The Attribution Detective
How I untangled a broken measurement stack and found R600k in misattributed spend — and why your attribution model is probably lying to you too.
Every channel looked profitable. Revenue was flat.
A brand spending R1.2M/month on paid media was getting strong reported ROAS across every channel. Google Search: strong. Meta: strong. Programmatic: above benchmark. The monthly report was all green arrows. Revenue growth had stalled for six months.
The CEO's instinct was that something was wrong with the measurement, not the channels. He was right.
Three layers of attribution dysfunction
When I went through the accounts forensically, I found three separate attribution problems compounding each other:
Users who saw a Meta ad and then searched the brand name were being attributed to Brand Search in the last-click model. Meta was driving the conversion — Brand Search was just the last click before purchase. R400k/month of Meta spend was appearing as Brand Search conversions.
The retargeting campaign had a 7-day view-through attribution window enabled. Users who saw the ad but didn't click were being attributed as conversions. The actual click-through ROAS was 40% lower than reported.
The GA4 and Meta pixel conversion events weren't deduplicated. The same purchase was being counted once in Meta Ads Manager and once in Google Analytics, inflating total reported conversions by approximately 35%.
Fix the measurement before touching the budget
My first recommendation was counterintuitive: don't change the campaigns yet. Fix the measurement first. Changing budget allocation based on broken attribution would mean optimising toward the wrong numbers.
We rebuilt the attribution framework over four weeks: implemented server-side tracking for deduplication, switched to data-driven attribution in Google, adjusted Meta attribution windows, and built a unified dashboard that read from a single source of truth rather than individual platform reports.
Once the measurement was fixed, the budget reallocation was obvious: R600k that had been attributed to brand search and retargeting was redirected to the upper-funnel paid social that had actually been driving the conversions.
ROAS up 28% without changing creative or total budget
Within 60 days of implementing the measurement fixes and reallocation, blended ROAS improved by 28%. Not by finding better creatives. Not by finding a new audience. Just by understanding what was actually working and putting money where the real performance was.
Optimise toward the right numbers, or you'll optimise toward the wrong outcome
Attribution isn't a reporting problem. It's a strategic problem. If your measurement is broken, your budget decisions are broken — regardless of how skilled your media buyers are. The audit that finds the measurement problems is worth more than any campaign optimisation.