The three formats
Every Meta account we manage runs three core creative formats: static single image (optimised for retargeting and low-funnel), short-form video 15–30 seconds (optimised for prospecting), and carousel (for AOV improvement and product range discovery). These aren't three options - they're three pillars. Every campaign that hits scale in our portfolio runs all three. Removing any one of them reliably reduces performance.
Testing methodology one: the hook test
The hook is the first three seconds of a video ad. It has a disproportionate impact on performance because it determines whether the algorithm serves the ad at full volume or throttles it. We run hook tests monthly - same product, same offer, five different openings. We test a direct product statement, a problem statement, a social proof hook (number of customers, review quote), a surprise or pattern interrupt, and a question. The winner's format rolls into the core campaign. The other four are killed regardless of how they feel creatively.
Testing methodology two: the offer test
We separate creative testing from offer testing deliberately. Once we have a strong creative format, we then test offer framing separately - the same ad with four different value propositions: price (save $X), urgency (limited time), social proof (most popular), and benefit clarity (what they actually get). This is where the biggest performance gaps appear. For Australian DTC brands we've found social proof hooks underperform in prospecting but overperform in retargeting - the inverse of what most people assume.
The one rule we never break
We never run a campaign without a 7-day minimum data window before making optimisation decisions. We've had clients push us to cut ads after three days of poor performance. Every time we've held the line and waited for the full window, the data has been right and the impulse to cut has been wrong. Meta's algorithm needs time to learn. The instinct to act immediately costs Australian brands money every quarter.