Performance Creative Testing Framework for Meta Ads in the Andromeda Era

You have been running Meta Ads for months. You do your audience research, build tightly defined interest groups, set up separate ad sets for cold and warm audiences, and spend hours fine-tuning demographics.

And lately, it has stopped working the way it used to.

Your costs have gone up. Results have become unpredictable. Campaigns that worked last year are struggling this year for no obvious reason.

There is a reason. And it is called Andromeda. Meta quietly replaced its entire ad delivery engine between late 2024 and early 2025. The old rules no longer apply. And if you are still playing by them, you are paying for it.

This article explains what Andromeda changed, why creative testing is now the most important skill in Meta advertising, and how to build a simple testing framework that works inside this new system.

What Is Andromeda and Why Should You Care?

Andromeda is Meta’s new AI-powered ad delivery engine. It completed its global rollout in October 2025 and is now fully live across Facebook and Instagram for all advertisers, including campaigns running in India.

To understand what changed, you need to know how ad delivery used to work. When you ran an ad, Meta’s system would take your targeting inputs, such as interests, demographics, and lookalike audiences, and use those as the primary filter to decide who sees your ad. You told Meta who to reach. Meta went and found them.

Andromeda flipped that logic entirely. Now, Meta does not start with your audience. It starts with your creative. The system scans your ad, reads the visuals, the hook, the language, and the tone, and then goes out to find the people most likely to respond to that specific content.

Think of it like this. Old Meta was like a postman delivering letters to addresses you gave it. Andromeda is like a smart recommendation engine, more like Netflix, that figures out who wants what based on what the content actually is, not just who you thought would like it.

The system uses machine learning models that are 10,000 times more complex than before. It evaluates millions of ad creatives in milliseconds and matches each one to users based on real-time behaviour, not static demographic profiles.

Key fact: Meta’s own internal tests show Andromeda delivers 8% to 10% better campaign performance, but only when campaigns are structured and fed correctly. If you are still using old targeting tactics, the same system that helps others hurts you.

What This Means for Your Targeting Strategy

Here is the uncomfortable truth for anyone who has spent years mastering Meta’s audience targeting tools.

Interest targeting, lookalike audiences, and narrow demographic segments have become significantly less effective. Not broken entirely, but no longer the lever they once were.

This is why campaigns structured around ‘Mumbai women aged 25-34 interested in fitness and organic food’ are now being outperformed by broad national campaigns with no interest filters at all. Andromeda does not need your help finding the right person. It needs your creative to tell it what the right person looks like.

Meta itself has made this explicit in its own guidance: the focus has shifted from audience targeting to creative diversification as the primary lever for finding relevant audiences.

What this means practically for Indian advertisers:

Stop over-segmenting. Running 5 ad sets with slightly different audiences fragments your data and gives the algorithm less to learn from. Consolidate into one campaign with broad targeting and let Andromeda do the matching.

Stop relying on lookalikes as hard boundaries. Lookalike audiences are still useful as a signal, a seed to help the algorithm start in the right direction. But the system will expand well beyond them. Restricting it to only that audience costs you reach.

Start trusting Advantage+. Advantage+ placements and Advantage+ audiences are no longer the lazy option. Under Andromeda, they are often the better-performing option because they give the algorithm the freedom it is designed to use.

Creative Is Now Your Targeting

This is the single most important sentence in this article, so read it twice.

Under Andromeda, your creative is not just the thing people see. It is the signal Meta uses to decide who to show your ad to. Your creative IS your targeting.

If your creative speaks to a working mother worried about her child’s education, Andromeda will find working mothers who are worried about their child’s education. You do not need to target them manually. The content does the job.

This is why creative testing has gone from being a ‘nice to have’ to being the most important thing you can do in your Meta advertising. The quality, variety, and distinctiveness of your creatives now determines both who sees your ad and whether they respond to it.

And here is what makes India’s market particularly interesting right now: with CPMs averaging around ₹210 and CPCs around ₹34, Indian advertisers can run significantly more creative tests than advertisers in Western markets for the same budget. This is a structural advantage. Use it.

The Creative Testing Framework Built for Andromeda

This is not the old ‘test one element at a time’ approach from 2022. Andromeda rewards genuine creative diversity. Here is how to test creatives in a way that works with the new system, not against it.

Step 1: Test Concepts, Not Tweaks

The biggest mistake advertisers make now is testing minor variations of the same ad. A different headline on the same image. A slightly different CTA button colour. A 5-second edit to an existing video.

Andromeda’s visual recognition models are sophisticated enough to recognise when two ads are essentially the same piece of content wearing a different hat. When it detects this, it may treat them as duplicates, which reduces your reach and increases your costs.

You need meaningfully different creative concepts. Each concept should represent a distinct angle, a different reason why someone might care about your product or service.

For example: if you sell an online CA coaching course in India, one concept could be a student testimonial video focused on fear of failure. A second could be a breakdown of the pass rate vs the national average. A third could be a day-in-the-life Reel of a student’s study routine. These are three different ideas, three different emotional triggers, three genuinely different pieces of content.

Aim for 8 to 15 genuinely different creative concepts per campaign. Not variations of one idea. Different ideas.

Step 2: Write a Hypothesis Before You Spend

Before launching any creative, write one sentence about what you expect to happen and why. This is called a hypothesis and it is what turns ad spend into learning rather than just cost.

It looks like this: ‘I believe a short Reel showing a real customer result will generate more leads than a studio graphic because our audience is sceptical of polished ads and responds better to proof.’

When the results come in, you compare against the hypothesis. Right or wrong, you learn something specific. Over time, these learnings become a playbook for your brand, a set of principles that make every future campaign smarter and cheaper.

Without a hypothesis, results are just numbers with no meaning attached.

Step 3: Set the Right Budget and Wait Long Enough

Most creative tests fail not because the creative was bad but because the test was cut too early. Andromeda needs time to exit its learning phase, the period during which it figures out who to show your ad to. Making changes or pausing ads during this window resets the learning and you start from zero.

Run each creative variation for a minimum of 7 days. For conversion campaigns, wait until each variation has generated at least 20 to 30 conversion events before drawing conclusions.

Step 4: Focus on the Hook First

On Meta, you have between 1 and 3 seconds to stop someone from scrolling past your ad. That window is controlled entirely by your hook, the opening frame of a video or the first line of a static ad.

Most brands spend the majority of their creative effort on the body of the ad and barely think about the hook. This is backwards. A weak hook on a great ad means no one ever gets to the great part.

Test your hooks aggressively and separately from everything else. A different opening line on the same underlying ad can double your click-through rate. Under Andromeda, a higher CTR signals stronger creative relevance, which means lower costs and better delivery.

Good hooks for Indian audiences tend to do one of three things: name a very specific problem the viewer recognises immediately, open with a surprising or counter-intuitive claim, or show a real result that the viewer wants for themselves.

Step 5: Read Results and Extract One Learning Per Test

When a test ends, do not just note which creative won. Ask why it won, and write down what you will do differently in the next test because of that answer.

The three metrics that matter most, in order:

CTR (click-through rate): Is the creative stopping the scroll? If CTR is low, the problem is the hook or the visual format, not the offer.

Cost per result: Is the campaign profitable? This is the number that connects ad performance to business outcomes.

Landing page conversion rate: Are the clicks turning into enquiries or sales? If CTR is strong but conversions are weak, the landing page is the problem, not the creative.

Keep a simple document recording every test: the hypothesis, the creative description, the result, and the learning. After 6 to 8 tests, patterns emerge. Those patterns become your brand’s creative playbook.

Common Mistakes That Break the System

Uploading too many similar creatives

Andromeda penalises creative similarity. If you upload 10 ads that are visually similar, the algorithm may group them together, reduce their combined reach, and push up your CPMs. More ads only helps if they are genuinely different from each other.

Making changes during the learning phase

Every time you edit a campaign, pause and restart it, or significantly change the budget, Andromeda resets its learning. Resist the urge to react to early numbers. Give campaigns the 7-day minimum before touching anything.

Scaling a winning ad too fast

When a creative wins, the instinct is to immediately triple the budget. This disrupts the delivery algorithm and often inflates costs fast. Scale in 20% increments every few days and monitor cost per result closely as you go.

Ignoring creative fatigue signals

When the same audience sees your ad too many times, performance drops. On Meta, track your frequency score. Once it crosses 3 to 4 on a cold audience, your creative needs refreshing. In India, peak seasons like Diwali and Holi push CPMs up significantly. Test and identify your winning creatives before these periods, then scale proven ads when competition and costs are high.

The New Game: Creative Is the Strategy

Andromeda did not make Meta Ads harder. It made them more honest.

The old system rewarded clever targeting, complex structures, and audience segmentation skills. It was possible to compensate for mediocre creative with good setup. That is no longer true.

The new system rewards one thing: creative that genuinely connects with people. Andromeda’s job is to find those people and put your ad in front of them. But it can only do that job if you give it something worth finding.

Businesses in India that understand this shift early have a real advantage. The cost of testing is low here. The audience is large and engaged. And most competitors are still running the same four ads they have had for six months.

Build a testing system. Generate real creative diversity. Write down what you learn. And let Andromeda do what it was built to do.

The brands that treat creative testing as a system, not a one-time task, are the ones that will consistently win on Meta in 2026 and beyond.

Not sure if your Meta Ads are structured for Andromeda? Request a creative performance audit from PS Digital.