Picture a media buyer opening Ads Manager on a Monday morning in early November. A campaign that worked well in the summer has slowed down. The audience hasn’t changed, the budget hasn’t changed and the targeting is exactly as carefully built as before. Yet results are drifting, and the usual fixes, a tighter audience or a new lookalike, don’t seem to move anything. Black Friday is under four weeks away.
That frustration is common right now, and it often traces back to an assumption that used to be safe. Think back to how a Meta campaign used to be built. You chose an audience, layered on interests, added a lookalike, set a few exclusions and then wrote the ads to suit. The audience came first and the creative followed. The skill lay in telling the platform who to reach.
Today, many advertisers say the order has flipped. They upload a range of genuinely different ads and let Meta decide who should see which one. The setting that used to carry the weight, the targeting, now does less of the work, and the ad itself does more. That’s a bigger change than it sounds. It means the creative team, not just the media buyer, is now steering who your campaigns reach, and it means a stale or repetitive set of ads can quietly cap your results however well the audience is built.
The name usually attached to that change is Andromeda. It’s also one of the most talked-about and least precisely explained topics in paid social, with plenty of confident advice online and much less that can be traced back to Meta itself. In this post we’ll explain what it is, what Meta has actually said about it, where agency advice goes beyond Meta’s own guidance, and what to do about it with the holiday peak only weeks away. We’ll finish with a checklist and answers to common questions.
What Andromeda Is, in Plain English
Andromeda is Meta’s ads retrieval system. Before any bidding or ranking happens, Meta has to narrow a huge pool of eligible ads, tens of millions, down to a shortlist of a few thousand that are worth scoring for a particular person. Andromeda is the system that does that first cut.
Meta described it on its engineering blog on December 2, 2024, and has since said it rolled out across its apps. It isn’t a campaign type you choose in Ads Manager. It’s part of how the auction works, which is why advertisers experience it as a change in what matters rather than a new button to press.
Retrieval happens before the auction, so an ad that isn’t shortlisted never gets the chance to win a bid. That’s the reason the quality and variety of your creative matters so much more than it once did.
The Numbers Meta Has Published
Meta reports that Andromeda raised retrieval model capacity 10,000 times, improved recall by 6% and improved ads quality by 8% on selected segments.
Those figures come from Meta’s own announcement, and the ads quality gain applies to selected segments, not to every advertiser. They describe the system’s performance, not what any individual account will see.
More recently, Meta published a new engineering post on October 28, 2026. According to summaries of it, the current system can process more than a billion ads per second on a single GPU server, described by Meta as “a 100x lift over the 2024 system.” Reporting also says Meta deployed a first generative model into its ads retrieval system in July 2026. The details will keep changing, but the direction hasn’t: more capacity to consider more ads and match them more precisely.
Meta’s advice to advertisers, as relayed by a spokesperson, is to diversify your creative, test and let the system optimise.
Creative Diversification: What Meta Actually Recommends
Meta’s guidance to advertisers has been consistent since early 2025. In March, Meta for Business explained that as businesses upload more creatives to support a diversification strategy, Andromeda powers the models that pick the right creative for each person. In April, Meta said the focus had shifted from niche targeting to creative diversification, and in December it published a post titled “Demystifying creative diversification.”
Diversification here means genuinely different ads, not small edits to the same one. Meta points to differences in format, such as video versus static or partnership content, and in theme, such as humour, nostalgia or savings. A new caption on the same footage is iteration, not diversification. The thinking is that distinct ads reach distinct kinds of people, while near-duplicates compete for the same ones.
It also helps that volume has exploded. Meta’s 2024 post noted that more than a million advertisers had used its generative AI tools to create over 15 million ads in a single month, according to summaries of it. With that many ads in the system, a clear, distinct creative has a better chance of being surfaced.
What Meta Has Said, and What Agencies Report
This is the part worth being careful about. Meta has not published a rulebook for advertisers. Its posts recommend creative diversification but don’t say how many ads to run, how to structure ad sets or how to read results.
A lot of the specific advice you’ll see online goes further: use fewer ad sets, put many more ads in each one, rely on campaign-level budgets and stack creatives across the stages of buyer awareness. Many agencies report good results from these approaches, and the logic is plausible, since pooling data in one place may help the system learn. But those results come from individual accounts, they aren’t Meta’s guidance, and what works for one brand won’t necessarily work for another. Treat them as hypotheses to test, not rules.
The same goes for precise thresholds that circulate, such as exact creative counts or similarity scores. Where a number isn’t in Meta’s own material, it’s someone’s observation, and your account may behave differently.
A Practical Plan
Whatever structure you choose, a few steps make sense on the evidence available.
Get your tracking right first. Meta can only learn from the data it receives, so check that your purchase events are sent reliably, including server-side, and that customer details are matched correctly. A poor signal makes everything else harder to read.
Build a creative map. List the angles you could take: the problem, the solution, the proof, the offer. List the formats you could use. Then make sure your ads cover several of each, rather than ten versions of one idea.
Test structure changes deliberately. If you want to try consolidating ad sets, change one account or one campaign at a time and compare it with a control, so you know what moved the result.
Plan a steady supply of fresh creative. Creative tends to fatigue, so a pipeline matters more than a one-off batch. This is a production problem as much as a media one, and it’s where our brand and creative work supports the media team.
Match the landing page to the ad. A story-led ad needs a page that continues the story, and a product ad can go straight to the product.
Read results at the creative level. Look at which ads are carrying the delivery, and check the platform’s numbers against your own analytics. Our three-lane framework for AI and human control explains why independent verification matters when a platform decides where the money goes.
For shops, feed quality matters. If you run catalogue or Advantage+ shopping campaigns, accurate product data feeds the system’s decisions directly.
A Kilowott Example: Gutta på Haugen
On Gutta på Haugen, our Oslo specialty food client, our paid media work ran on Meta and Google together. On Meta that meant Advantage+ and catalogue sales, with creative rotation and feed optimisation tied to the shop’s own priorities, and campaigns scaled seasonally across Christmas, Easter and summer. The results page reports a 4.24x blended return on ad spend across Meta and Google.
As with the other figures from that project, it comes from Kilowott’s own reporting, not an audit of the client’s ad accounts, and no spend figure is published, so read it as direction, not a benchmark. What the example shows is the pattern, not the number: a store built to convert, a clean product feed and a steady rhythm of creative, run together. We unpack the approach in what 4.24x blended ROAS looks like when the store is built right.
Why This Matters Before Black Friday
With Black Friday on November 27, the timing is practical. Holiday periods are when creative fatigues fastest and competition is highest, so having a range of distinct ads ready before the peak is more useful than scrambling during it. Our post on holiday ad costs covers how to plan budgets and limits, and our digital marketing team can help build the creative plan alongside it. Reporting that compares ads, formats and audiences is a natural job for automation, so the picture is ready every Monday without anyone assembling it by hand.
A Meta Andromeda Readiness Checklist
- Purchase events sent reliably, including server-side, with customer details matched
- A creative map covering several angles and formats, not variations of one idea
- Genuinely different ads, not edits of the same footage
- A plan for fresh creative every few weeks, not a single batch
- Landing pages matched to the message of each ad
- Structure changes tested one at a time against a control
- Results read at the ad level and checked against your own analytics
- Product feed accurate and complete for any catalogue campaigns
- Holiday creative ready and approved before the peak
Frequently Asked Questions
What is the Meta Andromeda update? Andromeda is Meta’s ads retrieval system, the stage that narrows tens of millions of eligible ads to a few thousand candidates before ranking. Meta announced it on December 2, 2024, reporting a 6% recall improvement and an 8% ads quality improvement on selected segments.
Does Andromeda replace targeting? Not entirely. It changes where the work happens. Meta’s guidance says the focus has shifted from niche targeting to creative diversification, so the variety of your ads matters more than before.
What is creative diversification? It means running ads that are genuinely different in format, theme and message, not minor edits to a single ad. Distinct ads can reach distinct kinds of people.
How many ads should I run? Meta hasn’t published a number. Agency advice varies, so test what works for your account and focus on making each ad genuinely different.
Should I consolidate my ad sets? Many agencies report good results from fewer ad sets with more ads in each, but that’s practitioner advice, not Meta’s rule. Test it carefully on one campaign before changing everything.
Where to Start This Week
If you only have time for a few actions before the holiday peak, these are the ones that matter most.
- Audit your current ads. Lay out everything running now and ask honestly how many are genuinely different from each other. If most are edits of the same idea, that’s your first gap.
- Brief three new angles. Pick three different messages or formats you haven’t used, such as a problem-led ad, a proof-led ad and an offer-led ad, and get them made and approved before Black Friday.
- Check your tracking. Confirm purchase events are arriving reliably, including server-side, before you spend more.
- Set up one clean test. If you want to try a simpler structure, choose a single campaign and compare it with a control, so you can tell what actually helped.
None of this needs a new platform or a large budget. It needs a short conversation between the people who make the ads and the people who run them, which is often the real missing link.
Make Each Ad Earn Its Place
The most useful way to think about Andromeda is as a system that rewards having something distinct to say. It’s less about learning a new setting and more about building a steady supply of different, well-made ads and measuring them honestly. The advertisers who benefit won’t be those who found the cleverest structure. They’ll be the ones who treated creative as the main lever and kept testing.
It’s also worth keeping some perspective. Meta’s systems will keep changing, and the specific advice you read this month may look different by spring. Reporting says Meta deployed a first generative model into its ads retrieval in July 2026, and it has published nothing that promises a fixed formula for advertisers. What stays useful through each change is the discipline underneath: genuinely different ads, reliable data, careful testing and results you check against your own numbers, not just the platform’s. Teams that build those habits adapt quickly whenever the next update arrives.
If you’d like help planning your creative, structuring tests or reading results before the peak, our digital marketing team is happy to talk it through, and you can see how we approach this across other brands in our case studies.
