Your Audience Targeting Isn’t Doing What You Think It’s Doing

If you’re spending time refining interest stacks, tightening demographic exclusions, and building out custom audience segments while your creative library has three variations of the same ad, you’re solving the wrong problem. Meta’s ad delivery system changed in late 2024 and has been running at full scale ever since. Our own 2025 Facebook Ads cost benchmarks showed lead generation costs rising around 21% year on year, and a lot of that pressure comes from advertisers competing with thin creative libraries, not bad targeting.

The core change is this: Meta’s system no longer primarily asks “does this user match the advertiser’s targeting?” It now asks “does this creative match this user?” That’s a different question than what drove ad delivery for most of the platform’s history, and it changes where you should be focusing your time and budget.

How Meta Used to Decide Who Saw Your Ads

For most of Meta’s advertising history, the advertiser was the primary signal. You told the system who you wanted to reach: age ranges, locations, interests, behaviours, lookalike audiences built from your customer list. Meta served your ad within those parameters. The better you understood your audience, the better your targeting and the better your results.

The result was an entire discipline built around audience architecture: precise ad sets for each segment, exclusion testing, tiered retargeting sequences, demographic adjustments. Granular control was the goal.

Two things eroded that model. iOS 14.5 in 2021 stripped much of the third-party data that made interest and behaviour targeting accurate. Then Meta began pushing toward AI-driven automation: Advantage campaigns, broad targeting options, and eventually Advantage+, each of which reduced the role of manual audience definition. Andromeda is the most significant of those steps because it changed what the algorithm is optimising for at a structural level.

What Andromeda Changed

The retrieval stage: what most advertisers don’t know exists

Meta’s ad system doesn’t just run an auction. Before your ads compete for impressions, they have to be selected as candidates, and that selection process is where Andromeda operates.

At any given moment, Meta has tens of millions of active ads. When a user opens their feed, the platform can’t evaluate all of them, so the first thing it does is narrow the field, from tens of millions to a few thousand candidates worth scoring and ranking. That’s the retrieval stage. Andromeda is the AI system running it.

Under the old retrieval logic, your audience settings were a significant filter in determining which ads entered the candidate pool for a given user. Andromeda, as Meta’s own engineering team described, uses a deep neural network that analyses the creative itself (what the ad communicates, what signals it carries, how it aligns with a user’s behaviour and interests) to decide which ads compete for each impression. The system delivered an 8% improvement in ads quality and a 6% improvement in retrieval recall after launch.

There’s a platform logic to this design too. When an advertiser can force-serve irrelevant ads just by outspending others for that audience, that’s a worse experience for users. Creative-based retrieval means the ads competing for each impression are the ones that match individual user interests, regardless of spend. Meta hasn’t described Andromeda as a user experience improvement, but it’s the natural consequence of how the system works.

Your creative is now doing more of the targeting work than your audience settings are.

What this means for your targeting settings

Your audience parameters still matter, but they’re no longer the primary mechanism determining who sees your ads. If your creative speaks specifically to the concerns of a 45-year-old tradie in Perth, Andromeda will find that person, even with broader audience settings. If your creative is generic enough to speak to nobody in particular, tight audience settings won’t save it.

Advertisers still spending most of their optimisation time on audience architecture are adjusting the volume knob on a system that has already moved to a different channel. The social media advertising landscape has been moving this direction for a while; Andromeda made it structural rather than directional.

If you’re running Advantage+ campaigns, which is where Meta is steering most advertisers, this shift is even more pronounced. Advantage+ uses broad targeting by default, which means the creative assets you feed it are the primary variable the algorithm has to work with. Give it a diverse creative library and it can match ads to the right people across a wide audience pool. Give it two or three versions of the same ad and it’s doing its best with limited inputs.

The Practical Gap for Australian SMEs

What most businesses are still doing

The default behaviour for most SMEs running Meta ads is still audience-first. Build the campaign, define the audience, produce creative to suit. Creative is treated as the final step rather than the strategic foundation. When results plateau, the instinct is to refine the audience: tighter age brackets, different interest combinations, new exclusions.

There’s a budget dimension here too. For businesses spending under $10,000 per month on Meta, granular audience segmentation creates a structural problem: the algorithm needs volume to optimise. When you split your budget across five or six tightly defined ad sets, each segment gets too little spend to generate enough data for meaningful learning. Campaigns that look like an audience problem are often a structure problem.

What Advantage+ actually needs from you

There are two things the algorithm needs from you that it can’t supply itself: creative diversity and budget concentration. Of those two, creative diversity is the one most Australian SMEs are under-delivering on.

Creative diversity doesn’t mean more content. It means more angles. If all your ads make the same argument in slightly different visual formats, the algorithm treats them as functionally the same asset. Three square images, one carousel, and two Reels all saying “we’re Perth’s best [service], contact us today” is one creative angle in six containers.

Meaningful diversity looks different: one ad speaks to cost anxiety, another builds trust through a client outcome, a third addresses a specific objection. Different problems, different emotional registers, different reasons to act. That’s what gives Andromeda something to match against different users in different mindsets.

For most SMEs at the $2,000–$10,000 per month mark, we’d recommend aiming for five to eight genuinely distinct creative concepts as a starting baseline, not the 15 or 20 you might see recommended at enterprise scale, but enough that the algorithm has real variation to work with. That’s our own read based on what we see across accounts, not a Meta-prescribed number.

What “Creative Diversity” Actually Means

Variety in angle, not just format

The most common creative mistake we see in Meta accounts is producing multiple assets that are formally different but conceptually identical. A static image, a Reel, and a carousel carrying the same message look like creative diversity on a spreadsheet but function as a single concept in the algorithm’s eyes.

True variety means different angles on why a customer would choose you. Three useful dimensions to vary across your creative library:

• The problem you’re addressing. Cost, trust, time, quality, risk: different customers are driven by different concerns. An ad speaking to someone anxious about wasting ad spend lands differently from one aimed at someone justifying a new agency to their CFO.
• The emotional register. Reassurance-oriented creative (“here’s why we’re a safe pair of hands”) attracts different people than aspiration-oriented creative (“here’s what growth looks like for businesses like yours”), even when the underlying offer is the same.
• The type of proof. A client testimonial, a specific result, a behind-the-scenes look at how you operate: each builds credibility in a different way. Our guide on where to find ad inspiration covers competitor research and creative gap analysis in more depth.

Where brand systems change the equation

Producing five to eight genuinely distinct creative concepts takes longer and costs more when your brand identity isn’t clearly defined. If your team has to make decisions about visual treatment, tone, and positioning from scratch for every campaign, production becomes slow and the output inconsistent. Assets start to look like they came from different companies, which is precisely the signal Andromeda doesn’t want to see.

A coherent brand identity (clear visual language, defined tone, consistent positioning) makes creative diversity faster to execute. Your creative team works within an established framework rather than rebuilding it for each brief. A five-concept campaign turns around in the time a poorly defined brand might take to produce two.

An ad from a recognisable brand converts at a higher rate than one that doesn’t look like it belongs anywhere, because familiarity and visual coherence build the micro-trust that makes someone click. The branding work and the paid social work reinforce each other. Clients who invest in brand identity before scaling paid social see the return in their ad performance, not just in how their brand looks.

What to Change About Your Campaigns

If you want to adapt to how the system works now, the changes are practical.

Audit your creative library first. Count how many genuinely distinct angles you’re running, not how many assets but how many angles. If you have eight ads and two angles, you have a thin library. The audit takes an hour and usually makes the problem obvious.

Consolidate your campaign structure. Fewer ad sets with larger budget pools give the algorithm more data to optimise with. If you’re splitting $3,000 a month across six ad sets, consider collapsing to two or three. The learning phase runs faster and the signals are cleaner.

Use Campaign Budget Optimisation. CBO lets Meta allocate spend across ad sets dynamically based on what’s performing. In an Advantage+ environment it generally outperforms fixed ad set budgets because the system can shift spend toward what’s working. It’s our default recommendation for most accounts.

Separate creative budget from media spend. Spending $5,000 per month on ads but nothing on producing new creative means you’re gradually running out of what the algorithm needs. A modest monthly investment in refreshed assets compounds across your media spend. If you want help mapping what your creative library is missing, our Facebook Ads management team can walk through your account and identify the gaps.

The businesses spending the most time on audience architecture are often the ones seeing the flattest results. It’s a symptom of optimising for a system that no longer exists in its old form. Andromeda changed the input the algorithm cares most about. Audience settings still matter, but creative is where the leverage is. If your Meta ads aren’t performing the way they used to, the answer is almost certainly in your creative library rather than your targeting parameters.

If you’d like to talk through what this looks like for your account, our social media advertising team works with Australian businesses across a range of industries and budgets. We can review your setup and help you work out where creative diversity is affecting your results.

Becca Humphrey

Paid Media Service Lead