Published on March 10, 2026
One of the most common questions advertisers ask is: how long should I let an ad run before making a decision? Kill it too early and you might have cut a future winner. Let it run too long and you waste precious budget on a loser. This guide gives you data-backed frameworks for making the right call at the right time.
When you launch a new ad, Facebook's algorithm enters what Meta calls the "learning phase." During this period — typically 3-7 days or until you accumulate about 50 optimization events — the algorithm is testing different audience segments and placements to find the best performers.
Making changes during the learning phase resets the process entirely. This means pausing and restarting an ad, changing the budget by more than 20%, or editing the creative all force the algorithm back to square one. Patience during this phase is essential.
After the learning phase, use this framework: If your cost per acquisition (CPA) is more than 2x your target after spending at least 2-3x your target CPA, it is time to kill the ad. If CPA is between 1x and 2x your target, the ad has potential — iterate on the hook or offer. If CPA is at or below target, congratulations: you have a winner ready to scale.
Never make decisions based on less than 1,000 impressions. Small sample sizes lead to wildly misleading metrics. A $5 ad that got 200 impressions and one sale is not a winner — it is a statistical fluke.
Across ecommerce verticals, the average winning ad runs for 30-90 days before creative fatigue sets in. High-ticket items and B2B offers tend to have longer lifespans because the audience is smaller and frequency builds slower. Fast fashion and trending product ads may fatigue in as little as 14-21 days.
Monitor your frequency metric closely. When frequency exceeds 2.5-3.0 in a prospecting campaign, performance typically starts to decline. At that point, refresh your creative while maintaining the winning hook angle.
When you have a confirmed winner, scale gradually. Increase budget by no more than 20-30% every 48 hours to stay within the algorithm's comfort zone. Alternatively, duplicate the ad set into a new campaign with a higher budget — this sometimes avoids the learning phase reset entirely.
Keep your winning creative running even as you test new variations. Many advertisers make the mistake of pausing winners to focus on new tests. Instead, run winners and tests in parallel so your proven performers keep generating revenue while you search for the next hit.
Explore winning ads in the ecommerce niche →
Explore winning ads in the courses niche →
Explore winning ads in the coaching niche →