Questions about video ad creative, answered straight

My ads get clicks and views but no sales. Why?

Clicks and views measure the ad; sales measure everything after it. When one moves and the other doesn’t, the gap can sit after the click (the page, the offer, the price, the checkout) or in who the ad reaches, and the ad report alone can’t tell which. Start by reading where in the video people stop playing and what the ad promised at that second, then check whether the page keeps that promise. Whether the ad sells at all is a different question, and only a holdout answers it (see below).

My ad died after three days. Is it creative fatigue?

Ads Manager will tell you. When Meta believes your audience has seen the same ad too many times, the Delivery column shows “Creative limited” (cost per result above your past ads, but under twice as much) or “Creative fatigue” (twice as much or more) (Meta). If neither status shows, look elsewhere first: a short drop can come from the learning phase, the auction or a narrow audience. If it is fatigue, Meta’s own advice is a new ad with “a new image or video that is materially different from the original creative”.

Why don’t Meta’s numbers match my store’s?

Because they count different things. Meta credits a sale to an ad when it falls inside the attribution window set on the ad set; your store counts every order, whatever brought it in. Neither number tells you whether the ad caused the sale.

Did the ad cause the sale, or would it have happened anyway?

Only a holdout can say. Split people at random, keep one group from seeing the ad, and compare sales between the groups: the difference is what the ad added. Meta calls this incrementality, measuring a campaign’s “causal impact on your business while controlling for all other marketing efforts” (Meta). Attributed conversions and ROAS show association, not cause. A winning A/B test tells you which of two ads did better, not whether either one added sales.

How long should I test a creative before calling it a loser?

Until it reaches the sample size the test was planned for, decided before it starts, not a set number of days. Stopping when the numbers look good or bad breaks the statistics: standard results are “wholly unreliable” when the stopping point is chosen by watching the test (Johari, Pekelis and Walsh). We size each test from your traffic and the smallest effect worth finding, write the stopping rule into the plan before launch, and report “no difference” when that is the answer.

The terms are defined in the glossary, and how we work is on the method page.