Situation. A services business with no outbound function, growing entirely through inbound demand and wanting a second engine. The standard approach is to hire a team, buy lists, and find out over the following year whether any of it works.

The number nobody had computed. Which lead source produced clients — not conversations, not appointments, clients. Outbound programs are typically judged on activity metrics because those are available immediately and the real one takes a full sales cycle to appear. The result is teams optimizing dial counts while the source mix quietly determines the outcome.

The test and how we made it readable. We built the function around source testing from the start rather than bolting measurement on afterward. Each lead source was tracked separately through to closed business, with enough volume per source to distinguish a real difference from a run of luck. Where deal counts were too low for a statistical read, we used pre-agreed milestone checks instead — defined in advance so a source could be judged on evidence rather than on how the last call happened to feel.

We also tracked conversion by rep, which matters more than it appears: variation above roughly 20% between reps on the same source means the constraint is in the sales motion, not in lead quality, and no amount of better leads will fix it.

Result. One source emerged as a consistent producer of new clients and became the backbone of the channel. Outbound went from nonexistent to a reliable second acquisition engine.

What transfers. Build measurement into a new channel at the start; retrofitting it means the first year of data is unusable. Judge lead sources on closed business, not on activity. And when volume is too low for statistics — which is common in B2B and services — replace them with milestones agreed in advance, so the decision is still made on evidence rather than on the most recent anecdote.

‍