How to Decide Which Lapsed Customers to Call First

Why this matters

Every shop that pulls a lapsed list discovers the same thing: the list is bigger than the time available to work it, by a lot. What happens next decides whether the exercise pays. Most shops sort by longest-lapsed first, because it feels like the customers most at risk deserve the most urgency. That ordering is close to the worst one available: the longest-lapsed records have the worst contact data, the weakest memory of you, and the highest chance of already being somebody else's customer.

This is not about who is worth keeping in general, and it is not the cleanup procedure that produces the list. It is the narrower question of call order when the list is 200 names and you have room for 60 attempts, and the ordering rule that gets the most bookings out of a fixed number of phone calls.

Step 1: Size the call budget in attempts, before you look at a single name

Decide how many attempts the office can actually make before you rank anything. Ranking without a budget produces a beautiful list nobody finishes, and an abandoned list is worse than no list because it teaches the office that outreach never gets done.

Work it out in real units. Count the hours per week your office lead can genuinely protect for outreach, allow about 12 minutes per attempt including the dial, the conversation or the voicemail, and the note on the record, and decide how many weeks the pass runs. Twelve minutes is a working default, not a law - time your own first ten and adjust.

Then convert attempts into records. Most reachable people take more than one try, so divide your total attempts by your average attempts per record to get the number of records the pass can cover. Cap attempts per record at two or three, and stop; a fourth call to somebody who has not picked up is a call you owe a different customer.

Step 2: Strip the list to callable records before ranking anything

Ranking a record you cannot reach is wasted work, so this happens first, not as a tie-breaker. Remove from the ranking, without exception:

• No working phone number or a returned address.
• No named person, only a company line that goes to a switchboard with nobody attached to your history.
• Any do-not-contact request, ever, regardless of age.
• Any open dispute, unresolved complaint, or unpaid balance in collections. These need a different conversation from a different person and must never appear in a routine outreach batch.
• Any record where the property is known to have sold. That is a new-customer approach at a known address, not a lapsed-customer call.

Removing these is not tidiness. Each one is a call that would have gone badly, and one of them landing in a batch is how an outreach pass produces a complaint instead of a booking.

Step 3: Rank on three ordered gates, not a blended score

A blended score creates ties, and ties get re-litigated by whoever is holding the phone. Ordered gates produce a strict sequence that anybody can reproduce.

Gate 1: overdue relative to their own interval, not in absolute months. Compute the overdue ratio: months since their last completed job, divided by their own median interval (or the service-type default if they have too little history). A customer 15 months out on a 12-month interval has a ratio of 1.25. A customer 15 months out on a 36-month interval has a ratio of 0.42 and is not late at all. Absolute months is the single most common ranking error and it fills the top of the list with people who are exactly on schedule.

Gate 2: the band the ratio falls in. Rank the bands in this order:

Gate 3: within a band, a live reason to call outranks no reason. A live reason is something on the record you can name in the first sentence: a recommendation they deferred, a component your tech flagged as near end of life, equipment past the age where the failure mode you have seen shows up, a second unit never serviced. A record with a live reason converts better and, just as importantly, gives the caller a first line that is about the customer rather than about your slow week.

Step 4: Fixed tie-breakers, applied in this order

Inside a band, among records with equally live reasons, break ties in a fixed published order so nobody improvises:

1. More completed jobs on record beats fewer. Proven repeat behavior is the best predictor you own.
2. Shorter natural interval beats longer. They come back more often, so a save is worth more per year.
3. Clean payment history beats a record with slow-pay notes.
4. Geographic clustering with work already on the calendar beats an isolated address, because the booking is cheaper to serve.

Publish the order. The point of a fixed tie-break is that two different people build the same list.

Step 5: The two records that jump the whole queue

A safety item flagged and never resolved. If a tech documented a hazard condition at the last visit - a gas-appliance combustion concern, water intrusion near electrical equipment, a stored-energy component like a loaded spring or an undischarged capacitor left accessible, a fall exposure created by removed guarding - that record is contacted today, ahead of everything, and the call is not an outreach call. Tell the customer plainly what was found, tell them to stop using the equipment and isolate it at the appropriate shutoff or disconnect, and get it scheduled. Never let a documented hazard sit in a ranked list waiting for its band to come up.

A customer who contacted you and never got a reply. A message with no logged response is not lapsed, it is dropped. It goes first, and it is an apology call, not an outreach call.

Step 6: Work a real list

A shop pulls 214 records that are past due against their own intervals.

Budget. The office lead protects 5 hours a week for four weeks. At 12 minutes per attempt that is 25 attempts a week, so 100 attempts in the pass. With an average of about 1.7 attempts per record at a two-attempt cap, the pass covers roughly 60 records. Sixty is the real number; 214 was never the plan.

Strip. Forty-one records fail step 2 on bad contact data, a sold property, or an open balance. That leaves 173 callable records to rank, so the strip removed 19% of the raw list before any ranking happened.

Band. Sorting the 173 by overdue ratio gives 58 in the 1.2 to 2.0 band, 62 in the 2.0 to 3.0 band, and 53 above 3.0. Those sum to 173.

Live reason. Inside the top band of 58, records carrying a deferred recommendation or a flagged component number 31; the other 27 have nothing specific on file.

The order, then. First the 31, then the 27, which is 58 records. The 60-record budget covers the entire top band plus the first two records of the next band's live-reason subset. Everything else waits for the next pass, and that is fine, because the next pass will re-rank and some of those records will have moved up a band on their own.

Result. Of the 60 records worked, 39 reached an actual person, a 65% reach rate on records worked. Fourteen booked paid work, which is 23% of the 60 records worked and 36% of the 39 conversations. State both bases every time, because those two numbers get quoted interchangeably and they are not the same measurement.

The check that justifies the ordering. The shop also worked 20 records pulled from the above-3.0 band in the same pass, as a comparison. Nine reached a person, 45% of those 20, and 2 booked, 10% of those 20. Against 65% reach and 23% booking in the top band, the direction is clear and it matches what the ordering rule predicted. Twenty records is a small sample and one unusual customer moves it several points, so treat this as directional evidence supporting the rule rather than a measured multiple. Run it again next pass and the pattern either holds or it does not.

What that comparison rules out is the intuition the shop started with. Working longest-lapsed first would have spent the entire 60-record budget in the band that produced 2 bookings per 20 records, and the pass would have been declared a failure and never repeated.

Step 7: Work the order you built, and do not re-sort mid-pass

Once the pass starts, the list is frozen. Re-sorting halfway through, usually because somebody remembered a customer they liked, destroys the only thing that makes the result readable: you no longer know what order produced what outcome. Add new arrivals to the next pass, not this one.

The exception is step 5. A hazard flag or a dropped message discovered mid-pass goes to the front immediately, and gets noted as an exception so it does not pollute the pass statistics.

Step 8: Re-rank on outcome, not on feel

At the end of the pass, record for each band: records worked, people reached, jobs booked. Those three numbers by band are the entire input to the next ranking. If your 1.2-to-2.0 band underperforms the 2.0-to-3.0 band two passes running, your interval defaults are probably wrong and people in the top band are not actually late - go back and recheck how the intervals were derived before you change the ordering rule.

What changes the answer

• Long-interval trades. Where the natural interval runs several years, the top band contains very few people at any moment and the ordering matters less than the property record. Rank on equipment age and known deferred work instead of on the overdue ratio.
• You have idle capacity next week specifically. Then geographic clustering moves from tie-breaker to a primary gate, because a booking you can serve on an existing route is worth more than one across town.
• The lapse has a known common cause. After a tech departure or a rate change, the affected segment is not a ranking problem, it is a targeted list, and it gets worked as a group with a specific message rather than folded into a general pass.
• The list is small enough to work entirely. Under about 40 callable records, skip the ranking and call everyone. Ordering exists to allocate scarcity.

How to verify you got this right

• The overdue ratio was computed against each customer's own interval. If a single shop-wide month count was used, the ranking is close to random for anyone off the average.
• The strip in step 2 ran before ranking, and no record with an open balance, a dispute, or a do-not-contact request appears anywhere in the worked list.
• Reach rate and booking rate are each reported with their base named, and nobody has quoted the conversation-based conversion as if it were the record-based one.
• The pass ended at the budgeted number of records rather than trailing off. A pass that stopped at 34 of a planned 60 has a capacity problem, not a list problem, and the next budget should be honest about it.

References

• See related: The Dormant List Audit SOP (producing and cleaning the list this ranks)
• See related: The Customer Reactivation SOP (running the contact cycle)
• See related: When to Stop Chasing a Dormant Customer
• See related: How to Use Service History to Predict the Next Call (deriving the intervals the ratio depends on)