Platform numbers overlap by design
Meta, TikTok and Google each claim the same order under their own attribution windows. Summed conversions routinely reach 120% to 160% of real orders. Use platform data to decide which ad wins inside that platform. Splitting budget between platforms needs a number the platforms cannot inflate.
GA4 on last non-direct click undercounts paid social, because much of the effect is view-through and dark social. Two systems disagreeing does not mean one is broken. They measure different things under different rules.
Blended MER is the number to run the business on
MER is total revenue divided by total marketing spend, pulled from Shopify or your ERP. It cannot be double counted. Track it weekly with new-customer CAC and contribution margin. When you raise spend, MER tells you what the business actually got.
Set an MER floor tied to margin and overheads. If 4x covers the plan, push spend while MER holds. When MER slips two weeks running, the incremental budget is buying nothing and should come back out.
Prove the channel with a holdout
Geo holdouts are the practical test for a UK brand. Split regions into exposed and control, hold the split for 4 to 6 weeks, then compare total orders. It is blunt and honest, and it survives every attribution change the platforms make.
Post-purchase surveys add the qualitative layer cheaply. Response rates are commonly reported in a 30% to 50% band when the question sits on the thank-you page. Answers drift toward memorable channels, so read them for direction alongside the holdout maths.