Variations vs Iterations: How to Evolve a Winner Without Triggering Meta's 'Similar' Penalty

Make a winning ad, then launch four near-identical copies of it, and Meta will now spend on roughly one of them. The other three barely get a look in.

That's the part of the Andromeda update that quietly changed how creative iteration works. The system got much better at recognising when two ads are basically the same thing, and when it decides they're "similar", it stops splitting spend across them. It picks the one with early traction and funnels budget there, leaving your "variations" to starve.

So the old trick (find a winner, swap the headline, swap a callout, relaunch, repeat) doesn't open up new spend anymore. Meta just shrugs and says "I've already got that one".

Here's the thing though. Iteration still works. Evolving a winner is still one of the highest-impact things you can do in an account. You've just got to do it in a way that reads as genuinely new to the algorithm, not as the same ad wearing a different hat. So let me pull apart the three things people muddle together, because the difference is the whole ballgame.

Three things people call "iteration" (only two of them work)

When a founder tells me they're "iterating on the winner", they usually mean one of three things. They're worth naming separately.

The lazy reskin. Same image, same person, same shot, same everything, with a new headline slapped on top or a different price callout in the corner. This is what most people do, and it's exactly what Andromeda now collapses. To Meta it's not a new ad, it's the same ad with a sticker on it. You'll launch six of these and watch one get all the spend. Wasted effort.

The variation. Same winning idea, genuinely new execution. You keep the thing that's working (often the headline or the core message) but you rebuild the visual from scratch so it looks new to the platform. Different scene, different person, different style. The message carries over, the creative does not. This reads as net-new, so it gets its own shot at spend.

The iteration. You take the winner and deliberately change a bigger variable to chase a different person or a different reason to buy. Not a cosmetic tweak, a real swing. A new angle, a new audience, a new way into the same product. Done with intention, this is where your next winner usually comes from.

The middle and the bottom of that list work. The top of it used to, and doesn't really anymore. The trap is that the lazy reskin feels like iteration. You're "making more of the winner", you're busy, it looks like work. It just isn't the work that moves anything.

Change the biggest variable, not the headline

Here's the principle I'd tattoo on every creative team's wall: when you evolve a winner, change the biggest variable you can, not the smallest.

People instinctively reach for the smallest change because it's the safest and the quickest. New headline. New first three words. New colour on the button. The problem is the smallest change is also the one most likely to read as "same ad" to Meta, and the one least likely to reach a genuinely new pocket of people.

So flip it. Start at the top of the ladder and work down only when you've exhausted the bigger swings.

The biggest variable of all is usually the avatar. Who is this ad actually for? I watched a brand selling a premium oral-care product burn weeks testing headline after headline aimed at one audience: younger blokes who wanted fresher breath for dates. The ads all blurred together and nothing scaled. The fix wasn't another headline. It was changing who the ad spoke to entirely, over to an older, higher-spending customer with the same need for a totally different reason. Same product, completely different person. That one change did what twenty headline swaps couldn't.

Below the avatar, in rough order of size:

• The angle. The same product sold on a different promise. "Save time" versus "look the part" versus "stop the embarrassment". Each one pulls in a different crowd.
• The desire. Two people might buy the same supplement, one for better sleep, one for fewer headaches. Speak to one desire per ad, not a mush of all of them.
• The awareness stage. A blunt "20% off" static for someone who already knows you is a different ad from a five-minute explainer for someone who's never heard of the problem, let alone your solution.
• The format and scene. Studio shot versus shot-on-phone, demo versus testimonial, one setting versus another.
• Then, last, the words on screen. The headline. The callout. The bit everyone starts with should be the bit you reach for last.

Define the bigger variable first, and the visuals fall out of it. The avatar and the angle decide what the ad should look like. The headline is a finishing touch, not a strategy.

What good evolution actually looks like

Let me make this concrete, because "change the biggest variable" is easy to nod along to and hard to do.

Say your winner is a clean before-and-after for a skincare treatment, a woman showing her results, and it's been running profitably for weeks. How do you make more without tripping the similar penalty?

The lazy version (don't): same woman, same shot, same lighting, new headline. Or the same woman from a slightly different camera angle in a different colour top. To you that feels different. To Meta it's the same ad. These collapse onto each other and one starves the rest.

A real variation: keep the winning message, rebuild the visual. Now it's a different person entirely, or it's the treatment being performed rather than the result being shown, or it's a hands-on demonstration instead of a glamour shot. The idea that won carries over. The execution is new enough that the platform treats it as a fresh ad and gives it room.

A real iteration: point the same product at a new buyer. The original spoke to women chasing clear skin, so now you build one for a man getting the same treatment, and another aimed at the curiosity angle ("what actually happens in a session") rather than the result. Different person, different reason, same product. That's a swing, and swings are what find your next winner.

The tell is simple. If you put the new ad next to the old one and a stranger would say "oh, that's basically the same ad", you've made a reskin. If they'd say "different ad, I can see they're the same brand", you've made a variation or an iteration. Aim for the second reaction every time.

The net-new test I'd run before launching a batch

Before a fresh batch of variations goes live, I run one quick gut-check, and it saves a lot of wasted spend.

Lay the new creatives next to your current winner and ask: what is the one big thing that's different here? Not "what did we change", but "what's the single biggest variable". If the honest answer for an ad is "the headline" or "the background colour", it goes back. It's a reskin, and Meta will treat it as one. If the answer is "a different avatar" or "a different angle" or "a genuinely new scene", it's cleared to launch.

The point of the test is to force the change to be big enough to matter. Each ad in the batch should be reaching for a different person or a different reason, not crowding into the same lane as everything else. If two of your "new" ads are both just a pretty shot of the same customer with the same message, kill one. They'll only cannibalise each other's spend.

This isn't about volume. I've seen teams test fifty cosmetic tweaks a week and find nothing, and teams test two real swings a month and find a winner with each one. Quality over quantity, every time. One deliberate iteration that opens a new audience beats ten reskins that Meta merges into one and ignores.

So next time you've got a winner and the instinct kicks in to "make more of it", stop before you touch the headline. Ask yourself who else this product is for, and what entirely different reason they'd have to buy it. The answer to that question is your next ad. The headline can wait.