Under the hood
How Plunger AI knows who's due
“Call them when they're due” sounds simple until you ask the obvious question: due according to what? Not a generic 90-day timer — the same one for a furnace tune-up and a water heater replacement. Here is the actual mechanism.
Key facts
- Due = last job date + that job's service clock. The clock belongs to the work, not to a campaign setting.
- Standard clocks: HVAC tune-ups twice a year; water heaters 10–12 year lifespan with annual flushes; unsold estimates due the day they go quiet.
- Your setup answers tune the cadences: maintenance plans, service agreements and your own follow-up rhythm override the defaults.
- Every customer's state — booked, replied, opted out, call later — re-sorts the queue after every call.
- The agent is measured on conversations, jobs booked and booked revenue: a call counts only when the customer actually talks (5+ seconds connected), unconnected dials never count, and every booking is owner-approved.
It starts from your list
The raw material is the list you upload: past customers, old leads and unsold estimates, with whatever you have on the work — the job, the date, the quote. Each record gets a clock from the work it describes. A March furnace tune-up starts a six-month clock. A water heater installed in 2015 starts the replacement conversation. A quote written Tuesday starts a follow-up clock measured in days, not seasons.
The service clocks
HVAC: two tune-ups a year, spring and fall — the two most predictable due dates in the trade. Water heaters: a 10–12 year working life and an annual flush, so both a yearly rhythm and a replacement window. Maintenance plans and service agreements: the cadence in the agreement. Quotes: due immediately, then again when their season turns. The industrial side runs on the same clocks: forklifts on quarterly PM under the service contract, racking on an annual safety inspection, dock doors on annual service — a door down is a dock stopped. These are the trade's standard clocks, and your setup answers adjust them to how your shop actually runs.
The queue, not the blast
Every morning the queue is simply the customers whose clocks have run out, ordered by due date. The agent works the queue inside your calling window, on the sequence you set — and each call rewrites its own record. Booked: out of the queue, into your schedule. Replied: out. Opted out: out for good. “Call me in spring”: the clock resets to spring. Every outcome is a new date on somebody's clock, so the list gets sharper every week it runs.
The call names the reason
Because the agent knows why it's calling, the call sounds like your shop, not a call center: “Your last tune-up was in March — the fall one's due.” “That water heater we quoted in June — still heating okay?” The customer hears their own history, which is the entire difference between a reminder from their shop and a pitch from a stranger.
How revenue agents are measured
Due dates decide who gets called. Measurement decides whether it worked — and the report never grades its own homework with soft numbers. The same four measures run through your portal report and the summary call your agent reads you, defined one way, in one pipeline:
An example run at a sample shop — what your report looks like. Plunger AI is new and has no client results to show, so every number below is illustrative, drawn with the exact definitions your real report uses.
Booked revenue, week by week
12 sample weeks829
Conversations · sample total, 12 weeks
195
Jobs booked · sample total, 12 weeks
$88,725
Booked revenue · sample total, 12 weeks
Where the bookings come from
Example month, by serviceShort labels, full names: Lead Recycling, Maintenance Plan, AI After Dark, 1 Minute Lead, Surveys. Surveys book the fewest jobs on purpose — their job is reviews: happy customers get routed to your review link.
Calls placed
Every dial the agent makes, up to the attempts you set. A dial is effort, not a result — it is never the number the agent is measured on.
Conversations
A call counts when the customer picks up and actually talks — five seconds or more connected. Rings, voicemail and unconnected dials never count as conversations.
Jobs booked
Counted only when the call's own record says a job was booked — and every booking waits for your approval before it goes on your schedule. No approval, no booking.
Booked revenue
Jobs booked × the average job value you set at setup — $455 in the sample above. An estimate in plain dollars, and the same math your summary call reads back to you.
Cancelled jobs come back out. If a booked job cancels, it is removed from the count — cancelled jobs are never counted and never charged.
The numbers, worked
On a 2,400-name list with standard clocks — half the list on twice-a-year tune-up cycles, the rest on annual rhythms — roughly 150–250 customers come due every month, steady all year. That's the quiet arithmetic generic timers miss: there is no slow month in the queue, only slow months in who bothers to work it.
At the working default of 3 attempts, a month's due list turns into real conversations with most of those 150–250 customers. Calls are unlimited, so the agent dials until the queue is worked — not until a counter runs out. If just one in four of those conversations books at a $455 average ticket, that's roughly 40–60 booked jobs and $18,000–$27,000 in booked revenue off a list you already own. The queue carries the rest forward. Nothing falls off the back.
Questions
What if our records don't have dates for every customer?
Then the clock starts from what you do have. Customers with a last-job date get precise due dates; customers without one join the seasonal waves — the spring and fall pushes every shop runs — and every completed job sharpens the record from there. A partial list is a working list.
Who decides the cycles — you or us?
You do. The defaults are the trade's standard clocks (tune-ups twice a year, water heaters at 10–12 years, annual flushes — forklifts on quarterly PM, racking on annual inspections), and your answers during setup set the cadences your shop actually runs — maintenance plans, service agreements, your own follow-up rhythm.
Does it learn from call outcomes?
The record does. Every call updates the customer's state — booked, bowed out, call me in spring — and the queue re-sorts around it. A customer who says “after the holidays” comes due after the holidays, not in the next generic blast.
What stops the agent from calling the same person twice?
The queue. Each customer has one record and one state; a booking, a reply or an opt-out moves them out of the active sequence immediately. Nobody gets worked by two campaigns at once.
Can we see why a customer was called?
Yes — every record shows the call, the summary and the reason it was placed: the job, the date, the clock that came due. Nothing in the queue is a mystery.
Free first
See your number before you pay for anything
Run the free Revenue Audit on your real numbers — past customers, unsold estimates, missed calls. Takes about a minute — no email up front, no card. The audit shows it. The agent collects it.