Almost every conversation about paid social starts in the same place: our cost per lead is too high.
It might be. But cost per lead on its own tells you very little, and acting on it alone is how good campaigns get switched off and bad ones get more budget.
The number has a denominator problem
Cost per lead is spend divided by leads. That means it moves for two completely different reasons — the cost went up, or the definition of "lead" changed. Most of the time it is the second one, and nobody noticed.
If your form got shorter, your cost per lead fell and your cost per customer probably rose. If you added a qualifying question, your cost per lead rose and your sales team got happier. The headline number moved in the opposite direction to the business result in both cases.
What to measure instead
Track the whole chain, and hold each step to its own number:
- Cost per lead — what you pay for a form fill
- Qualification rate — how many of those are a real prospect
- Cost per qualified lead — the first number that means anything
- Close rate — how many become customers
- Cost per customer — the only figure that decides whether to keep spending
A campaign at double the cost per lead with three times the qualification rate is the better campaign. You cannot see that from the ads dashboard alone, which is exactly why the dashboard should not be the only thing you look at.
The tracking is probably wrong
Before drawing any conclusion, check that the numbers are real. In our experience most accounts under-report, and a few wildly over-report.
Common causes we find:
- The conversion event fires on page load rather than on submission, counting everyone who reached the form
- Two events are firing for one submission, halving the apparent cost
- iOS and browser privacy features are suppressing conversions the campaign genuinely produced
- The thank-you page is missing, so there is nothing distinct to measure
That last one is worth dwelling on. If your form posts back to the same URL, there is no clean event to count. Give a successful submission its own page and the measurement problem largely solves itself.
Attribution windows change the story
The same campaign will report very different numbers on a 1-day click window versus 7-day click. Neither is wrong. They answer different questions.
Short windows credit only the impulse buyers. Long windows include people who saw the ad, thought about it, and came back. For a considered purchase — property, professional services, anything over a few hundred pounds — a short window will make your ads look far worse than they are.
Pick a window that matches how long your buyers actually take to decide, and then leave it alone. Changing it mid-quarter makes the trend meaningless.
Frequency, not creative, is often the culprit
When a campaign that was working starts to slide, the usual assumption is creative fatigue. Sometimes it is. More often the audience is simply too small for the budget, and the same people are seeing the ad six or seven times.
Check frequency before rewriting the ad. If it has crept above three or four in a short window, the fix is a bigger audience or a lower budget, not a new headline.
When the number really is too high
Sometimes the answer is genuinely that the channel does not work at your price point. That happens, and it is worth saying out loud rather than spending another quarter proving it.
The honest test: if cost per customer is above your gross margin on a first sale, and you have no repeat business or referral to make up the difference, paid social is not your channel right now. Fix the offer, the margin or the follow-up first. More budget will not rescue arithmetic.
The reporting we send
For every campaign we run, the monthly report leads with cost per qualified lead and cost per customer, with cost per lead shown underneath as context. Spend, leads and close rate sit beside them.
It is a less flattering way to report, because it makes it obvious when something is not working. That is the point — it is also the only version of the report you can make a decision from.
