The Customer Lifecycle Stages a Service Shop Actually Has

Why this matters

Every shop already sorts customers into stages. It just does it in someone's head, differently every week, and usually only when the phone rings. That works until the list passes a few hundred names, at which point the shop can no longer tell the difference between a customer who is quiet because they are fine and a customer who is quiet because they left.

A stage model fixes that by turning a feeling into a test. This article is about the multi-year states a customer moves through with your shop, each defined by something you can check in the service history in under a minute. It is not a journey map. A journey map traces the touchpoints inside one purchase decision, from the search to the invoice, and it is a useful and different tool. Stages are where a customer sits across years, and the whole point is that you can compute them from dates you already have.

A stage is a state you can test, not a feeling

Three properties make a stage model usable:

• The entry test is objective. It reads job count, job dates, and nothing else. If two people in the office would sort the same customer differently, the definition is not finished.
• Each stage changes what the shop does. A stage that produces the same action as the one next to it should be merged. Categories that do not change behavior are decoration.
• Movement is dated. You should be able to say when a customer entered a stage, not just that they are in it. That date is what makes intervention possible.

The dates you need are already on your invoices. A shop with no software and a shoebox of paper copies can build this; a shop with software can build it in an afternoon.

The seven stages

Prospect. They have contacted you or you have contacted them, and no work has been performed. A quote sits here. So does a call that never converted. The trap is counting quoted-but-not-sold names as customers, which inflates every retention number you calculate afterward.

First-time. Exactly one completed job, ever. This is the highest-risk stage in the whole model and the shortest. Most shops lose more customers here than anywhere else, and lose them silently, because a first-timer who never returns produces no complaint and no cancellation.

Repeat. Two completed jobs. The second job is the single most meaningful event in the relationship, because it is the first evidence that the customer chose you rather than found you. Nothing else in the model carries as much information per event.

Established. Three or more completed jobs, with the most recent one inside their own normal interval. This is where a personal service interval becomes measurable and where the shop finally has something to plan around.

Slipping. They have a history, and the gap since the last job has passed their own normal interval by a meaningful multiple, but has not yet reached your dormancy line. This is the only stage that exists purely to create a window for action, and it is the one most shops do not have.

Dormant. The gap has crossed the threshold where you stop assuming they are simply between jobs. The threshold is not universal and should be derived from your own interval data rather than borrowed.

Lost. Dormant plus a reason: a stated departure, a known move, a property sale, a competitor sign in the yard, or a documented reactivation attempt that got a clear no. Lost is not a longer version of dormant. It is dormant with information attached, and that difference is what keeps you from working a dead list forever.

The stage table

The two clocks

Everything in the table above runs on one of two clocks, and confusing them causes most misclassification.

The interval clock is the customer's own natural rhythm: how long they typically go between jobs. It is a property of their equipment, their property, their habits, and their trade mix. A customer on a twice-yearly seasonal rhythm and a customer who calls only when something breaks are on completely different clocks, and the same elapsed silence means opposite things.

The relationship clock is job count, and it only ratchets forward. A customer with nine completed jobs stays Established-by-history even during a long silence, which is why the interval clock has to be the thing that moves them to Slipping. Job count tells you how much they have invested; elapsed time tells you whether they are still investing.

Read either clock alone and you get a wrong answer. Job count alone marks a nine-job customer as your best customer three years after they switched. Elapsed time alone marks a healthy annual-service customer as at-risk every October.

Why "within their interval" has to be personal

The temptation is a single shop-wide number: everyone is fine under 12 months, everyone is dormant over 18. It is easy, and it systematically hides your best customers.

Consider the arithmetic. A customer whose natural rhythm is 6 months has to be silent for twice their normal cycle before a 12-month line notices, and three times their normal cycle before an 18-month line does. A customer whose natural rhythm is 18 months trips the same line while behaving completely normally. So a global threshold generates false alarms on your low-frequency customers and false silence on your high-frequency ones, and high-frequency customers are usually the profitable ones. You get noise where it does not matter and quiet where it does.

The fix is to compute each customer's own typical gap and measure them against that. Use the median gap, not the average. One emergency call in the middle of a planned rhythm pulls an average down hard and makes the customer look more frequent than they actually are.

Worked example: one address across four years

A shop pulls the history for a single residential address.

• Year 1, March: first call, an after-hours failure. One completed job. Stage: First-time.
• Year 1, October: the tech's flagged item, done on the named month. Two jobs. Stage: Repeat. Gap 1 is 7 months.
• Year 2, April: seasonal service. Three jobs. Stage: Established. Gap 2 is 6 months.
• Year 2, November: seasonal service. Gap 3 is 7 months.
• Year 3, February: unplanned repair. Gap 4 is 3 months.

Four gaps: 7, 6, 7, and 3 months. The average is 5.75 months, the median is 6.5 months. The average is the wrong number here and it is wrong in a specific direction: the 3-month gap was an unplanned failure, not part of the customer's rhythm, and averaging it in makes the shop expect this customer sooner than they will ever actually call. Use 6.5 months as the personal interval.

Now walk the clock forward from that February visit in Year 3.

• September of Year 3 is 7 months out. That is one interval. Normal. No action.
• December of Year 3 is 10 months out, roughly 1.5 times their 6.5-month interval. This is the Slipping entry, and it is the moment worth a personal call.
• March of Year 4 is 13 months out, exactly 2.0 times their 6.5-month interval.

Here is the part that matters. If this shop runs a global 18-month dormancy line, this customer is still sitting in the Active bucket in March of Year 4 with five completed jobs, and nobody will look at them until August of Year 4 at the earliest. By then the silence is 18 months against a 6.5-month rhythm, and whatever caused it has had a year and a half to harden into a habit with somebody else. The global rule did not fail to fire. It fired on schedule, and the schedule was wrong for this customer.

Against the personal interval, the same record raised a flag in December of Year 3, eight months earlier, while the relationship was still recent enough that a call would land as attentive rather than as a shop noticing an absence a year late.

One honest caveat on this example: five jobs and four gaps is a thin basis for a median, and a single unusual year can distort it. Treat a personal interval computed from fewer than three gaps as provisional, and fall back to the interval typical for that service type until the customer has enough history to speak for themselves.

How to verify you got this right

Every active name has exactly one stage, and you can name the date it entered. If a customer's stage cannot be stated with an entry date, the model is being applied as a label rather than as a state, and it will drift back into opinion within a month.

The Slipping bucket is not empty and not enormous. Empty means the multiple is set too high and you have effectively deleted the stage. Enormous, meaning a large share of your Established base, usually means the interval is being computed from averages that include emergency calls, so everyone looks overdue. Recompute with medians before you change the multiple.

Stage counts move between quarters. A stage distribution that is identical quarter over quarter is not being recalculated; someone is copying last quarter's sheet. Run the sort fresh, on the same day of the month, every month.

Lost has reasons in it. If your Lost bucket has no reason recorded on most rows, it is a second dormant bucket wearing a different name, and you will keep reactivating people who told you they moved.

References

• U.S. Small Business Administration (SBA), customer relationship and retention guidance for small business
• Trade-standard practice for service-interval scheduling and recall programs
• See related: Customer Journey Mapping for Service Business, The Dormant Customer Definition Worth Setting, How to Segment a Customer List by Lifecycle Stage, The Loyalty Ladder for a Service Business