Five things we go looking for.

• Slow lead response

  Leads + jobs

  What it looks for. Every enquiry that came in, and how long it sat before somebody got back to it. Then it splits them in two: the ones answered quickly, and the ones that waited.

  What it costs out. Both groups rang the same phone. If the quickly-answered ones book more often, that difference — times the leads that waited, times what a job is actually worth to you — is the number.

  From the sample report: 2h 47m median first reply

• Aging quotes

  Quotes

  What it looks for. Proposals you sent that never got a yes or a no. Not won, not lost, not expired — just sitting there since the day you finished pricing them.

  What it costs out. We measure how often your older quotes eventually did close, and apply that rate to the pile still waiting. The expensive part — the visit, the pricing — is already paid for.

  From the sample report: 38 of 119 open quotes past 30 days

• Dormant customers

  Jobs

  What it looks for. Customers who used to come back and have not. Your own history sets the line — how long a gap is normal for the work you do, rather than a borrowed rule of thumb that means nothing in your trade.

  What it costs out. How many are overdue on that line, how often people who go that quiet still come back, and what a repeat job is usually worth. All three come out of your records, not ours.

• Uncollected invoices

  Invoices

  What it looks for. Work you already did, already invoiced, and never got paid for. This is the one that is not a maybe — the customer agreed to the amount.

  What it costs out. Not the whole overdue pile. Plenty of late invoices land eventually, and counting them all would be exactly the inflated number this product exists to refuse. It is discounted by how often invoices that old have genuinely gone bad in your own books.

• AI-booked job audit

  Jobs

  What it looks for. If an AI receptionist answers your phones, it follows those jobs the way it follows the ones your team booked: booked, completed, invoiced, paid. Vendors report the booking. Almost nobody reports what happened to the money after it.

  What it costs out. The gap between the two, per booked job — with your own human bookings as the control group. No vendor benchmark, no outside data. And it measures a gap, not blame: AI usually gets the after-hours and first-time callers, which is a harder book.

  From the sample report: 22 records declined — under the 30-row floor

Software computes every one of these figures. AI can explain one to you; it is never allowed to produce one. And where a detector does not have enough records to be sure, it says so and stays quiet — an empty finding is an honest one.

On the sample report, all five together read 1,832 records and came back with 4 open findings and ~$38K/mo revenue at stake — each printed beside its own arithmetic, like 38 × $3,420 × 0.34 × 0.38 = $16,500/mo. See the sample report →