Sold is a stage. Built is a job. Collected is the only one that pays anybody.

A job can be sold and then die: the adjuster reduces the scope, the homeowner changes their mind, the mortgage company sits on the cheque, the crew never gets scheduled.

Every one of those has already consumed a measure-up, a sales appointment and a contract, and none of them produced revenue.

Ranking channels on sold jobs therefore rewards whichever source produces the most contracts, not the most roofs, and the two lists are different.

Exporting the sold-and-never-built jobs alongside the collected ones makes the difference visible per source, which is the number worth acting on.

It is also usually a process finding rather than a channel one: a source whose jobs die at one particular stage is pointing at that stage.

On an insurance job the price is set by an adjuster, the timeline is set by a carrier, and the contractor's sales process is mostly about documentation.

On a retail job the price is the contractor's, the timeline is the homeowner's, and the sale is a sale.

Blending them produces an average value that describes neither and a conversion rate that moves with the weather rather than with the marketing.

Where the export carries a job type, the report splits on it and each gets its own ranking and its own sensible window.

Most contractors find the sources feeding each barely overlap, which argues for two budgets rather than one.

Door-knocking produces a large share of exterior work in storm markets, and the cost of it — wages, commission, vehicles — is entirely knowable.

Because it is a payroll line rather than an advertising invoice, it sits outside the marketing report and is never ranked against paid search.

A canvasser code on the lead puts it into the same comparison on the same collected revenue, against what it genuinely cost.

The finding is often about which crews and which neighbourhoods produced collected work rather than signatures, which changes next season's plan more than a channel ranking would.

A hail event fills the pipeline for a fortnight. Those leads arrived because of the sky, and crediting them to whatever campaign happened to be running produces a conclusion that will not repeat.

Reading those weeks separately — by job type, by an insurance flag, or simply by the date range — keeps the rest of the year honest.

The point is not to discard the revenue, which is real, but to stop it being used as evidence that something works.

A contractor who does this consistently ends up with a defensible picture of the baseline, which is the thing that decides whether to spend in a quiet season.

It also stops the reverse error, which is cutting a channel that looked poor during a month when every crew was busy with storm work and nobody answered the phone. That is a capacity problem wearing a marketing problem's clothes.