Field Service

Tracing Ad Spend to Paid Jobs for a Home-Services Company

A home-services company in Canada with two brands bought ads and could not trace that money, or its calls, to a paying job. We joined calls, ads and jobs every night into a database the company owns, so cost per booked job and return per campaign could be asked of the owner's own data.

// The outcome
  • Told as it happenedno invented figures
  • Your data, your accountswhere the project allows
  • Anonymized on purposeno client names
Published 2026

Results

  • Calls, ad campaigns, search data and jobs from four services were collected every night into a database the company owns

  • Cost per booked job and return per campaign became questions the owner's own data could answer

  • Both brands sat in the same data, and one customer was linked across them and counted once

  • The owner got read-only access to 43 data objects, so his own tools could connect to the database

A home-services company in Canada with two brands bought ads and took calls, but the money could not be followed from an ad or a call to a job that was actually paid. The calls were in one service, the ads in another and the jobs in two separate job systems, one per brand. Ad budget decisions were made without that link.

Where the money was

The leak was the trail. Ad and call money went out, jobs came in, and nothing in the company’s own data connected the two. The owner worked out returns by hand from downloaded reports. We built one part for it. Calls, ads and jobs were joined every night into a database the company owns, so cost per booked job and return per campaign became questions that data could answer.

What we built

Every night a set of connectors collected the figures from four services: call tracking, Google Ads, Search Console and the job system, which ran as two accounts, one per brand. They wrote into a database the company owns, in its own account, and every run left a line in a sync journal. If a night failed, an alert reached us, and we fixed it.

A matcher then joined each call to the job that followed it. It used the normalised phone number and a 30-day window, and the first ad-attributed call got the credit. On top of that sat views for cost per lead, cost per booked job, and revenue and return by campaign, so those figures came out of the database without a person assembling them.

Later we added the second brand’s job account and linked one customer across both brands by phone, so a customer’s later jobs were credited to the original source and each job was counted once. We put a fixed set of read-only questions behind a chat connector on the company’s own server, and gave the owner a read-only data endpoint over 43 objects, so he could connect his own dashboard. The nightly jobs ran on a server in the company’s own account, and the work was delivered in milestones.

After this the owner could ask which campaign returned what on jobs, instead of working it out by hand.

What it did not do

We did not measure the money effect. We have no figure for changed ad spend, saved budget or more booked jobs, and we do not claim one.

Only part of the calls could be matched to a job. Matching by phone number has that ceiling. Call recognition and analysis could raise it, and that is planned, not built.

We did not build a dashboard for this company, call recognition, Local Services Ads or GA4 connections, or any advice on how to spend ad money. The system measured; the decisions stayed with the owner.

If your calls, ads and jobs live in different tools and nobody can say what a booked job cost, our integration work starts the same way: map what is there, then build one record you own.

Tech stack

  • Supabase
  • n8n
  • TypeScript
  • Node.js

Losing money somewhere similar?

Tell us where you think you are losing money. If we go further, the first paid step is the one-week Money Leak Check for $800. Builds after it are fixed price.