Method
How work is measured
Three numbers we report, three we refuse to report as outcomes, and why our case studies are less impressive than everyone else's.
Media reporting has an unusual property: the numbers are abundant, precise, and largely disconnected from the business. A report can be entirely accurate and tell you nothing about whether the money worked.
What we report
1. Outcomes as counted by your ledger
Platform-reported conversions are an input, not the answer. Each platform claims conversions by its own rules, and those rules do not add up across platforms — a customer who saw social, then searched, then bought is claimed in full by both.
The number we report against is the one in your accounting system, with the gap to platform-reported figures attributed to named causes.
2. Where the money went, at property level
Not a channel summary — the actual placement report. This is the number that most often changes a client's mind about their media buying, and it is usually the first time they have seen it.
3. Incrementality, where budget justifies it
The only measurement that answers the question the business actually has: what would have happened without this spend? Attribution cannot answer it, because the counterfactual is not in the data. A holdout can — withhold spend from a comparable group and compare.
It costs volume, and from roughly the point where a month's spend exceeds the cost of the test, it is worth it.
What we do not report as outcomes
- Impressions and reach. Quantities purchased. Reporting them as results is reporting the invoice as an achievement.
- Engagement. Unless your business sells engagement, in which case we will measure it properly and say so.
- Platform-reported return on spend, unreconciled. The single most over-trusted figure in media.
Why our case studies look weaker
The standard format is a large percentage and a logo. "+340%." As evidence it is close to worthless, because it omits the baseline, the period, what else changed, and how it was measured.
It also omits survivorship. An agency that runs twenty engagements and publishes its best three has described the top decile and presented it as the method's typical result.
A question worth asking any agency, including this one: how many engagements did you run in the period these case studies come from? Three case studies from four engagements is a track record. Three from forty is a selection — and the arithmetic is not usually volunteered.
What a monthly report contains
- Spend, and outcomes as counted by the ledger.
- Cost per outcome, with the known sources of error stated.
- The placement report, and what was excluded this month.
- What changed, what we expected it to do, and where it disagreed.
- The decision for next month — and what would make us reverse it.
Point 5 is what makes the report worth reading. A recommendation with no stated condition for reversal is an opinion. With one, it is a plan that can be checked.
Last reviewed: 11 September 2026