Case Study
Campaign Math: four attribution rules, one budget, and the data’s own defects
Which campaign worked best depends on a rule nobody chose. Campaign Math shows how much the answer moves, and whether the data can answer at all.
Figures on this page are from the live tool; the data is synthetic and seed-locked.
The dropdown Google took away
Attribution rules decide which campaign looks like the winner, and in November 2023 GA4 removed the transparent ones: no more first-click, linear, or position-based. The default that remains is data-driven attribution, which cannot be reproduced outside Google. Campaign Math computes the four transparent rules itself, by hand-checkable arithmetic, and lets them bracket the black box.
What Campaign Math does
The verdict. One synthetic quarter: $5,064 of $72,621 in paid spend bought clicks the brand’s own analytics cannot tie back to the campaign that paid for them, and 9 of 12 paid campaigns change rank by two or more places depending on the attribution rule. One campaign’s return on spend runs from 0.69x to 1.88x: the same spend, the same sales, a different story from a dropdown.
The honest twist. The top campaign holds under every rule, and the page says so plainly: that result is real, and it explains why (44% of revenue comes from one-visit buyers, where every rule agrees by arithmetic). The verdict copy and the rule for choosing it were committed before the data was generated, so the page could not have been written to flatter the data.
The hygiene audit. Nine classes of tracking defect, each planted at a published or stated rate: missing UTMs, Safari cookie expiry, payment-gateway self-referrals, bot traffic and the rest. Each row carries a “documented case” column, and every row but one says “none published”, because nobody publishes the budget they misallocated over tracking defects. The one exception proves the point: Uber cut roughly two thirds of a nine-figure ad budget and installs did not move.
How the proof is possible
Quellwood Provisions is invented and its quarter is generated from a fixed seed, with every defect rate published or stated and none adjusted after the result was seen. The full GA4-shaped event export, the spend table, and the generator’s own truth files are downloadable from the tool, and the event export’s SHA-256 hash is published so anyone can confirm the data behind the page is the data they downloaded. Every number on the page can be recomputed from those files by hand. No analytics vendor publishes that.
What this looks like on your data
The engagement runs the same four rules and the same hygiene audit on your GA4 export and spend sheet: which campaigns win under every rule, which wins are an artifact of the dropdown, and how much of your spend sits on traffic your own analytics cannot trace. The defects here were planted; yours are free-range.
Data is synthetic; methodology and deliverables are real.