A platform grading its own numbers
Meta claimed £15 per signup for the client’s ad spend, below the business’s blended £22 across every channel. A paid channel coming in that far below a blend that includes free and organic traffic warranted a closer look.
Without reconciling Meta’s top-of-funnel data against what converted downstream, nobody could say whether the platform’s number was real.
Reconcile Meta with GA4
Rig brought Meta’s click and impression data and GA4’s conversions into one warehouse, joined, so Anthony could take the top of the funnel from Meta and the bottom from GA4. Reconciled that way, Meta was costing £104 for each signup.
Build a custom attribution model
Anthony then built a custom waterfall attribution model, ranking each customer’s data by quality to tag where they came from. It could place about 80% of customers, and it put the cost per signup higher still, at £139 to £170.
Check it with statistics
A regression of daily Meta spend against signups gave about £48. Using only the weeks when Meta and Google spend moved in opposite directions, to separate the two channels, gave about £55. Anthony judged that range the most credible, with error bars showing how sure the answer can be.
He is now tracing why the methods disagree, from the way tagging is done to customers who drop out of a journey and restart it on another device.