How Many Ad Creatives Do You Need to Find a Winner?
Want my team to build this for you? Work with Advergent. Rather build it yourself? Join the coaching waitlist.
You don't need a fixed number of ads to find a winner, because you're looking for a winning angle. In my accounts: 5 to 10 ads per angle when an ad set gets about $100 a day, 20 to 30 new ads a week up to $1,000 a day, and 50 to 100 once one angle gets $1,000 a day.
The count is the easy part. What decides whether those ads find you anything is what you count as a creative, how much each test spends before you judge it, and whether the creative is even what's broken. The campaign structure underneath is in how to scale Meta ads.
Why do you test angles instead of individual ads?
Because the angle decides who stops scrolling. An ad has two levers: the angle, meaning the persona, their situation and the problem your offer solves, and the format it's shot in. No format takes a business from $0 to $100K a month on its own. A winning angle can, and formats come after it.
So when you count creatives, count angles first. My testing campaign runs one angle per ad set on broad targeting, so the ad set tests the idea and the ads inside it test the execution. Launch 30 ads that all make the same argument and you haven't run 30 tests. You've run one test 30 times.
That matters more since Meta's Andromeda update, because ads with the same hook, body, CTA and message now read as overlapping and your CPMs go up. One creative with 30 hooks bolted onto it doesn't work anymore. What works is one concept per audience, with different hooks, holds and angles inside it.
Angles come from your buyers, not a brainstorm. On a small budget I'd stick to 3 or 4 avatars most likely to buy and get specific about each: why they'd buy, the pain they live with now, what they want, and the proof that would land with them. Let's say you sell sales training to agency owners. The owner still taking every sales call themselves, the owner whose closer stopped closing, and the owner about to hire a first closer are three different angles, even though the program is the same.
The best first angles already convert somewhere. If a piece of content went wild on your Instagram, there's a very high chance the same message works in your ads. Then pull 3 to 5 audience types out of your reviews, comments, emails and DMs.
How many ads should you put in each ad set?
It depends on the budget behind the angle. In my accounts, an ad set getting $100 a day carries 5 to 10 ads. If your total spend is up to $1,000 a day, 20 to 30 new ads a week is the sweet spot. With $1,000 a day behind one angle, launch 50 to 100.
Volume matters because of speed. Launch 50 ads in a week and you get the learning that 10 ads a week would give you five weeks later, while a faster competitor is already scaling what they found.
But volume only helps if each ad gets enough spend to tell you something. On a smaller budget, keep the count lower and go deeper on quality, because a big batch sharing a small budget means most of the ads never get a fair chance.
There are two ways to feed the account, and both work. The volume path is 30 to 50 new creatives a week, for a team with editors and a production system. The message diversity path is 15 to 20 completely different concepts a month, each built on a different reason to buy, for a smaller team or a solo operator. I'd start with message diversity to find what resonates, then switch to volume once you know what's worth making more of.
Once an account is scaling, 20 to 30 net new creatives a week is the bare minimum I hold clients to, and me and my team produce that volume for them.
What counts as a new ad creative?
A new creative is a net new, individualized ad: a new script, a sharper callout to the persona, or a more specific sub-angle. A re-edit of the same footage, a new thumbnail, or 30 hooks on one body doesn't count, because the algorithm sees the same idea again and so does the buyer.
This is where creative counts get inflated. A team re-edits the 10 ads it recorded six months ago, relaunches them at the same personas and reports 10 new ads. Meanwhile a competitor pushes 25 to 30 fresh angles at new personas and opens up more of the market with every launch. On paper both are testing. Only one of them is finding new buyers.
To make a real new ad, go back to the numbers behind the last one. What did the script fail to say? Was the callout speaking to the wrong persona, or to too many at once? Is there a sub-angle that talks to a narrower slice of the audience with one specific problem? Answer those and you have a new ad. Change the music and you have the old one again.
Statics count if they're genuinely good. I test video and statics at the same time and let cost per booked call pick the winner. Bear in mind that if you sell B2B or info products, you'll need video eventually, because cold strangers need the credibility a well-edited video builds. And an ad workflow made entirely of AI-generated ads isn't something I'd expect to carry an account right now.
How much should a creative test spend before you judge it?
Let each angle's ad set spend three times your target cost per qualified call, then compare it with the target using a 20% grace range. At a $100 target, that's $300 per angle before you decide anything. Judge earlier and you're reacting to noise, which is how good funnels get killed in week two.
Notice the unit: cost per qualified call, not cost per call or cost per lead. An angle that books plenty of cheap calls with people who can't buy isn't a winner. It's the most expensive ad in the account, because it eats closer time and teaches Meta to find more people like them. I go deeper on that in why cheap Meta leads cost you more.
Here's how I read an ad set once it has spent its 3x:
| What you see after 3x spend | What it tells you | Next creative move |
|---|---|---|
| At target or within 20% above it ($120 on a $100 target) | The angle works | Keep it, raise the budget 10 to 30% every one to three days, and start horizontal scaling: 80% of new ads on this angle, in new formats |
| Above the grace range | The angle or its execution is off | Kill it, then rewrite from the numbers: script, callout, sub-angle |
| No qualified calls at all | The angle didn't land with buyers | Kill it and pull the next angle off your list |
| Cheap bookings, but the calls don't qualify | The ad is attracting the wrong person | Say who it's not for in the ad, and check the event you send back to Meta |
Read these numbers weekly, never daily, because daily numbers on a cold funnel are noise. On a brand-new ad account and pixel, give the account at least $1,400 over 14 days before you decide Meta doesn't work for you.
The 3x rule is also why your budget caps how many angles you can test at once. Let's say you spend $600 a day and $300 of it sits in a scaling campaign. That leaves $300 for testing, which is three angles at $100 a day each. Split the same $300 across more angles and each one takes longer to reach its 3x. Fewer angles, funded properly, find a winner faster than many angles funded thin.
What if none of your ad creatives are winning?
Check whether the creative is the problem before you make more of it. If fewer than about 35% of clicks turn into VSL viewers, the ad promised something the page didn't deliver. If viewers don't apply, the page or VSL failed. New ads can help with the first. They won't fix the second.
These are my floors from my accounts, walked through on 1,000 clicks: about 350 people should watch the VSL, and about 10% of those, 35 people, should fill in the application. If the drop is between the click and the VSL, there's a gap between what the ad promised and what the page says, so one of them has to move closer to the other. If the drop is between the VSL and the application, no amount of new hooks fixes a page that doesn't sell the call. I cover that case in what to do when your VSL gets watched but nobody books.
The number I judge this early phase on is cost per application, not cost per click or cost per lead, because an application takes effort and effort is the first honest signal in the funnel. If it's unstable or absurd, your script and your page are broken, and touching audiences or budgets won't change that. And don't expect validation data from $100 a day. If your target is $80 an application and you need 10 a day to see a pattern, that's $800 a day. That's a hypothetical, but run that math before you decide your creative failed.
We launch every funnel on a landing page structure we've run many times, as a control, so when no winner shows up I already know the layout isn't the variable and the next round of creative goes at the messaging.
And sometimes it isn't the creative or the page. If you'd take literally anybody who can pay, you don't have an offer worth filtering for yet, and no ad count will find a winner for it. Get five customers by hand, define who they are, then come back to ads.
What do you do once you find a winning creative?
Shift 80% of your new ads onto the winning angle and keep 20% on fresh angle tests. Then make that angle in every format you haven't tried: talking head, carousel, split screen, B-roll, skit. Some versions will fail, some will match the winner, and some will beat it.
This is horizontal scaling: more formats behind the angle that's working, instead of only more budget behind the ad set. You'll use both, but I lean on horizontal first, because you've already paid to learn what this persona responds to, so a new format on a proven angle is a smaller bet than a fresh angle test. It's also what makes 20 to 30 ads a week manageable, because most of the week's work becomes new versions of something you know converts.
Keep the 20% running, though. Every angle hits a ceiling, and you'll see it when more budget drags cost per qualified call up or thins call quality. If the 20% has been running all along, the next angle is already waiting. Stop feeding the account new creative and performance drops again. How that plays out as spend rises is in how to scale Meta ad spend without lead quality collapsing.
If you want help reading which of your angles is actually working, that's the kind of thing we work through in coaching. Or start with the guide below: it has my five-stage scorecard and how to tell Meta to buy qualified booked calls instead of form fills.
FAQ
How many ads should be in one Meta ad set?
In my accounts, 5 to 10 ads per ad set when the ad set gets around $100 a day, and 50 to 100 when an angle has $1,000 a day behind it. Every ad in the set should be a different execution of the same angle, not a re-edit.
Should I test hooks or angles first?
Angles. Thirty hooks on one creative is still one idea, and since the Andromeda update Meta treats ads that sound alike as overlapping and charges you more for them. Find the angle first, then test formats and hooks around it.
How do I know an ad is a winner and not a fluke?
Wait until its ad set has spent three times your target cost per qualified call, then check it against the target with a 20% grace range. Read it on weekly numbers, not daily ones, because daily numbers on a cold funnel are noise.
Do static ads count toward creative testing?
Yes, if they're genuinely good, and they're a cheap way to test angles early. Run them alongside video and let cost per booked call decide, but expect to need video to build credibility if you sell B2B or info products.