Case study · Home services
3.1× the booked jobs, from the same households
A five-location HVAC group was buying leads from three vendors and could not say which ones turned into work. We rebuilt the demand around households rather than clicks — streaming, mail and search reaching the same addresses — and wired the phone to the CRM so the answer stopped being an argument.
Placeholder case study — figures, quote and creative are illustrative until a signed client release is on file.
booked jobs per month
cost per booked job
to the first tracked booking
locations on one program
The business, and what was actually wrong
Residential HVAC across two metros: five locations, forty-one trucks, replacement work carrying the year and maintenance plans carrying the shoulder seasons. Healthy business, run by an owner who could tell us his revenue to the dollar and his marketing cost to the nearest shrug.
Three lead vendors were invoicing him every month, and each one claimed the same jobs. Nobody was lying, exactly — they were all counting a call they had touched. What nobody could produce was a number for what a booked job cost, which meant every budget conversation was a negotiation rather than a decision.
- Three vendors, three sets of numbers, no shared definition of a lead
- Calls ringing into five different lines with no source attached to any of them
- Peak-season spend set by last year’s invoice rather than by capacity
- A brand that looked like five separate companies across trucks, mail and web
At a glance
- Sector
- Residential HVAC
- Footprint
- 5 locations, 2 metros
- Average job
- $9,400
- Engaged
- 14 months, ongoing
- Channels run
- Streaming, mail, search
- Team on it
- 4 people, one lead
What we actually did
In this order, because measuring last would have meant arguing about the results for a year.
Measured first
A tracking number per channel and per location, wired into the CRM, before a dollar moved.
Picked the household
Postcode-level targeting drawn from where the crews already made margin, not a radius on a map.
Shot it once
One day on site produced the television spot, nine social cutdowns and the mail creative.
Landed three times
Streaming, mail and search timed to reach the same address inside a fortnight.
Sent the truth back
Booked jobs pushed to the platforms, so bidding optimised on work rather than on phone calls.
What came out of one shoot day
Everything below was made in-house, from the same morning on site with their own crews and their own trucks.
Placeholder frames. The finished creative goes in as the client clears each asset for publication.
“We stopped arguing about which vendor was telling the truth. There is one number now, everybody can see it, and the arguments are about what to do next instead.”
Placeholder quote — replaced with an approved one when the client signs the release.
How the numbers were measured
A multiplier is only worth reading if you know what it was measured against. Here is the whole basis.
The baseline is the twelve months before we started, taken from their CRM rather than from any platform report: jobs marked completed with a revenue line against them. The comparison period is the twelve months after the first full month of tracking, so the three weeks of setup are excluded from both sides.
A booked job counts once, at the location that ran it, credited to the first channel that touched the household inside a thirty-day window. That is a choice, not a law, and it is generous to the top of the funnel — a last-click model would flatter search and make the mail look worse than it is.
- Seasonality: both periods cover a full year, so the summer peak sits in each
- Capacity: they added six trucks in month nine, which raises the ceiling and is not our doing
- Pricing: their average job rose 7% over the period, which we have not adjusted for
What did not work
The first mail drop went out a fortnight before the call tracking was live. Six weeks of response cannot be attributed to anything, which is our mistake rather than theirs — we should have held the drop and did not.
Programmatic display ran for a quarter and never produced a booked job we could trace to it. We stopped it and moved that budget into a second mail cadence, which is where it should have gone at the start.
The maintenance-plan email sequence took three attempts. The first two read like an agency wrote them, because an agency had; the third was rewritten around what their own dispatchers say on the phone, and it is the one that works.
What we ran, and what each part was for
Six of the eight disciplines. The two we left out were brand strategy and e-commerce, because neither would have moved this business.
Streaming TV
Reach at the top of the household, bought for two metros rather than a nation.
Direct mail
The same addresses again, on paper, inside the fortnight the ad had already reached them.
Google Ads
The bottom of the funnel: whoever was ready to call that afternoon.
Creative and production
One shoot day, planned for cutdowns rather than for a single finished film.
Websites and apps
A landing page per metro, because a shared one converted at half the rate.
Tracking and attribution
The layer that made every other number on this page arguable rather than assertable.
More of the work
$41 per lead
Storm-season search paired with follow-up that answers in under a minute.
62% more calls
Local SEO plus call tracking that showed which pages actually rang the phone.
2.4× consult rate
A qualifying form and a proof-led landing page that stopped paying for tyre-kickers.
Want to know what your version of this costs?
The audit is the same work this engagement started with: read the account, read the tracking, read the follow-up, then say which two or three moves would pay first. Free, and yours whatever you decide.
Get a free audit