Rebuilding After an Account Shutdown

A cause-driven ecommerce brand had climbed to close to $1M a month early in 2020. Then a boosted post that mentioned the pandemic got its ad account disabled, with millions of dollars of data on it. The brand was off the platform for two months, hired an agency, and hadn't turned a profit on ads in about three months when it came to the agency where I was senior media buyer and analyst. It became that agency's first client paid on performance.

The brand still had cash, a strong product and a strong offer, and an average cost per sale of about $130. Rebuilding meant testing in an account with none of its old history. Then Facebook's cost per thousand impressions nearly doubled, two or three days into testing.

I audited the account in far more depth than anyone asked for. Testing opened with a calibration run: three to five thousand impressions at each optimisation stage, from add-to-cart through purchase, to see how far the algorithm leaned. It leaned hard toward purchases, so we could test aggressively. I checked whether it was us first when prices nearly doubled. The same spike showed in other client accounts and among the members of the coaching group I ran, so it was platform-wide. That week was the obvious moment to launch new creative and a new landing page, and I held them back: with the ads changing mid-spike, nothing could be compared against the baseline we'd already built. We cut budgets, moved part of the spend into a second ad account, which helped, and waited the spike out. The creative and audience tests ran against a clean baseline once prices settled.

• Account audit
• Calibration test across optimisation events
• Creative test, then audience and creative test
• Second ad account to isolate the spike
• Facebook Ads

The creative test turned up one ad producing purchases at $13 to $18. An audience and creative pairing then came in at $11 to $12 per sale, against the account's previous average of about $130. Both were test results, and the next step was validating them at a larger budget.

The owners got a cost per sale they could trust, measured against a clean baseline, and an explanation for the price spike backed by other accounts instead of a guess.

A test only tells you something when one thing changed. The platform is already the thing changing during a platform-wide price spike, so new creative launched that week can't tell you anything.