---
title: "Meta ads creative diversity: what Andromeda rewards (and what it ignores)"
description: "Meta groups lookalike ads as one creative. What creative diversity actually means under Andromeda, how many different concepts to run, and what to do today."
url: https://dumbcmo.com/blog/meta-creative-diversity-andromeda
category: "Paying to be seen"
tags: ["meta ads", "facebook ads", "instagram ads", "andromeda", "creative diversification", "advantage+", "how-to"]
published: 2026-11-06
updated: 2026-11-06
reading_time: 12 min
author: "The DumbCMO"
---

# Meta ads creative diversity: what Andromeda rewards (and what it ignores)

> **TL;DR:** Andromeda can shortlist more ads, but Meta groups ads that look or feel alike and treats them as one creative. Tiny tweaks don’t count. Genuinely different concepts do. On a small budget, start with 4–6 different ideas, not 20 copies of one photo.

[Notebook](https://dumbcmo.com) › [Blog](https://dumbcmo.com/blog) › Paying to be seen › Meta ads creative diversity: what Andromeda rewards (and what it ignores)

We already covered [how Andromeda works](/blog/meta-andromeda-2026-how-it-works): it’s Meta’s first filter, the step that shortlists a few thousand ads out of millions before the clever ranking models ever see them. This note is the creative half of that story.

Because Meta’s own advice, repeated through 2025 and 2026, keeps landing on the same word: **diversity**. Not “make more ads.” Make ads that are actually different.

Agencies have turned that into “you need 50 creatives a week or Andromeda will punish you.” That’s not what Meta wrote. Let’s separate Meta’s words from the LinkedIn carousel.

## In normal-people words

**Creative diversification** means making ads that are genuinely different in look, feel, storyline and message, aimed at different kinds of buyers or different reasons to buy. Meta’s system can then match each ad to the people it fits.

**Creative iteration** means small tweaks to the same ad: same photo, new headline; same video, different button text. Useful for learning. Not the same as diversification.

**Why Andromeda cares:** Meta says its system “groups together ads that share key visual and thematic attributes.” Lookalikes get treated as variations of one creative. Ten near-identical ads give Andromeda roughly one option, dressed ten ways.

> **The DumbCMO:** If your “creative refresh” is the same selfie with a new emoji, Meta already noticed. Andromeda did too.

## What Meta actually says (with sources)

Keep two piles. **Meta said it**, and **our read**. We won’t invent numbers.

### Meta said it

1. **Iteration ≠ diversification.** Same visuals with different button text = iteration. A Reels ad for Gen Z and a creator ad for new moms = diversification. ([Demystifying Creative Diversification](https://www.facebook.com/business/news/demystifying-creative-diversification), 16 Dec 2025.)
2. **Similar ads get grouped.** “Our ads system also analyzes creatives and groups together ads that share key visual and thematic attributes. When multiple ads look or feel alike, these are seen as variations of the same creative,” and learnings are shared “at the creative level – not just the individual ad level.” (Same post.)
3. **Andromeda is the shortlist.** When someone opens Instagram or Facebook, Andromeda “quickly scans tens of millions of ads and selects a few thousand strong candidates so only the most relevant and distinct creatives move forward.” Then models like GEM rank them. (Same post. Engineering detail in our [Andromeda 2026 note](/blog/meta-andromeda-2026-how-it-works).)
4. **Targeting gave ground to creative.** “The focus has shifted from niche targeting to creative diversification as the best lever to find the most relevant audiences.” ([The Creative Advantage](https://www.facebook.com/business/news/the-creative-advantage-unlocking-the-power-of-diversification-with-meta-andromeda), 22 Apr 2025.)
5. **Too-similar ads can hit the same people.** “When creative is too similar, our system may not recognize it as different and could deliver it to the same sets of audiences.” That overlap can lead to Creative limited or Creative fatigue in Ads Manager. ([Creative differentiation](https://www.facebook.com/business/m/small-business/creative-differentiation), Meta for small business.)
6. **Fatigue ≠ similarity.** Fatigue = people seeing the same creative too often. Similarity = ads that look too alike even when they’re separate files. Similarity can cause fatigue. (Dec 2025 post.)
7. **Numbers Meta publishes** (Meta’s averages, measured by Meta):
   - Advantage+ image generation: **11% higher CTR**, **7.6% higher conversion rate**. Text generation: **3% higher CTR**. (Dec 2025 / Apr 2025.)
   - Advertisers new to Advantage+ sales who turned on AI targeting: on average **22% higher ROAS**, and Meta cites **$4.52** return per $1 spent. (Dec 2025.)
   - Streamlined Advantage+ sales setup: on average **9% lower cost per action**. (Apr 2025.)
   - Older holiday guidance for Advantage+ shopping: keep **at least 20 diversified ads**, put **20–30%** of budget into testing new ads, and leaders launch weekly. ([Three steps](https://www.facebook.com/business/news/three-steps-to-optimize-your-performance-with-creative-diversification), holiday 2024.)
8. **One brand example, not a rule.** Dribbleup’s Head of Ecommerce said they went from 3–4 new creatives a week to almost 50. Meta quotes them. That’s one ecommerce brand with an in-house team, not a quota for a café in Pune. (Apr 2025.)

### Our read (inference, not Meta’s words)

- Andromeda’s job is to find *distinct* candidates worth ranking. If five of your ads look the same, Meta’s grouping language suggests they compete as one idea. You’re paying for five files and getting one bet.
- “Entity ID” is a phrase some agencies and blogs use for Meta’s internal fingerprint of a creative. We could not find Meta documenting a public “Entity ID” field or score for advertisers. Treat it as practitioner shorthand, not a button in Ads Manager.
- Account Insights can show a **creative similarity** insight when images or videos “appear too visually identical.” Meta’s Help Center notes the feature may not be available to every account yet, and it doesn’t publish a numeric similarity score or threshold. If you see it, treat it as a nudge to make something new. If you don’t, don’t invent a score.

**Meta never published “change X% of the pixels and you’re safe.” The bar is “truly distinct in visuals, messaging and formats.”**

## What counts as different (and what doesn’t)

Use this as a gut check. Meta’s examples and small-business checklist talk about different themes, messages, visuals, formats, storylines and reasons to care (price, performance, ease of use). The table below is our reading of that, not a Meta scoring rubric.

**Looks like variety vs actually different**

| Looks like variety (iteration) | Actually different (diversification) |
| --- | --- |
| Same product photo, five headlines | Customer video of them using it |
| Same Reel, new caption or music | Founder talking to camera about why they made it |
| Same ad, red button vs green button | Clear price / offer ad with different framing |
| Same model, three crops or colour grades | Before/after or demo of how it works |
| Same UGC clip, three on-screen text hooks | Gift-buyer ad vs self-buyer ad (different person, different reason) |
| AI background swap on the exact same product shot | Lifestyle scene in a real kitchen / shop / street vs studio packshot |

## The chai stall version

Think of a thali. Ten plates of the same dal with different spoons is still one dish. Dal, roti, raita, pickle and a sweet is a thali. Andromeda wants a thali. Agencies keep selling you ten spoons.

A cheap way to build real difference without a studio:

1. **Write 4–6 reasons people buy** (or kinds of people who buy). One ad per reason.
2. **Mix formats:** at least one talking-head or demo video, one clear product image, one social-proof ad (review, testimonial, UGC) if you have it.
3. **Change the first two seconds** of every video. Same opening = same ad to a tired thumb.
4. **Phone + daylight + a real customer** beats a polished clone of last month’s winner.
5. **Use Meta’s Advantage+ creative / GenAI tools as extras**, not as your only “diversity.” Meta’s own averages above are for those tools; you’re still responsible for what they write or generate ([we covered that](/blog/meta-ai-ad-edits-your-responsibility)).

## How this ties to Andromeda

Short version, so you don’t have to re-read the engineering post:

- **Andromeda = retrieval** (the shortlist). Ranking models like GEM only see what makes the shortlist.
- Meta rebuilt Andromeda so it can score a much bigger pile of eligible ads, in part because generative tools helped advertisers create “hundreds of millions of new ads.” More volume only helps if the ads are actually different.
- Meta’s business posts explicitly connect Andromeda to creative diversification: more distinct creatives → more options the system can personalise with.

So the practical link is boring and useful: **Andromeda can look at more of your ads. It will not magically invent difference you didn’t upload.**

For the funnel, GPUs and the 2.4% conversion claim, see [Meta Andromeda updates 2026](/blog/meta-andromeda-2026-how-it-works). For October product changes (formats, defaults, reporting), see [Meta ads October 2026 updates](/blog/meta-ads-october-2026-updates).

## How many ads should a small business run?

**What Meta says:** at least **20 diversified ads** inside Advantage+ shopping / sales campaigns, with **20–30%** of budget for testing, and weekly launches for “leaders.” One quoted brand hit ~50 new creatives a week.

**Our read for a small budget:** if you spend a few hundred rupees (or dollars) a day, 20 ads is a destination, not day one. Start with **4 to 6 genuinely different concepts** in one main campaign. Add one or two new different ads a week (or every two weeks). Kill or pause the ones that are clearly the same idea in a new shirt. Leave winners alone and put new concepts *beside* them.

> **The DumbCMO:** “We launched 40 ads” is not a flex if 38 of them are the same jar of pickle from three angles.

### How to structure it (practical, not a secret framework)

- Prefer **fewer campaigns, more different ads inside them**, so budget and learning aren’t split eight ways. Meta pushes Advantage+ sales for this reason; test it if you sell online.
- Keep hard limits that are real (city you deliver to, age if required). Drop interest stacks you can’t explain. Creative does more of the targeting now — Meta said the shift out loud.
- Don’t A/B-test button colours instead of concepts. Iteration can inform the next concept. It shouldn’t replace it. Meta said that too.
- Watch Ads Manager’s Delivery column for Creative limited / Creative fatigue, and Account Insights for creative similarity if it shows up. Those are Meta’s own flags.

**Your creative, after Andromeda**

- **If** You have 8–10 ads that are the same photo with new text **→ then** Keep the best one. Replace the rest with 3–5 different concepts.
- **If** Every video starts with the same face saying the same line **→ then** Reshoot openings. Change the first two seconds and the reason to care.
- **If** You’re on a tiny budget and feeling guilty about not hitting 20 ads **→ then** Ignore the guilt. Hit diversity first. Grow the count when you can keep them different.
- **If** One ad is working at a cost you can live with **→ then** Leave it. Add new concepts beside it. Don’t keep editing the winner.
- **If** An agency wants a weekly retainer to “feed Andromeda” **→ then** Ask how many of those ads would pass the thali test. Pay for concepts, not volume.

## Things your agency won’t say

- “50 ads a week” is one brand’s story Meta quoted, not a Meta quota for you.
- GenAI backgrounds on the same product shot are iteration with a GPU. Nice to have. Not diversification by themselves.
- A phone video of a happy customer is often more “diverse” to Meta than ten Canva reskins.

## Do this today

- [ ] Open Ads Manager. List live ads. Circle the ones that share the same visual or opening. That’s your real creative count.
- [ ] Write 4–6 different reasons people buy your thing (or kinds of people who buy). One ad idea per line.
- [ ] Make or brief one new ad that would fail the “same photo, new text” test. Ship it this week.
- [ ] Mix formats: video + static + proof (review/UGC) if you can.
- [ ] If you use Advantage+ creative / image gen, treat outputs as extras beside real concepts, and check the copy before it runs.
- [ ] Check Delivery for Creative limited / fatigue, and Account Insights for creative similarity if available.

## Not-so-frequently asked questions

### Does Meta treat near-identical ads as one?

Meta says it groups ads that share key visual and thematic attributes and treats lookalikes as variations of the same creative, with learnings shared at the creative level. It has not published a public “similarity score” or a pixel threshold.

### What about “Entity ID”?

That’s agency / blog shorthand for Meta’s internal creative fingerprint. We couldn’t find Meta documenting a public Entity ID advertisers can see. Don’t buy a course about hacking it.

### How many different concepts do I need?

Meta’s published best practice for Advantage+ shopping is at least 20 diversified ads. For a small budget, our view is start with 4–6 truly different ones and grow from there.

### Is changing the hook enough?

Sometimes for learning. Meta’s own iteration-vs-diversification examples say same visual + new CTA is iteration. Change look, feel, storyline and message if you want diversification.

### Will Advantage+ creative fix this for me?

It can help volume and some performance (Meta’s 11% / 7.6% averages). It won’t turn one idea into five. You still need distinct concepts.

### Is this only for ecommerce?

No. Meta’s small-business diversification page covers retail, real estate, finance and education too. Different reasons to care still apply if you sell a service.

## Where to go next

- How the engine works: [Meta Andromeda updates 2026](/blog/meta-andromeda-2026-how-it-works)
- What changed in product this month: [Meta ads October 2026 updates](/blog/meta-ads-october-2026-updates)
- If you use Meta’s AI edit tools: [Meta’s AI can now edit your ads](/blog/meta-ai-ad-edits-your-responsibility)
- Everything on ads: [PPC, dumbed down](/ppc)

## Sources

1. Meta for Business, [“Demystifying Creative Diversification”](https://www.facebook.com/business/news/demystifying-creative-diversification), 16 December 2025: iteration vs diversification examples; grouping of similar ads and learnings at creative level; Andromeda shortlist language; fatigue vs similarity; GenAI CTR/CVR averages; Advantage+ ROAS average.
2. Meta for Business, [“The Creative Advantage: Unlocking the Power of Diversification with Meta Andromeda”](https://www.facebook.com/business/news/the-creative-advantage-unlocking-the-power-of-diversification-with-meta-andromeda), 22 April 2025: shift from niche targeting to creative diversification; 9% lower CPA with streamlined Advantage+ sales; Dribbleup 3–4 → ~50 creatives/week quote; image gen 11% CTR / 7.6% CVR; Ben & Jerry’s +7% link clicks example.
3. Meta for Business, [“Three steps to optimize your performance with creative diversification”](https://www.facebook.com/business/news/three-steps-to-optimize-your-performance-with-creative-diversification), holiday 2024: ≥20 diversified ads in Advantage+ shopping; 20–30% budget for testing; weekly launches.
4. Meta for Business (small business), [“Successful ads tell a story in more ways than one” / creative differentiation](https://www.facebook.com/business/m/small-business/creative-differentiation): too-similar creative may not be recognised as different; same audiences / Creative limited or fatigue; diversify angles like price, performance, ease of use.
5. Meta Business Help Center, [Account Insights](https://www.facebook.com/business/help/1784925068944145): creative similarity when images/videos “appear too visually identical” (feature may not be available to all accounts).
6. Engineering at Meta, [Andromeda 2026](https://engineering.fb.com/2026/10/28/ml-applications/meta-andromeda-next-gen-ads-retrieval-gpu-co-design/), 28 October 2026, and [Andromeda 2024](https://engineering.fb.com/2024/12/02/production-engineering/meta-andromeda-advantage-automation-next-gen-personalized-ads-retrieval-engine/), 2 December 2024: retrieval vs ranking; capacity for more ads. Plain-words walkthrough in our Andromeda note.
7. Practitioner / press (labelled as such, not Meta fact): Marketing Brew, [how advertisers are adjusting to Meta’s creative diversification guidance](https://www.marketingbrew.com/stories/meta-ad-creative-diversification-best-practices-advertiser-approach) (agency quotes; Meta spokesperson says guidance is outcome-based, not model-hacking); various blogs claiming a public “Entity ID” or numeric similarity score — we could not verify those in Meta’s own docs.

*DumbCMO is a free explainer site. We’re not affiliated with, paid by, or endorsed by Meta. We don’t run ads for anyone. Numbers above are Meta’s own claims unless we say “our read.”*

## Related notes

- [Meta ads can show discounts you never approved](https://dumbcmo.com/blog/meta-ads-detected-offers-discounts): An ads specialist spotted a Commerce Manager setting, Detected offers, that lets Meta show discounts from your website in your Facebook and Instagram ads, even in copies of ads with promotions off. Meta hasn’t announced it. If you sell products on Meta, check it before copying Black Friday ads.
- [See the fake Google Ads clicks you didn’t pay for](https://dumbcmo.com/blog/google-ads-invalid-traffic-card): Google Ads is rolling out an Invalid traffic card on the Overview page. It shows estimated savings, the fake clicks filtered before billing, and credited cost, refunds for fake clicks found later. You’re not charged for invalid clicks. Check it before buying a click-fraud tool.
- [Google Ads learning phase is now 7 days](https://dumbcmo.com/blog/google-ads-learning-phase-7-days): Google’s help pages now say the Learning status lasts 7 days at most and shows only for a new bid strategy, added or removed campaigns, ad groups or keywords, or a big goal change. Budget, target, ad and targeting edits don’t trigger it. Results still take 1 to 2 conversion cycles to settle.

---

From [DumbCMO](https://dumbcmo.com): The marketing notes you were too embarrassed to take.
