A Creative Testing Framework for Paid Social

Test angles first, executions second

An angle is the argument — why someone should care. Saves time. Made by hand. Cheaper than the alternative. Trusted by people like you. An execution is how the angle is expressed: the format, the actor, the music, the edit.

Testing five executions of one angle tells you which video performed. Testing five angles tells you what your market responds to, which is knowledge that outlives the ad account and informs the website, the sales script and the packaging.

Change one variable at a time — properly

The reason most tests are uninformative is that two ads differ in six ways. A different hook, different footage, different music, different length, different call to action. When one wins, you have learned nothing transferable.

Hold everything constant except the thing you are testing. This is slower and it is the only way results accumulate rather than reset.

Decide the sample size before you look

Ad platform dashboards invite you to call winners far too early. A creative with 40 clicks and a good conversion rate is noise.

Set thresholds in advance and hold to them:

• Hook rate (3-second views over impressions) — judgeable at around 3,000 impressions.
• Click-through rate — around 1,000 impressions per variant.
• Cost per acquisition — at least 25 conversions per variant, preferably 50.
• If a variant cannot reach the threshold within the test budget, it should not be in the test.

Diagnose where the creative fails

Different metrics point to different problems, which turns a vague failure into a specific fix. Low hook rate means the first two seconds failed — the opening frame or the visual pattern break. Good hook rate but low click-through means the middle failed — the argument did not land. Good click-through but poor conversion usually means the ad set an expectation the landing page did not honour.

That last case is the most common and the most misdiagnosed, because it looks like a creative problem and is actually a mismatch problem.

Keep a creative library

Record every test: the angle, the execution, the numbers, and a one-line conclusion. It takes ten minutes a week and it is the difference between an account that improves and one that oscillates.

After six months, patterns emerge that no single test would reveal — that founder-led video consistently beats studio production for you, that price-led angles work in one category and fail in another. That library is a genuine competitive asset, and it is the only output of testing that appreciates.