Turn ads off too early and you never find winners. Leave them on too long and you burn budget. The right answer depends on data, not on days, but there are clear rules of thumb that keep you out of both traps.
Understand the learning phase
When you launch or significantly edit an ad set, Meta enters a learning phase while it explores who responds. Results are often volatile during this time. Meta generally aims for around 50 optimisation events in a week to exit learning, so very small budgets can stay in learning for a long time.
Judge on data, not time
A useful approach is to set a spend threshold based on your target cost per purchase. Give each new ad enough spend to have a fair chance, then decide.
| Spend so far | Typical decision |
|---|---|
| Under 1x target CPA with no add to carts | Wait unless CTR is very poor |
| 1 to 2x target CPA with no purchase | Consider turning off, check CTR and landing page |
| Purchases at or below target CPA | Keep running and prepare to scale |
| Purchases well above target after 3x CPA spend | Turn off or rework the creative |
Early signals you can read
- Hook or 3 second view rate tells you if the opening works
- Link click-through rate tells you if the ad creates interest
- Cost per add to cart hints at purchase performance
- Landing page view rate shows whether clicks reach the site
When to scale
Scale ads that hold a profitable cost per purchase for several days. Increase budgets gradually, or duplicate the winner into a new campaign with a higher budget. Keep feeding new creative so you are not dependent on one ad.
When to turn ads off
- Cost per purchase stays above your break-even point after enough spend
- Frequency keeps rising while click-through rate falls
- The offer has ended or stock has run out
- Comments are turning negative
Why long-running ads matter for research
This logic is why running time is the best public signal of a winning ad. Brands make the same decisions you do. If a competitor's ad has been running for 90 days, it has almost certainly passed their kill rules many times over. Filter the ad library by running time to find those proven winners quickly.
A simple weekly routine
- Monday: review last week's ads against your kill and scale rules.
- Tuesday: brief new concepts based on winners and competitor research.
- Wednesday to Thursday: launch new tests.
- Friday: check early signals and fix tracking or landing page issues.
Summary
Give ads enough spend to learn, judge them against clear cost targets rather than a fixed number of days, scale winners gradually, and use running time in the ad library as a shortcut to see what has already passed other brands' tests.
Search millions of real Facebook and Instagram ads with video and image previews. Free to start, no card needed.
Frequently asked questions
How many days should I run a Facebook ad?
Judge on spend and results rather than days. Many advertisers give new ads three to seven days if budget allows.
How long is the learning phase?
It depends on how quickly the ad set collects optimisation events. Meta generally targets around 50 events per week.
Should I turn off ads on weekends?
Usually no. Turning ads on and off disrupts delivery. Let Meta optimise across the week.
Can an old ad stop working?
Yes. Ad fatigue causes performance to decline as the same audience sees it repeatedly.
What does a long-running competitor ad tell me?
That it has likely been profitable for that brand, making it a strong reference for your own creative.