Most paid media playbooks start with creative testing: run three ad concepts, see what wins, scale the winner. We've learned to hold off on that until we've checked something less exciting first — whether the account can actually tell the difference between a good outcome and a bad one.
Why measurement comes before creative
We took over an account where the previous team had been running aggressive creative testing for months, cycling through dozens of ad concepts, chasing CTR and engagement. When we audited the conversion tracking, the "purchase" event was firing on the cart page, not the order confirmation page — meaning every abandoned cart was being counted as a completed sale. The account had been optimising toward a metric that didn't mean what everyone thought it meant, for months, through dozens of "successful" creative tests that were actually measuring something closer to cart-add rate.
No amount of creative testing can fix that. It just produces confident, well-documented wrong answers faster.
What we check before touching creative
Does the conversion event fire where it's supposed to? We manually walk through the purchase flow ourselves, checking that the pixel or conversion tag fires exactly where it should — not one step earlier, not one step later — before trusting a single number the account has ever reported.
Is the account counting duplicate conversions? Some setups fire a conversion event on every page load after purchase, including if the customer refreshes the confirmation page. This inflates conversion counts in ways that are easy to miss and can make an underperforming campaign look mediocre instead of bad.
Are view-through conversions inflating the picture? Platforms often count a conversion as "attributed" if someone merely saw an ad and later converted through an unrelated path, sometimes days later, with no click in between. Depending on the settings, this can meaningfully inflate reported performance for campaigns that aren't actually driving much.
Does the reported revenue match the client's actual revenue? We ask for real order data and compare it to what the ad platform reports, even roughly. A mismatch of more than 15-20% usually points to a tracking issue worth resolving before any optimisation decision gets made based on the platform's numbers.
The order of operations that actually works
Fix measurement first, even though it's the least visible, least exciting part of the work and the part a client is least likely to notice or thank you for. Then test creative, audiences, and bidding strategy against numbers you can actually trust. Skipping straight to creative testing on an account with broken measurement isn't a shortcut — it's just running the same experiment as everyone else, with a broken scoreboard.