How Long Should You Run Facebook Ads Before Judging Results?

Run a new Facebook ad campaign for 7 days before judging creatives and 14 days before judging the campaign's economics, with only delivery failures and total engagement flops killed earlier. The first days are the learning phase doing its exploration, and numbers from that period systematically mislead in both directions. Most money lost by small stores in Meta ads is lost to decisions made on day 2.

Why early numbers lie

Two forces distort the first days:

1. The learning phase is exploring. Delivery tests your ad across audience slices, and early cost per purchase reflects that exploration tax, not the settled state (learning phase explained).
2. Small numbers are noisy. At $15 per day, a single early sale makes day-2 ROAS look genius, and its absence looks like failure; neither is signal. Purchase data needs volume before it means anything.

The two-week window exists because that is roughly when a modest budget accumulates enough events to read.

The check schedule

Day 3: is it functioning?

Not "is it working": is it spending, delivering impressions, and free of disasters (rejected ads, broken links, zero clicks after meaningful impressions). Fix mechanics; judge nothing.

Day 7: which creatives earn attention?

Click-through rate stabilizes before purchase data does. Compare creatives against each other, pause the clear losers, and let winners carry the second week. Do not touch budget or targeting; edits reset learning.

Day 14: does the economics work?

Now judge cost per purchase against your break-even, over the full window rather than day slices (break-even math here). Keep, scale gently, or conclude honestly.

The early-kill exceptions

Waiting is not piety; some failures announce themselves immediately:

• No delivery or spend after a day: structural problem, fix it
• Rejected ads: fix the trigger (rejection reasons)
• CTR near zero after a thousand-plus impressions: the creative is failing at its first job; replace it without ceremony
• Broken landing page or tracking: every day of data collected is garbage; stop, fix, restart

The pattern: kill on mechanical failure, never on early economics.

What "judging" should conclude

At day 14, three honest outcomes:

1. Below break-even cost per purchase: keep and scale ~20% at a time (scaling guide).
2. Near break-even: the creative or the offer needs iteration, not abandonment; test the next angle.
3. Nowhere close, with real volume: the product-ad fit question is real. Diagnose the funnel layer (clicks? carts? checkout?) before blaming the channel.

Then the cycle continues: winners get refreshed before fatigue, and the next test launches. Judgment is a rhythm, not an event.

Frequently asked questions

How many days should I run Facebook ads before making changes?

Seven days before judging creatives against each other, fourteen before judging campaign economics. Earlier intervention is warranted only for mechanical failures: no delivery, rejections, broken pages, or near-zero engagement at volume.

How much should I spend before judging a campaign?

Enough for the data to mean something: at $10–$20 per day, the standard two-week window lands around $150–$300, which buys a readable test for most product price points.

Why do my ads perform badly in the first few days?

Learning-phase exploration plus small-sample noise. Delivery is still discovering who converts, and early cost metrics carry that tax; the settled performance emerges across the second week.

When is it okay to kill an ad early?

On mechanical failure: rejected status, no spend, broken links, or click-through near zero after a thousand-plus impressions. Economic verdicts before day 7 are guesses wearing dashboards.

What if results are still bad after two weeks?

Diagnose by funnel layer: no clicks is a creative problem, clicks without carts is a page or price problem, carts without purchases is checkout friction. Fix the failing layer and rerun; see our first-30-days plan for the structured version.