The three numbers you need first.
What a customer is worth to you, including repeat work — not the value of the first job, which is how most businesses undervalue a customer and conclude that acquisition is too expensive.
What share of leads become booked work today, by source. Not form fills, not "opportunities" — booked work.
How long you take to respond, measured honestly, including evenings and your busiest week.
Without these, a lead source cannot be evaluated. With them, the decision is usually obvious and frequently is not the one you expected.
Shared leads are a speed auction.
Many paid lead products sell the same enquiry to several companies. You are not buying a customer — you are buying an entry in a race, and the race is decided in minutes.
That can be a perfectly good trade if you are genuinely fast. It is a very poor one if you are not, because you will pay the same price and lose most of them to whoever called first.
Attribution has to reach the job, not the form.
A source that produces plenty of enquiries and few booked jobs will look excellent on a cost-per-lead report and be losing you money. This is the single most common way marketing spend is misallocated.
Track through to booked, completed and paid work. The ranking of your sources usually changes once you do, and sometimes reverses.
The cheaper thing to do first.
Before buying more, look at the demand you already generated and failed to convert: estimates sent with no outcome recorded, calls with no callback, customers who stopped. That work is already paid for.
It is also the honest test of whether you have a demand problem at all. Many businesses convinced they need more leads actually need to finish handling the ones from last quarter.