Measuring what a campaign actually did
Measurement is where most campaigns quietly go wrong: not in the ads, but in what everyone agreed to count.
Two campaigns can spend the same money, sell the same amount, and produce opposite reports. The difference is never the ads. It is what was counted, when it was counted, and who was given credit for it.
Get the plumbing in before the launch
Two things do the work. The X Pixel sits on your site and reports what visitors did there. The Conversions API sends the same events from your server, which survives the browser conditions that lose pixel events — blocked scripts, a person who left before the page finished, a purchase confirmed later.
Use both. They report the same events through different routes, and the platform reconciles them. Fitting them afterwards means the campaign’s first weeks are the ones you can never explain.
Count events that mean something
The temptation is to count everything. Resist it: an account tracking fifteen events has fifteen numbers and no answer.
Pick one primary event that equals value — a purchase, a qualified sign-up, a booked call — and a couple of supporting ones on the way to it. Define them in plain words and write the definition down. “Lead” means something different to marketing, sales and finance, and the version in the ad account decides what the platform optimises for.
If your value is not a website event at all, say so early. Some of the best campaigns end in a phone call or a shop visit, and the honest answer is a proxy event plus a tracked question at intake, not a metric invented to fill the dashboard.
Attribution is a choice, not a fact
Every platform reports conversions it believes it caused, within a window it chose. Those windows overlap between platforms, which is why the sum of everyone’s claimed conversions usually beats the total you actually sold.
Decide the rules before the campaign runs, not after the numbers land: which window, which events, and which source is the tiebreaker. Then compare the platform’s number against your own system’s total, and treat a gap as information rather than an accusation.
Read the report against the objective
A report is only readable against the thing the campaign was for. Awareness measured in clicks looks like failure; conversions measured in impressions looks like triumph. Both readings are useless.
Start every report with the objective agreed before launch, then show the spend, the result, and what changed because of it. Put the changes in it too — bids raised, audiences narrowed, a creative set retired — because a number with no decision attached teaches nobody anything.
Signals that a report is flattering you
- The metric changed halfway through the period.
- The comparison is against the worst week, not the usual week.
- Rounded percentages appear without the counts underneath them.
- Conversions are up and revenue is flat.
- Nobody can say what would have counted as a bad result.
Keep it boring and comparable
The value of a report is mostly in repetition: same events, same window, same shape, week after week. A dashboard that changes definition every month cannot show a trend, and a trend is the only thing that tells you whether the work is compounding.
Change a definition when it is wrong, never when it is unflattering, and mark the date in the report so the line’s kink has an explanation next to it.