Acquiring a stranger costs whatever the ad market charges this month, and the price only moves in one direction. The people who already paid you and had no complaint cost almost nothing to reach. At most businesses, nobody ever reaches them.
The list nobody works
Every completed job leaves a record: a name, a phone number, what was done, and when. Three years of those records is a maintenance schedule that nobody is reading. Meanwhile the same business pays by the click for strangers who have never heard of it, and treats that spend as normal because it arrives as an invoice while the neglected list costs nothing visible at all.
The argument here is not against advertising. It is that the cheapest audience in the building is already sitting in the job system, and it is the only marketing asset that compounds — every job completed this week makes the list longer.
Lapse is defined by the trade, not the calendar
Most software ships with a default like "inactive after twelve months." That is a database setting, not a definition. The real definition of lapsed comes from the product's own interval, and every trade has one.
- The homeowner whose air conditioning was serviced in April 2024 and never since is overdue for the pre-summer check — and in West Texas, the visit has to happen before June, because nobody books maintenance in the first 100-degree week.
- A dental patient at fourteen months since the last hygiene visit is not vaguely inactive; on a six-month recall, she is two cleanings behind, and the practice's own chart says so.
- The producer who bought filters and belts ahead of planting last season and has not called ahead of this one has either bought them somewhere else or is about to.
Once lapse is defined this way, finding lapsed customers stops being a project and becomes a question the existing records can answer: everyone whose interval passed without a return visit. That query, run monthly, is the whole engine.
Consent comes before the first message
This is the section most vendors skip, and it is the one that matters most. A text that replies to someone who just called your business is transactional — responsive and expected. A "spring tune-up special" sent to a 2023 customer list is marketing, and under the TCPA, marketing texts require prior express written consent. A signature on an old work order almost certainly did not grant it.
The rules have also tightened recently. Under the FCC's 2024 consent-revocation order, most of the revised rules took effect in April 2025; one remaining piece — a single STOP revoking consent for everything the business sends — has been delayed, currently to January 2027. The posture does not wait on that deadline: STOP has to work immediately, permanently, and across everything you send. Every message has to name the business. Texts stay inside quiet hours — 8 a.m. to 9 p.m. where the recipient is, not where you are.
So the first task in any reactivation program is not writing a message. It is an audit of what consent actually exists. Going forward, a checkbox on the invoice or intake form collects written consent cheaply, one job at a time. Looking backward, the honest answer is that much of the old list can be reached by email under different rules, or by a phone call or a postcard — but not by marketing text. None of this is legal advice, and a practice or firm in a regulated trade should put the question to its own counsel. But a program built to respect these rules from the first message costs far less than one retrofitted after a complaint, and a vendor who waves the question off is telling you how they handle everything else.
The message names the service, not the sentiment
"We miss you" is noise, and customers correctly read it as noise. The record supplies something better: the last service date and the work performed, pulled into the message itself. Your system's last service was April 2024; we're booking pre-summer checks for the next two weeks. One concrete next step — a booking link, or "reply YES and we'll call you."
Then restraint. One message, one follow-up spaced weeks later, and stop. The list's entire value depends on not burning it, and the fastest way to burn it is to treat consent as a license to be constant.
Count return visits, not open rates
Open rates and click rates measure the message. The business runs on whether the customer came back. So the measurement is set up before the first message goes out: freeze the list of who will be contacted, and ninety days later, count how many of those names booked. If the list is large enough, hold back a slice, contact the rest, and compare the two groups — the difference is the program's effect, measured in the units the owner already uses for everything else: jobs on the board and money in the account.
This design also tells you when the program is not working, which is worth exactly as much as finding out it is. A reactivation effort that cannot be measured this way is a subscription, not a system.
When this is the wrong tool
A trade with a twenty-year purchase cycle has little to reactivate — a roof does not come due the way a cleaning does, and pretending otherwise just generates unwanted mail. Thin records are a harder stop than people expect: if the job history is first names and dead numbers, the first project is record-keeping at the counter, not outreach. Customers who left over a bad experience are not lapsed, and a cheerful reminder lands on them badly; reactivation cannot repair the reason someone quit calling. And a list without text consent is reachable only through slower channels. That is a constraint to respect, not a problem to engineer around.
Where to start
Export the customer list with last-visit dates, and count how many are past your trade's interval. That one number — produced before any message is written, before any tool is bought — tells you the size of the asset, and whether the consent work that has to come first is worth doing.