The question showed up under a Meta ads video this week, almost word for word the way it shows up in every ad account eventually: running a successful campaign for 60 days, crushing it for the first 35, and now results have dropped and CPL is climbing day by day. Should I duplicate the winning creative into a new campaign or ad set?
The instinct is to treat the campaign as broken and rebuild it. Most of the time nothing broke. The campaign did exactly what it was told — find the people most likely to convert — and it has now found most of them. What is left is the same people again, at a higher price.
No — duplicating the winner makes it worse. A CPA that climbs after a good first month usually means the audience is running out, and two columns show it: frequency going up while first-time impression ratio goes down. New creative first, a broader audience second, duplication never.
The two columns
Before changing anything, add two columns to the ad set view and look at them by week.
- Frequency — average impressions per person. If it has climbed from around 1.5 to 3 or more over the same weeks that CTR slid, the same people are seeing the ad repeatedly, and each genuinely new impression costs more to win.
- First-time impression ratio — Meta defines it as the number of accounts reached for the first time divided by impressions, per day. Meta's own guidance on the metric is the diagnosis in one sentence: if it is dropping significantly along with the ad set's performance, it may be time to broaden your audience or find a new one.
Put the two together with the CPL curve. Rising frequency, falling first-time impression ratio, falling CTR, rising cost — that is one story, and the story is saturation, not a bad ad. A bad ad looks different: CTR was never good, and frequency is still low.
Why duplicating the winner makes it worse
Duplicating the winning ad into a new ad set feels like a fresh start. It does two things, both bad. The copy targets the same pool, so the people who were tired of the ad in one ad set now see it from two. And a new ad set starts its own learning phase: Meta's help centre puts the exit at about 50 results in the week after the last significant edit, and says ad sets in learning are less stable and usually run a higher CPA. You have paid the learning tax a second time to reach the same exhausted audience.
The same page lists what resets learning: editing the ad, ad set or campaign in ways that meaningfully change delivery, and frequent budget changes. Which is why the panicked week-five sequence — raise the budget, then cut it, then duplicate — reads to the delivery system as three restarts in a row.
What to change instead
- Keep the ad set. Its learning is the asset; do not throw it away.
- Add two or three new creatives to it that differ in angle, not in colour — a different opening problem, a different proof, a different format. The people who ignored the old ad have not necessarily rejected the offer; they have rejected that ad.
- Let the old creatives fade rather than pausing them the same day. Pausing everything at once is itself a significant edit.
- If frequency stays high after the creative refresh, the audience is the constraint, not the creative. Then broaden — that is what Meta's own note on first-time impression ratio tells you to do — rather than layering a lookalike on top of the same seed.
- For lead gen, read lead quality over the same window. The cheap first 35 days often took the high-intent slice of the audience first; a rising CPL on the tail looks identical from the outside and is partly just the tail.
Rule of thumb: if frequency is up and first-time impression ratio is down, you have run out of people, not out of luck. New creative first, broader audience second, duplication never.
Where an agent fits
This is the kind of drift a person catches in week seven and a daily reader catches in week five. It is also the case where the cheap moves are safe and the expensive move is the tempting one. Everix reads the account every day and handles this case the way the post does: letting losing ads fade and pausing what is clearly done happen on their own; anything that raises the budget waits for a human, because the week-five reflex to spend more on a saturated audience is exactly the reflex that should have a gate on it.
If you are in week five right now, do the two-column read first. It takes a minute, and it decides whether you are refreshing creative or shopping for an audience — two different jobs that a duplicate would have done neither of.