Identical Facebook Ad Sets, Different Results? Why

Duplicating an ad set copies its settings, not its delivery. The copy enters different auctions, at a different moment, with no history of its own, and when both copies chase the same people Meta lets only one of them bid each time. A gap between twins is the expected outcome, not a sign that one copy is broken. Read the gap as a delivery effect first, and use Meta’s A/B test whenever you need a fair comparison.

Why don’t two copies of the same ad set split delivery evenly?

Because Meta will not let your ad sets bid against each other. Meta’s page on auction overlap says that when two or more ads from the same advertiser enter the same auction, it chooses the ad with the highest total value to compete, and your other ads sit that auction out. Its page on overlapping audiences is more specific: it enters the ad set with the best performance history and prevents the others from competing.

Identical ad sets overlap completely, so every auction that one qualifies for, the other qualifies for too. Whichever copy gets a few early conversions builds the better history, wins more of the shared auctions, and gathers still more history. The other copy is left with the auctions the leader skips. Meta adds that auction overlap can stop an ad set from spending its full budget or getting enough results to exit the learning phase, and that too much overlap makes performance less predictable.

You can see this in Ads Manager: one copy spends steadily while the other underspends or reads Learning limited, even though their settings match.

Why does each copy learn something different?

Each ad set runs its own learning phase. Meta says the delivery system explores the best way to deliver an ad set when it is created, performance is less stable while it explores, and an ad set usually exits after about 50 results in the week after its last significant edit. Two copies explore separately. They do not share what they learn, so one can settle on a cheaper pocket of the audience while the other is still testing expensive ones.

Meta also warns against high ad volumes: when you create many ad sets, the system learns less about each one, and combining similar ad sets combines their learning. Two copies of one ad set split the signal that a single ad set would have used to exit learning faster.

Different launch moments meet different auctions

A copy launched on Tuesday competes against Tuesday’s advertisers, not Monday’s. Meta tracks this with a metric called auction competition change, which compares the competitive bid on a given day with the average of the previous 3 days and calls a change over 20% significant. Competition moving that much day to day is one reason a copy started later can look better or worse than the original from its first hours.

Edits reset one copy and not the other

If you touch one copy, the two stop being twins. Meta lists the significant edits that send an ad set back into learning: any change to targeting, creative or optimization event, adding a new ad, changing the bid strategy, and pausing for 7 days or longer. Large budget changes may also count. One edited copy is back at the start of learning while the other keeps its history.

How big a gap can chance alone create?

Bigger than most people expect at small volumes. Purchases arrive at random, and a common approximation treats the spread around a count as about its square root. If two copies would each average 25 purchases a week, each one’s count typically lands within about 5 either side, and the gap between them has a spread of about 7. A week of 20 against 30, a gap of about 1.4 times that spread, is within normal noise, even though, on equal spend, it reads as one copy costing 50% more per purchase.

At 100 purchases each, the same arithmetic gives a spread of about 14 on the gap, which is small next to 100. More results shrink chance relative to the signal. For the judgment of whether a whole new campaign is failing or just early, read how do I know if a new campaign is failing or just volatile? That post covers the day 1 to 7 decision. This one explains why two twins diverge.

How do I check which cause applies?

Work through these in Ads Manager before you pause anything.

1. Open Ads Manager, click Ad sets and select both copies. Compare Amount spent: a copy far under budget points to auction overlap.
2. Check the Delivery column. If one copy reads Learning limited and the other is Active, the copies are splitting results.
3. Add the Last significant edit column (Columns, then Customize columns). Different dates mean one copy restarted learning.
4. Add the Results and Reach columns and compare totals. Under about 50 results each, treat the gap as unproven.
5. Open Audiences, tick both saved audiences and click Actions, then Show audience overlap. Meta says the tool needs audiences of at least 10,000 accounts to be useful.

For a full walk-through of the overlap tool, read how do I check and fix audience overlap on Facebook?

How do I compare two versions fairly?

Use Meta’s A/B test instead of running copies side by side. Meta says an A/B test shows each version to a separate segment of your audience, makes sure nobody sees both, and keeps the audiences evenly split and statistically comparable. Meta specifically advises against informal testing, such as switching ad sets on and off by hand, because it leads to overlapping audiences and unreliable results.

1. In Ads Manager, click Ad sets, select the ad set and click Duplicate.
2. Choose New split test campaign, then pick the one variable you want to change.
3. Set the test to run for 1 to 30 days, and give both versions the same budget.
4. Raise the budget or length until the estimated power reaches at least 80%, which Meta recommends.
5. Choose the metric that decides the winner, such as cost per purchase, and click Create test.

Meta notes in its duplication guide that this option is being introduced gradually and may not be in every account yet; the Experiments tool is the other route. Test a real difference. Meta says versions that are too similar may not produce a confident winner, so an A/B test of two identical ad sets will usually end without one.

Should I keep both copies running?

Usually not. If the copies exist to “give the ad two chances”, you are paying for overlap. Meta’s own fix for overlap is to combine similar ad sets with their budgets, or to turn off the ones that are learning limited or have the fewest results and move that budget to the active one. Keep a second ad set only when it targets a different audience or placement on purpose. The learning phase guide covers how many results a week each ad set needs, and how many conversions does a Facebook ad set need to exit learning? explains Meta’s 50-result guide.

FAQ

Does duplicating a winning ad set copy its results?

No. The duplicate is a new ad set with no performance history, so it starts in the learning phase. If it targets the same people, Meta lets only one of your ad sets into each shared auction, and the original’s history usually gives it the edge. Expect the copy to deliver differently, often worse at first.

Can I avoid overlap by putting the copy in another ad account?

Not reliably. Meta says running ads from separate ad accounts may not help you avoid overlap, because its delivery system may identify the accounts as belonging to the same advertiser. Meta suggests consolidating campaigns into fewer accounts to manage overlap instead.

Is a 30% difference between identical ad sets significant?

Not by itself. It depends on the counts behind it. At 20 purchases against 26, chance alone explains the gap. At 400 against 520, it is very unlikely to be noise. Check the result counts, the Last significant edit dates and spend before you trust any percentage.

Why did my duplicate ad set beat the original?

The original may be carrying an edit that restarted learning, a fatigued audience, or a higher-competition launch day it never recovered from. The duplicate may also have caught a lucky early run of conversions that then won it the shared auctions. Confirm with an A/B test before you move budget.

Sources

Sources: Meta Business Help Center, Understand auction overlap, About overlapping audiences, About the learning phase, Significant edits and learning phase, About learning limited, Auction competition change, About A/B testing, Create an A/B test by duplicating an ad set or ad and Tips for improving your A/B tests (checked 2 Oct 2026).