at very different action than a reactivation offer. Feeding both lists into the same campaign is a common and expensive mistake in home services operations.
Topics: churn · service interval · segmentation · customer data
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.