Meta Andromeda for Shopify: How Many Ad Concepts?

Quick answer: Meta has not published a number of ad concepts to run. What it has published is that Andromeda, its ad retrieval stage, narrows tens of millions of ads to a few thousand for each person, and that supplying more creative variety gives the system more to test. For a small Shopify team, a sensible starting point (our planning rule, not a Meta rule) is two to six concepts, scaled to how many purchases a week Meta reports, that each give a different reason to buy, made in two formats and run inside one consolidated campaign so every ad still collects enough results to be judged. Count concepts, not files: a new headline on the same idea is the same concept.

If you run Meta ads for a Shopify store, someone has probably told you that Andromeda changed everything and that you now need 15, or 20, or 50 new creatives a month. Some of that is grounded in what Meta has said. A lot of it is agency folklore with a confident number attached.

First, what I do and do not do. Prism3 sells Shopify SEO, GEO and CRO. We do not run ads or email. I am writing this because Meta creative is one of the levers on cost per customer, and it is worth understanding even if someone else makes it. The brief you hand your designer or your agency shapes who clicks, and the page they land on shapes how many of them buy. The second half of that is the part of the system we work on.

This guide separates what Meta has published about Andromeda from what has been added since, then turns it into a creative brief a small team can run: how many concepts, what makes two concepts different, which formats to vary, and how to read the results without fooling yourself. Every Meta fact below comes from a page I opened myself, and I have noted which. Meta’s pages change, so treat the dated statements as true as at 31 October 2026 and check them again before you lean on them. Meta published an engineering post and two business pages about Andromeda on 28 October 2026, so many explainers you will find predate them.

What Meta has actually published about Andromeda

Andromeda is not a creative setting you switch on. It is a part of how Meta’s ad system works. The pages that describe it, in date order:

Andromeda is a shortlist step, not the whole decision. Meta describes retrieval as narrowing the pool before larger ranking models choose the final ads, and its auction then weighs your bid, ad quality and the estimated chance the person takes your action. A creative that survives retrieval still has to win the auction.

The percentages are Meta’s own, system-wide figures. The 8% (ads quality) and 2.4% (conversions on Facebook) describe Meta’s platform, with methods the pages do not explain. They are not a prediction for your store.

Meta’s advice is directional. The most specific line I could verify is the October 2026 tip to supply more creative variety so the system has more to test and learn from. It gives no count.

If you want Meta’s own summary of what to do, it is the Performance marketing page, which lists five practices: account simplification, automation, creative diversification, data quality and results validation. Creative diversification is one of five. It sits next to combining ad sets and connecting your data through the Conversions API, which the folklore tends to leave out.

Folklore versus the record

Here is what you will hear, and what I could and could not verify. “Not verified” means I did not find it on any Meta page I opened, not that it is false.

Why it matters: the Entity ID story sends you to make visually different versions of the same promise, hoping to get past a filter. What Meta’s pages support sends you to find different reasons for different people to buy. That is better work whether or not any filter exists.

What “distinct” should mean

The Meta pages I read do not define a concept, so here is a working definition that holds up either way.

A concept is one reason a particular buyer would choose your product. The reason is the idea. Everything else is execution: the format, the hook, the headline, the model, the background colour.

Here is a test I use. Take two ads. Swap their headlines and visuals. If the ads still make the same promise to the same person, they are one concept. If swapping changes who the ad is for or what it promises, they are two.

An illustrative example, not a client: a store sells a burr coffee grinder.

• “Stop wasting beans on bad grinds” and “Grind fresher than a café” are both about grind quality for a coffee fan. One concept, two headlines.
• “A gift for the person who has everything” is a different buyer and a different reason. A second concept.
• “Quiet enough for a unit with thin walls” answers an objection from a different person. A third concept.
• “Still using the pre-ground bag? Here is the difference in the cup” is education aimed at a person who is not yet convinced grinding matters. A fourth.

Same product, four reasons, four likely buyers. Four headline variants over one photo is iteration: useful, but a different job.

Eight angles to mine for raw material

You do not need to invent angles. Your store already holds the evidence of why people buy and why they hesitate.

Session recordings and review text are cheap research. Our guide to reading Shopify session recordings shows how to find the hesitations that make good objection angles. Three real objections beat twelve invented ones.

How many concepts should a small store run?

Meta does not give a number, so the honest answer is that it depends on how much data your budget can produce. The reason is mechanical, and Meta documents it.

Meta’s learning phase page says an ad set usually leaves the learning phase after about 50 results in the week after its last significant edit, and that during learning performance is less stable and cost per action is usually higher. Its ad volume page says that when too many ads run at once, each one delivers less often, fewer ads exit the learning phase and more budget is spent before the system can optimise. Its guidance on splitting says to combine ad sets, because every extra ad set spreads your results thinner.

Here is worked arithmetic. These are illustrative numbers, not a benchmark and not a client.

• A store spends A$100 a day on Meta, which is A$700 a week.
• Meta reports a cost per purchase of A$35, so the store gets about 20 purchases a week.
• To reach about 50 results in a week, it would need roughly 50 × A$35 = A$1,750 a week, or A$250 a day.
• At A$100 a day, a single ad set sits below that line. Split the budget across four audience ad sets and each gets roughly five purchases a week.

That is the real constraint on a small store. The risk is spreading a small number of purchases across too many ad sets and ads, so none collects enough results to teach the system or you.

My planning rule, which is judgement and not a Meta threshold:

If your purchase numbers are smaller than the table assumes, the answer is still not “more concepts”. It is fewer ad sets, stronger differences between the concepts you do run, and more patience.

There is a counterweight. Meta’s Advantage+ shopping campaigns page (Meta says it has renamed these Advantage+ sales campaigns) recommends importing your existing eligible ads so the system can automatically test up to 150 creative combinations. As at 31 October 2026 that page still uses the older name, so check the current limit in Ads Manager. The platform can handle variety; your budget decides how much of it you can learn from. Our guide to creative testing on small Shopify budgets goes further on structuring tests, and the post on when Advantage+ sales campaigns make sense covers the campaign choice.

Fewer ads, more varied assets

Meta’s ad volume page makes a specific recommendation that fits small teams well: decrease the number of ads per ad set, but keep diverse creative assets in each, noting that one ad can hold up to ten assets. As at 31 October 2026, that page still refers to changes from June 2024 and to dynamic creative and flexible ad format, so check your own Ads Manager for the current options before you build around it.

In practice, for a three-concept test you might build three ads, one per concept, each carrying a static image and a short video, rather than six separate ads. Bundled assets can limit how cleanly you can read each one, so check what breakdown Ads Manager gives you for your setup.

What to vary inside a concept: formats

Once you have distinct concepts, vary the format, not the idea. Here is what Meta says that you can act on.

• Mix static and video. Meta says its direct response research found that campaigns with both static images and video achieved conversion lift at a 17% higher rate than campaigns with static images only. The page gives no sample, dates or method, so treat it as a reason to test a video, not as a benchmark.
• Keep videos short and single-minded. For conversion testing, Meta recommends 5 to 10 seconds, one benefit, animation in the first three seconds, and a call to action with a consistent end card.
• Build for the phone. Meta recommends 1:1 or 4:5 so the ad covers more of a mobile screen, and its ad creative page recommends choosing six or more placements and a mixed-format strategy.

A simple grid keeps this from sprawling. Down the side, your concepts. Across the top, two formats. Fill each cell once.

You do not need every cell on day one. Start with cells you can produce from material you already own: customer photos, a phone video, a review.

Check what Meta’s automation changes

Meta’s Advantage+ creative page says some enhancements may be on by default, that you can turn them off, and that ads created or significantly edited with certain generative AI features may carry an “AI info” label. Open the enhancements list before you publish and choose on purpose: visual touch-ups, a music track or a generated background can change how a premium product looks.

The one-page creative brief

This is the part you can take to a freelancer, a designer or your own phone. One page per concept. If it does not fit on one page, the concept is not clear enough.

CONCEPT NAME:            (e.g. "Gift for the person who has everything")
WHO IT IS FOR:           one sentence, in the customer's words
THE REASON TO BUY:       one promise, not three
THE OBJECTION IT KILLS:  what stops this person buying today
PROOF WE CAN SHOW:       review, photo, spec or demo we actually have
FORMATS:                 static + one short video (5 to 10 seconds)
FIRST 3 SECONDS:         what the viewer sees and reads first
CALL TO ACTION:          the one next step
LANDING PAGE:            the exact URL, and the first thing it says
WHAT WOULD STOP US:      the result that makes us retire it

The last two lines are where most briefs fall down: the landing page line makes you plan the click’s destination before you spend, and the stopping line stops you chasing a concept because you like it.

Make the page continue the promise

Meta’s October 2026 explainer says ad quality reflects the whole experience, from the ad itself to what happens after the click. It names shipping speed, support responsiveness, product quality and checkout friction as things to audit the way you review creative, and says Meta draws on more than 200 million customer survey responses a year covering things like purchase satisfaction, shipping delays and unexpected charges.

That is Meta telling you the post-click experience is part of how its system judges your ad. It also means each concept needs a destination that keeps its promise.

Do not send every concept to the home page: a person who clicked “quiet enough for thin walls” and lands on a generic hero has to start again. Our product page CRO checklist lists what to look at, and mobile conversion on Shopify covers the phone-specific problems, which matter if most of your paid social traffic arrives on a phone (check your own split in Shopify).

Running it without breaking the test

Set up once, then leave it. Meta’s learning phase page recommends waiting to edit until the ad set is out of the learning phase, because editing during it resets learning. It lists avoiding unnecessary edits, avoiding high ad volumes and using realistic budgets. Decide your concepts, launch them together and resist the daily tinkering.

Combine ad sets. Meta’s page on audience fragmentation says splitting an audience across several ad sets with the same objective and creative could decrease performance, and suggests using reporting breakdowns for segment-level views rather than separate ad sets. If you have several ad sets with the same creative and slightly different interests, that is the first thing to consolidate.

Name and tag every concept. Put the concept name in the ad name and in a utm_content value so you can see orders by concept in Shopify and Google Analytics. Platform-reported numbers and Shopify’s will differ, which is normal. Our post on why Meta shows more sales than Shopify explains the main reasons. Judge concepts on cost per purchase, and compare against your own order data.

Send clean conversion data. Meta’s Performance marketing page lists data quality as one of its five practices and names the Conversions API. If your purchase events are doubled, missing or late, your concept results are wrong before you read them. See our Meta Conversions API setup guide.

Read concepts as a set, not a league table. When the whole campaign makes 20 purchases a week, an ad with six purchases is not proven and an ad with two is not dead. Until the ad set is out of learning, treat per-ad numbers as directional, and look for patterns: did a concept get clicks and no orders?

Know what fatigue looks like, and what Meta’s tool can and cannot see. Meta’s creative fatigue page says “Creative limited” appears when cost per result is above what your past ads cost but under twice as much, and “Creative fatigue” when it is at least twice as much. It also says the feature is only available for ad sets with one creative, and not for Advantage+ catalogue ads, dynamic creative or Advantage+ app campaigns. So if you run several creatives in one ad set, you may not see those labels, and you will need to watch cost per purchase yourself. Its recommended fix is a new ad with an image or video that is materially different, and it notes that keeping the original active instead of pausing it may maximise results.

Rotate by concept, not by file. When a concept tires, the replacement is a new concept from your angle list, not a recolour of the old ad. That is the clearest working meaning of diversification I can defend from Meta’s pages.

When more concepts will not help

Creative is a cost-per-click and click-through problem. It is not a store problem. If a campaign is bringing visitors who do not buy, adding concepts spends more to learn the same lesson.

An illustrative sum, not a benchmark: say your clicks cost A$1.00 each.

• At a 1% store conversion rate, 100 clicks cost A$100 and produce one order: A$100 per order, before product cost.
• At 2%, the same 100 clicks produce two orders: A$50 per order.

Same ads. Same creative. Half the cost per customer. Meta’s own explainer says the platform optimises toward the outcome you ask it to, and that a better customer experience can help an ad beat a higher bid. If you have good concepts and clicks that do not buy, read why Facebook ad clicks do not turn into sales, and if your cost per customer is climbing and you are not sure why, the diagnostic order for rising Meta CAC is the place to start.

A first-month plan

Week 1: pull the top support questions and review phrases, and watch ten session recordings. Week 2: write three one-page briefs and run the swap test on each. Week 3: make one static and one short video per concept from real material, and check each landing page on a phone. Week 4: launch in one consolidated setup, tag every ad, confirm purchases fire once, then leave it alone. After that, review weekly as a set.

Key takeaways

• Meta has not published a number of ad concepts. Treat any specific number, including mine, as a planning rule.
• What Meta has published: Andromeda is the retrieval stage, it considers more of your ads and creative variations, and supplying more creative variety gives the system more to learn from. Its 8% and 2.4% figures are system-wide and self-reported.
• “Entity ID” and “similarity score thresholds” are third-party terms. I found neither on a Meta page I opened. What Meta does document is auction overlap between your own ads that share an audience or assets.
• A concept is a reason to buy for a particular person. Headlines, colours and crops are execution. Use the swap test.
• On a small budget the risk is spreading purchases too thin. Meta says an ad set usually needs about 50 results in the week after its last significant edit to leave learning.
• Fewer ads per ad set with varied assets is Meta’s own advice. Mix static and short video, and build for the phone.
• Each concept needs a landing page that continues its promise. Meta says ad quality includes what happens after the click.
• If clicks do not buy, more concepts will not fix it.

FAQs

How many ad creatives should a small Shopify store run on Meta?

Meta gives no number. As my own planning rule (not a Meta rule), two to six distinct concepts, scaled by how many purchases a week Meta reports: two or three under about 20 purchases, three to five from 20 to 50, four to six above 50. Each concept should give a different reason to buy. How many matters less than whether each one teaches you something.

Does Andromeda penalise similar ads?

I could not find a Meta page that says so, and “Entity ID” and similarity-score thresholds appear only in third-party write-ups. Meta’s October 2026 post says the system scores every eligible ad. What Meta does document is auction overlap: when ads from one advertiser enter the same auction, only the one with the highest total value is considered, and overlap can limit delivery when ads share an audience or assets. Making your ads genuinely different avoids the issue anyway.

What counts as a distinct creative concept?

A different reason to buy for a particular person, not a different file. Swap the headlines and visuals of two ads: if they still make the same promise to the same person, they are one concept. A new background, crop, caption or voice-over on the same idea is iteration.

How often should I refresh my Meta creative?

Meta gives no schedule for sales campaigns on the pages I read. What it publishes is a signal: “Creative limited” and “Creative fatigue” labels based on cost per result against your past ads, available only for ad sets with one creative. Watch cost per purchase and frequency of your best ads, and rotate in a new concept when results sag, not on a calendar.

Do I need video for Meta ads?

You do not need it, but test it. Meta says its direct response research found campaigns with both static images and video achieved conversion lift at a 17% higher rate than static-only campaigns, without giving the method. A short, phone-shot video of the product in use is a fair first test.

Should I still split ad sets by audience?

Meta’s pages advise the opposite: combine ad sets so results are not spread thin, and use breakdowns for segment reporting. In my view, if your budget produces under 50 purchases a week, one ad set with varied creative is a safer starting point than several narrow ones.

Keep reading

• Meta creative testing on small Shopify budgets
• Why Meta CAC is rising: the diagnostic order for Shopify stores
• Facebook ad clicks but no sales on Shopify

Want a second pair of eyes on what happens after the click?

More creative concepts give Meta more to test; none of them changes what happens when the visitor reaches your store. Book a free 30-minute diagnostic: we look at where your store loses buyers after the click, using your own numbers, and tell you what to fix first. We do not run ads or email. No pitch. No email gate.

If your ads are bringing visitors who do not buy, see how our Shopify CRO service approaches conversion and tracking.