How Targeting, Creative & The Meta Algorithm Work Together: Paid Social Fundamentals.

Much like the world around us, Meta Ads have changed.

Not dramatically on the surface – Ads Manager’s still kept a pretty familiar look – but under the hood, the way campaigns find the right audience is very different from a few years ago.

Where media buyers once spent hours building intricate audience targeting layers, Meta’s algorithm now does most of that work automatically. 

The focus has shifted from manual targeting to creative, signals, and smart monitoring.

If you're dipping your toes into the world of Paid Social for the first time – or the first time in a while – our Strategy Director / Development Mentor, Siann Nutting has a bit of an introductory refresher for you. This blog will walk you through the fundamentals of modern Meta advertising (incl. how the algorithm works, what marketers should actually be focusing on today, and which metrics matter most when diagnosing performance).

For brands getting started with Meta Ads, understanding these basics is the difference between campaigns that quietly burn budget and campaigns that actually scale.

From manual targeting to algorithmic targeting.

Less than two or three years ago, Meta advertising used to be all about defining the audience manually. Granularly, specifically, manually.

Want to promote chocolate? Target people who like chocolate brands.

Selling fishing gear? Target men interested in fishing.

Simple, logical… And increasingly outdated.

Meta’s algorithm is now significantly more advanced. Instead of relying heavily on manually defined audiences, the platform analyses engagement signals and creative assets to determine who should see an ad – based on how applicable it might be to them. 

Signals might include:

• Pausing to watch a video
• Clicking through to a website
• Interacting with similar ads
• Past purchase behaviour
  
  

Meta constantly collects these signals, feeds them back into the system, and refines delivery over time. You might’ve heard of GEM as well, talked about in relation to Meta.

It’s a Generative Ads Recommendation Model (GEM!) driven by AI launched in 2025 that revolutionises advertising across Facebook and Instagram. 

Now Meta’s largest recommendation system, it uses LLM-inspired technology to analyse user behavior, understand creative content, and predict conversions – resulting in a 5% increase in conversions on Instagram and 3% on Facebook, just in its earliest days.

It also works in tandem with Meta’s Andromeda, which acts as the retrieval system to collect and curate the ads for delivery. 

Think of Andromeda as a “concierge” that maps relevance for each user. It searches through millions of potential ads to create a short list of options for a user in real-time, whilst GEM is the one that analyses user behaviour to begin with, thereby refining the short list and identifying the "will convert" winners after. 

And as AI frequently does, it’s now gotten even better. 

The more data the platform receives, the smarter it becomes at identifying people who are likely to convert. It’s why (shockingly!) broad targeting has become the recommended starting point for many campaigns of late.

Broad targeting might work better than you think.

The biggest shift, perhaps. And it can feel counterintuitive at first.

Surely narrowing the audience should make ads more efficient?

Not necessarily. The algorithm, powered by Meta’s GEM, can often identify high-value audiences that marketers would never think to target manually.

For example, a chocolate campaign might assume the audience should be chocolate lovers. But the algorithm may discover a strong segment of people buying gifts for partners (aka. an audience in a headspace to shop, but would likely be missed with narrow targeting).

Broad targeting gives Meta room to learn. And that’s where the platform really shines.

Creative is the new targeting.

If we had a dollar for every time we said that, watch us rolling in dough.

But the fact is, ad creative is now the heaviest determinant of who actually sees an ad.

Meta’s algorithm scans ad copy, images, and videos to identify signals within the content itself. Those signals help the system decide which users are most likely to engage.

An image of a chocolate bar? Meta may serve that to users who frequently engage with food content. A reference to a game like Animal Crossing? Expect the algorithm to lean toward users who have interacted with gaming-related content.

This dynamic has been compared to speed dating for attention.

The hook, the visual, and the message all need to land quickly. If they do, then Meta’s algorithm gets the signal it needs to keep pushing the ad to the right audience.

If they don’t… The passive scroll wins, which is why creative testing is one of the most important levers in Meta advertising today. You’ll hear us preach that 101 times.

About the Meta Ad Auction, and why relevance wins.

Every ad impression on Meta is the result of an auction.

But winning that auction isn’t just about bidding the most money. 

Meta prioritises relevance. If the platform believes an ad will resonate strongly with a particular user, it’s likely to win the auction – even if other advertisers place a higher bid.

This is why creative quality and audience signals directly impact cost efficiency.

When relevance is strong, ads perform better and auctions become cheaper to win. When relevance drops, costs tend to rise. Correlation, hey? 

Check out this case study where we proved exactly that, in partnership with The Y.

Metrics that actually matter.

Monitoring the right metrics is essential when diagnosing campaign performance.

Some of the most important include:

• CPM (Cost Per Mille)
  The cost to reach 1,000 people. Rising CPMs can signal creative fatigue, audience saturation, or increased competition.
  
  
• Click-Through Rate (CTR)
  The percentage of users who click the ad. A strong CTR usually indicates that the creative is resonating, and people are one step closer to conversion.
  
  
• CPC (Cost Per Click)
  How much each click costs.
  
  
• CPA (Cost Per Acquisition)
  The cost required to generate a conversion.
  
  
• ROAS (Return on Ad Spend)
  One of the most important profitability metrics in performance marketing.
  
  
• Frequency
  How often one person sees an ad. If frequency climbs too high, it may mean Meta is struggling to find new audiences and serving ads to the same eyes repeatedly.

‍

Metrics rarely exist in isolation. For example, a higher CPM might still be perfectly acceptable if the click-through rate (CTR) is strong and conversions remain healthy.

Context matters. 

And seasonality matters too. During major retail periods like Black Friday, CPMs often rise simply because more advertisers are competing for attention.

Understanding the Meta campaign structure.

Meta Ads follow a simple three-layer structure.

Campaign

This is where the objective is defined (e.g. driving sales, leads, or traffic).

Ad Set

This layer controls targeting and budget. Most modern strategies favour broad targeting at this level, with a few smart exclusions. For example, excluding customers who purchased within the last 30 days to avoid bombarding existing buyers with acquisition ads.

Ad

Now this is where the creative lives. Images, videos, and copy all sit here — and this is where creative testing becomes critical.

Different messages, hooks, and formats allow the algorithm to learn which creative resonates best with different audience segments.

The modern media buyer’s job.

Automation hasn’t removed the need for marketers, it’s simply changed the job.

Instead of manually defining audiences, today’s media buyers focus on:

• Creative strategy
• Structured testing
• Monitoring performance signals
• Diagnosing performance changes
  
  

When performance shifts, marketers essentially become detectives.

The goal is to isolate the cause, form hypotheses, and run controlled tests to uncover what’s really happening. Insights might come from unexpected places – such as spikes in website traffic from certain demographics, or changes in conversion rate during specific periods.

If the platform / GEM receives better signals (incl. website activity and customer data), Meta’s algorithm can optimise way more effectively.

Final Takeaways? Feed Meta all the right signals.

That’s how today’s best-performing campaigns do it, instead of focusing on tightly defined audiences and generic ad creatives.

That means:

• Starting with broad targeting
• Investing in strong, varied creative
• Monitoring the right performance metrics
• Continuously testing and learning
  
  

For brands and marketing teams looking to build these capabilities internally, understanding these fundamentals is the first step.

For teams that want to go further, our DataSauce Academy offers tailored training and upskilling programs designed to help marketing teams build practical expertise in channels like Meta Ads and performance marketing more broadly – bespoke, of course.

You can learn more about it here. 

Alternatively, if you'd like our experts to do it all for you? Works too, get in touch here.