The teardown · free · nothing to fill in

772 leads in 26 days, at $26.87 each. Here is the entire campaign.

One Florida contractor, Meta lead ads, three markets, $20,740 spent. Below is what we ran, which concepts produced the leads and which ones we killed, where each market landed, and what happened to a lead after it arrived. About a three-minute read.

Every figure is reconciled against the ad account, the database and the CRM. Where a number covers only part of the run, it says so.

What actually produced the leads

Eight concepts ran against the same audience, in the same markets, at the same time — so the only variable was the idea. Two of them ran away with it, and neither is about roofing. One is weather. One is insurance. The concepts a contractor would expect to win — guarantees, financing, reviews — finished at the bottom.

Ranking from the opening window of the campaign, not the full 25 days — these do not sum to 772 and should not be read as shares of it.

It is still getting cheaper. The newest wave of concepts, live since 25 August, is running at $21.34 a lead against the $26.87 blended average — the test never stopped, so the winners keep replacing the average.

The lesson transfers past roofing. People do not respond to your craftsmanship; they respond to the thing that just happened to them. Find the event that puts your trade in someone's head this week and lead with it.

What happens to a lead after it arrives

Cheap leads are worth nothing on their own. This is the part that decides whether $26.87 was a bargain or a waste, and it is the half most agencies never build.

1. 01

   A homeowner enters an address.

   Nothing else is asked of them. No upload, no account.

2. 02

   The roof is measured from satellite.

   Area and pitch computed — no truck roll, no tape measure, no appointment booked to find out.

3. 03

   Their own pricing produces the tiers.

   Locked rates and deterministic math, never a model guessing at a number.

4. 04

   It lands in the CRM screened and texted,

   on a follow-up board with the cadence clock already running.