The ROAS was a fiction and everyone knew it
Norrland's channels claimed 6.4× return. The bank account claimed otherwise. We rebuilt measurement from the ground up and found 38% of reported revenue was being counted twice.
The problem
Meta claimed 4,100 conversions. Google claimed 3,800. The shop had 5,200 orders. Both platforms were taking credit for the same customer, and the marketing team was making budget decisions on the sum.
Ad blockers and ITP meant client-side tracking missed roughly three in ten orders — disproportionately the desktop, high-value ones.
This is the least glamorous problem a growth team can have and it silently corrupts every decision downstream.
What we did
- 1
Move tracking server-side
We stood up server-side GTM and wired the Conversions API for Meta and enhanced conversions for Google. Order tracking accuracy went from 71% to 99.2%, measured against the actual order table rather than a platform's word.
- 2
Build one number everyone agrees on
We made the ERP the single source of revenue truth and reported blended ROAS against it. Platform-reported numbers stayed visible but stopped being decisions.
- 3
Reallocate on the real picture
Once the fiction cleared, two campaigns that looked like winners were breaking even, and a neglected Google Shopping segment was quietly at 5.1×. We moved the money.
What it returned
The honest ROAS was less than half the reported one. That was a difficult meeting and it was the most valuable thing we delivered.
Revenue rose SEK 4.1M on flat spend, purely from pointing the same budget at things that were actually working.
“They told us our numbers were wrong in week two and spent the next four months proving it, then fixing it. Painful, and we would not go back.”