Creative Testing Roadmap with Pre-Committed Decisions

Most creative tests in paid social cannot be resolved at the budget they run on. Separating a twelve percent difference in hook rate takes roughly ten thousand impressions a cell and resolves in days. Separating a twelve percent difference in purchase rate takes several hundred conversions a cell, which at a sixty dollar acquisition cost is real money per cell before anyone learns anything. So the first thing this roadmap does is sort your questions into the ones your budget can answer and the ones you should stop pretending to test. The ladder: Hook tests are cheap, fast and high volume, and they resolve on an upstream metric. Angle tests cost more and need the hook held constant, which most accounts forget. Format tests are expensive because each cell needs its own production. Offer tests are the most expensive of all and the most often run casually. We put them in that order and hold the line, because the common failure is an account testing four things at once and learning none of them. Cell design on Meta, specifically: Broad targeting and Advantage+ placements mix delivery, so a six-cell test inside one ad set resolves into whatever the system decided in hour four. There are two honest responses. Separate ad sets with matched budgets and a hard floor on learning-phase exits, which costs more and gives you a creative verdict. Or accept in-ad-set delivery and read the result as a delivery preference rather than a creative verdict, which is cheaper, still useful, and is what most accounts should do most of the time. What we will not do is run the cheap design and report it as the expensive one. Decisions written before the data: Every test in the plan carries a stop rule and a pre-committed action for each outcome, agreed by whoever will want to overrule it in week nine. This is the part that changes results. An account that decides in advance what a losing hook means will kill it. An account that decides afterwards will find a reason to keep it, and usually does. Incrementality where it is worth buying: Geo holdouts need enough matched markets to survive normal variance, and below a certain spend the effect is smaller than the noise between markets. We size that honestly, and where the answer is that your account cannot afford a clean lift read this quarter, we say so and put the money into hook volume instead. What you receive: A twelve-week calendar with each test dated and sequenced against your production capacity, so nothing is scheduled that you cannot shoot. A spec sheet per test: hypothesis, cell design, primary metric, the threshold that counts, the stop rule, and the action for each outcome. A naming convention that maps every creative to its angle, hook and format so reporting groups itself instead of requiring a person. A reporting view built in Sheets or Looker Studio against your existing export. A one-page decision log your team keeps, because the value compounds only if the reasoning survives the quarter. What is not included: We do not buy media or operate the account. We do not implement attribution platforms, media mix models, or server-side tracking. We do not produce the creative being tested. Who this is not for: Accounts under roughly fifty thousand dollars a month on a single platform. The tests will not resolve inside a planning cycle and a roadmap would be theatre. Teams whose current reporting cannot attribute a result to a specific creative because nothing is named consistently. That is a naming and taxonomy problem and it has to be fixed first. We will tell you that in the first call and it is not this engagement. Anyone who wants a testing framework they can apply to any account. The whole point is that the plan is sized to your spend, your production capacity and your conversion volume, and a portable framework has none of those in it.