You don't need more ads. You need more signals.
By Denzel Geng · Updated October 5, 2026
You can put 30 ads into an ad set and still give Meta very little new information. If 10 are the same UGC with different hooks, 10 are the same concept with different CTAs, and the rest are minor visual changes, you haven't created 30 ideas — you've created 30 executions of a few ideas. This is the difference, why it matters more after Andromeda, and where genuinely different concepts come from.
Creative diversity is how many genuinely different ideas — distinct angles, triggers, and avatars — your account is feeding Meta, as opposed to creative volume, which just counts files. Two ads with different hooks, CTAs, or colors can still be one idea. Meta's own advice is to diversify by concept, because people shop for the same product for different reasons — extra files of one idea don't add a new reason.
See the concepts winning in your niche- A volume problem vs a diversity problem
- What people mean by "Entity ID" — and what it isn't
- The first-frame grid test
- Does the next ad add new information?
- The biggest mistake: starting with the hook
- Where new angles come from, part 1: customer voice
- Don't rank customer voice by likes
- Where new angles come from, part 2: competitor ads that survived
- Two research streams, one usable brief
- Stop measuring creative output. Measure creative signals.
A volume problem vs a diversity problem
A lot of teams think they have a creative volume problem. They don't. They have a creative diversity problem.
The usual ways teams create more "creative" — same video with a different headline, same UGC with a different CTA, same product shot on a different background — are easy to produce because you're changing the surface of the ad. You're not changing the underlying idea.
That distinction matters even more after Andromeda. Meta's engineering post on Andromeda describes a retrieval system built for a fast-growing volume of creatives, says that, as a planned next step, the architecture is expected to deliver "a more diverse set of ad candidates", and says "increased ad diversity can improve people's experience with ads and drive better advertiser outcomes." Meta's best-practice page puts the creative side plainly: "Develop ad creatives based on a variety of motivations to appeal to different audiences" (Meta Performance 5). Our read is that the creative now carries more of the audience signal. So the goal isn't to keep feeding the account more files. The goal is to keep feeding it meaningfully different signals — and that's a research problem, not a production problem.
What people mean by "Entity ID" — and what it isn't
You've probably seen people talk about Entity ID in this context, so one clarification first: Entity ID is not an official Meta term. There is no "Entity ID" field in Ads Manager. It's a concept advertisers use to describe a pattern they report — you can create 30 different Creative IDs, but that doesn't mean Meta treats all 30 as independent ideas.
The exact mechanics aren't public, so don't treat Entity ID as a literal number you can find in your account. The useful takeaway is much simpler: two ads can have different IDs and still be very similar. Our read, and what operators commonly report, is that near-identical ads don't expand delivery the way distinct concepts do; they tend to compete with each other for it. The closest official thing is Meta's creative diversity label in Ads Manager, in development, which Meta says is calculated "by comparing the images and video thumbnails in your ad set to each other to determine similarity" (Meta Business Help). The page names only images and video thumbnails, not copy, script or angle.
Compare that with a studio product demo becoming a customer talking from her kitchen, or a rational product explanation becoming a personal story about why every previous solution failed. Those changes move the context, the narrative, the visual language, and often the audience the ad speaks to. That's what real creative diversity looks like.
The first-frame grid test
Take the first frame of every ad you're running and put them in a grid. Show it to someone who had nothing to do with making them and ask which ones feel like the same ad. You're not reverse-engineering Meta's system — you're measuring how much diversity actually exists in your own library.
If 30 ads collapse into four obvious groups, you don't have 30 concepts. You have four concepts wearing different outfits.
Does the next ad add new information?
For the count itself — how many ads to run at once and how many to add each week — use the creative testing playbook; this page is about what counts as a different ad.
The better question is not "do I have enough ads?" It's "does the next ad add new information?" If it does, make it. If it's another version of something you've already tested, that time is better spent developing a new concept. Ten genuinely different ideas can teach you more than thirty variations of one.
The biggest mistake: starting with the hook
This is where creative teams get stuck. The request comes in as "we need 20 new hooks," the hooks become UGC scripts, a few CTAs and formats get added, and by the end there are 30 new ads. On paper the volume looks great. In reality the team may still be testing one idea.
Creative diversity starts further upstream. At the concept level you decide what you're actually saying: the angle, the trigger you're tapping, the problem you're solving, the mechanism that makes your solution different — and who the idea is actually for. That last one is your avatar, and it matters more than most teams realize. A beginner thinks "I don't know where to start." Someone who has already tried five solutions thinks "why is none of this working?" Same product — completely different problem, motivation, and message.
Only after the concept is clear do you get into execution: hook, format, script, visual treatment, CTA, landing page. Those matter, but they're downstream. If the underlying angle doesn't change, changing the hook five times doesn't give you five new ideas — you're optimizing inside the same concept. This is why teams can spend weeks "testing creative" without learning much: they keep changing the bottom layer while barely touching the top. The difference between the two layers, shown with real ads, is laid out in ad angles vs hooks.
Where new angles come from, part 1: customer voice
You can ask a copywriter to brainstorm hooks all day, but eventually you run out of useful things to say — because the source material isn't there. New ideas come from outside the creative team: what customers are saying, and what competitors have already proven. Start with the customers. The full method for turning these conversations and competitors' surviving ads into angles is in the ad angles guide.
Find where the conversation actually happens
Don't start by opening Reddit and typing your category keyword. First figure out where people in your market talk about the problem — Reddit, Facebook Groups, Discord, YouTube comments, Quora, product reviews, niche forums. It differs by category.
Search for different kinds of conversations
If you have a brand, search the product name — but also people switching away from it, specific usage problems, category pain points, and comparison threads. "Is Brand X good?" tells you almost nothing. "Why did I stop using Brand X?" gives you an objection. "Brand X vs Brand Y" tells you what people weigh at the moment of purchase.
Open the post — the gold is in the comments
Search snippets tell you what's worth opening; the valuable material is in the full post and its comments. A post may ask a simple question while someone in the comments explains exactly why they rejected three solutions. That comment is often the line you actually want.
Weigh repeated behavior over popularity
Five independent people recommending the same product tells you more than one comment with a thousand upvotes — you're seeing repeated behavior, not amplification.
Don't rank customer voice by likes
Here's where teams lose the signal a second time. You collect 100 customer quotes, sort them by upvotes — and now you're optimizing for popularity instead of insight. One viral thread can generate dozens of comments reacting to the same event, which makes your research look far more diverse than it is.
Instead, cluster the quotes by theme. "I don't know where to start." "There are too many options." "Every tutorial assumes I already know what I'm doing." Different sentences — same underlying problem: beginners feel overwhelmed before they start. That's an angle.
Then count independent conversations, not comments. Ten comments in one viral thread aren't ten signals. Ten people in ten different threads saying roughly the same thing — that's interesting. This is the moment customer research turns into creative strategy: each solid cluster hands you a trigger, a pain, or an avatar.
Where new angles come from, part 2: competitor ads that survived
The other side of the system is the Meta Ad Library. There are thousands of ads in there, and most aren't worth studying. The ones that matter show two signals. Lifespan: the longer the window between when an ad first appeared and when it was last active, the longer the brand kept backing the idea. That isn't proof it made money (Meta publishes no spend or results for regular ads), but it's worth a closer look. And variation: a brand running UGC, founder, and static executions around the same underlying concept isn't keeping one lucky ad alive — it's investing in the concept.
Don't save the ad. Deconstruct it. For each interesting ad, record the hook and its type, the problem being introduced, the mechanism that explains why the product works, the proof, the offer, the CTA, and the persona. Then read the structure as a whole: Hook → Problem → Mechanism → Proof → Offer → CTA.
Do this across 20, 30, 50 ads and patterns appear — different brands, actors, and visual styles running on the same underlying structure. Name the patterns and store them. Now your team has a library of proven ways to build ads instead of reinventing the structure every time a brief opens.
Two research streams, one usable brief
Customer research answers: what should we say? Competitor research answers: what are proven ways to say it? Combine them and "make another UGC ad" becomes a brief your team can actually execute:
Avatar: experienced user who has already tried existing solutions. Trigger: frustration after repeated failed attempts. Angle: the problem isn't your effort — it's the mechanism. Execution: problem → failed solutions → new mechanism → proof → CTA.
Run the first-frame grid test on a competitor's live ads in the Ad Library: a brand running 20 or 30 creatives can turn out to be running a handful of concepts. Different actors, different colors, different hooks. Same idea.
Stop measuring creative output. Measure creative signals.
The old question was "how many ads did we make this month?" The better sequence: how many genuinely different concepts did we add? How many different avatars did we speak to? And only then — how many executions should we build from those concepts?
The order matters. Concept first, execution second. More ads don't automatically mean more learning; more genuinely different signals give you more opportunities to learn.
Common questions.
Creative diversity is the number of genuinely different ideas — distinct angles, triggers, and avatars — running in your account, as opposed to creative volume, which counts files. A concept change moves the underlying idea (who it's for, what problem it names, what mechanism it claims); an execution change moves the surface (hook, CTA, format, color). Meta's own advice is to diversify by concept, because people shop for the same product for different reasons; our read is that near-identical ads mostly compete with each other, while genuinely distinct concepts reach new buyers.
It changes what counts, not the count: the count depends on budget and our creative testing playbook covers it, but only conceptually different ads add a test, so thirty variations of one idea are still one test.
Entity ID is not an official Meta term and doesn't appear anywhere in Ads Manager. It's the name advertisers use for a pattern they report: very similar creatives seem to behave like one idea, however many Creative IDs they have. We couldn't find a Meta source for it and the mechanics aren't public, so treat it as a mental model, not a field to optimize — the practical implication is simply that surface-level variations don't add a new reason to buy.
No — but volume without diversity is. Meta's own advice is to diversify by concept and to develop ad creatives based on a variety of motivations to appeal to different audiences, so what compounds is the number of meaningfully different ideas you feed Meta, not the file count. Teams still shipping 30 ads a month do fine when those ads span genuinely different concepts and avatars; they stall when the 30 collapse into three ideas.
Run the first-frame grid test: screenshot the first frame of every running ad, arrange them in a grid, and ask someone outside the team which ones feel like the same ad. Groups that form instantly are one concept each. A stricter check: write down each ad's angle (trigger + problem + mechanism + avatar) in one line — if two ads produce the same line, they're the same concept in different outfits.
Feed the machine different ideas, not more files.
Every team can produce more ads. Very few can keep finding the next genuinely different idea.
ViralMojo continuously pulls customer conversations, competitors' surviving ads, and emerging pain points from your niche, then turns them into ads of your own — static, AI video, your product in a winning ad, or a real creator — so your next 10 ads are 10 different ideas, not one idea in 10 outfits.
- Customer-voice clusters from real conversations — angles ranked by independent signals, not upvotes
- Competitors' long-running ads broken down to the concept layer: hook, mechanism, proof, persona
- Each concept aimed at a named avatar and trigger, with the test variable spelled out
- Works cold from your URL — no ad-account history required
Plans from $99/mo, flat — see ViralMojo pricing.