I run paid media across Meta and Google for a supplement subscription brand in a tightly regulated health category. The work that mattered wasn’t making more ads. It was figuring out what the account was actually telling us.
Campaigns were judged on cost per trial. But a trial that never renews costs money. Some "cheapest" ads were bringing in customers who didn’t stay.
So: Every ad is now judged on cost per paid renewal, and the Meta algorithm is fed the conversion we actually want.
I used ad longevity in the Meta Ad Library as a proxy for performance (brands keep paying for ads that work) and studied 7 competitors. Their long-runners open with a symptom the customer feels. Every one of ours opened with an offer.
So: Six single-variable test briefs, each changing one thing, each with a written condition that would prove the idea wrong. Tests designed to teach us something, not just to find a winner.
With confidence intervals on each campaign’s cost per order, the middle 9 of 11 campaigns overlapped completely.
So: Stop reshuffling budget between look-alike campaigns. The account became a Scale campaign for proven ads and a Test campaign with weekly batches and written pause and graduate rules.
Comparing our ads, the number of distinct proof elements in a creative (label, ingredient, result, reviewer) tracked with better cost per purchase.
So: That became a brief rule: every new concept carries enough proof to stand on its own, instead of relying on taste.
Auditing where ads sent people, I found a price that didn’t match between pages, a wrong servings count, and an unsupported certification claim.
So: Those got fixed before spend scaled. Great ads can’t rescue a landing page that contradicts them.
These are real ads from the account I run, all public in Meta’s Ad Library. The earlier creative led with the offer. The newer creative follows what the research showed: lead with proof, use concrete amounts, open with the customer’s problem.
Source: Meta Ad Library, public. Performance figures stay private.
Supplements sit under Meta's health-ad policy, FTC guidance, and the product label itself. I built compliance into the triage of all 180 files (30 concepts). Each concept got a Launch, Fix, Hold, or Cut verdict with a written reason:
On Google I wrote a 25-asset text set (headlines, long headlines, descriptions) that steers clear of prescription-drug terms Google restricts, and recommended pausing the Performance Max campaign that wasn’t converting until the assets and measurement were right.
Business figures (revenue, spend, conversion and retention rates) stay private at my employer’s request. Happy to walk through the real work on a call.
I'd set the renewal-based metric before the first dollar of spend. Optimising for the wrong number early trains both the algorithm and the team on the wrong thing.