Twelve leads, four jobs, one channel that paid for itself.

Small contractors have a revenue distribution with a long tail: dozens of small repairs and a handful of installs worth more than all of them together.

In a one-month file, whichever channel produced the install wins by an order of magnitude, and the conclusion is noise rather than signal.

Six months is the practical floor, and a year is better. The report shows the median job as well as the total by channel, because those two numbers disagreeing is itself the finding.

Where job type is in the export, splitting repair from install removes most of the distortion outright.

A homeowner who liked the plumber calls the plumber again. Over two years, repeat work is often the largest single line in a small contractor's book.

Attributed to nothing, it disappears into Direct / Unknown and makes every paid channel look worse than it is. Attributed to the original channel with no thought, it makes them look far better.

The report separates first jobs from subsequent ones and leaves the judgement to you, with the number in front of you rather than assumed.

That number is also the honest argument for the more expensive channel: a source whose customers come back twice is worth paying more for at the front.

A small contractor's phone rings from a van decal, a map listing, a search advert and a neighbour, and none of those produce a form submission.

Every one of them lands in the same place — the phone — so without tracking numbers there is nothing to join on and the report is mostly Direct / Unknown.

The fix is cheap: distinct tracked numbers on the website, on the vehicles and on any paid listing. One change makes every subsequent report worth reading.

Until that exists, the report states the size of the untracked bucket rather than guessing a distribution across the channels that do have data.

Lead marketplaces sell the same enquiry to several contractors and bill whether or not you win it. The cost per lead is known; the cost per job never is.

Matching invoices back to marketplace leads gives the only number that matters, which is what a won job actually cost once the losses are included.

Contractors who run this usually find one marketplace is defensible and another is not, where the cost-per-lead figures were nearly identical.

The same method applies to any channel that bills for contact rather than for outcome, which is most of them.

It also settles the argument about response time. Where a marketplace lead only converts when it is called within minutes, the invoice file shows it, and the fix is an operational one rather than a change of channel.