The way Meta Ads work has fundamentally changed. Not incrementally, not in a way that requires minor adjustments to your existing playbook; the underlying logic of how campaigns are structured, how audiences are defined, and where competitive advantage actually lives has shifted.
For advertisers who built their approach on the old model, interest-based ad sets, layered lookalikes, and carefully segmented retargeting buckets, the instinct is often to hold on. The data increasingly suggests that is the wrong instinct.
Advantage+ Shopping Campaigns (now rebranded as Advantage+ Sales Campaigns) and Advantage+ Audience are Meta's AI-native campaign structures. They are not simply automated versions of the manual campaign structure. They represent a different philosophy about where optimisation decisions should be made.
In a standard manual campaign, you define the audience and Meta serves the ad within your parameters. In Advantage+ campaigns, you provide creative assets, a conversion objective, and optionally some audience signals, and Meta's algorithm determines who sees what, when, and at what frequency. Budget allocation, placement selection, and audience expansion are all handled by the system.
Meta’s Business Help Centre describes Advantage+ Audience as a system that treats your manually specified targeting as a "suggestion" rather than a hard constraint. The algorithm will expand beyond those suggestions when it calculates a higher probability of conversion.
This is not a minor detail. It means that the detailed targeting inputs that used to define your campaign are now, at best, guardrails and, at worst, friction.
The evidence for broad targeting outperformance is not anecdotal. It reflects the structural reality that Meta's ad auction is now optimised for outcome prediction at a scale that manual targeting cannot match.
When you specify a narrow interest-based audience, you are capping the pool available to the algorithm. That cap forces the system to find buyers within a constrained set, even when the highest-probability converters may sit outside it. Broad targeting removes that cap and lets the model work across the full eligible population.
This works particularly well when accounts have strong conversion signal: sufficient pixel events, CRM data uploaded via Customer Match, and consistent purchase or lead data feeding the algorithm. Advertisers with fewer than 50 conversions per week often find that broad targeting produces inconsistent results because the model does not have enough data to identify patterns reliably.
Jon Loomer, one of the most technically rigorous commentators on Meta advertising, has documented the trajectory of these changes extensively on jonloomer.com. His analysis of how Advantage+ Audience interacts with manual targeting inputs is worth reading before making any structural changes to your campaigns.
The classic Meta campaign structure looked like this: separate campaigns for cold prospecting (interests, lookalikes), warm audiences (website visitors, video viewers), and hot retargeting (cart abandoners, past purchasers). Each had its own budget, bid strategy, and creative.
That structure made sense when Meta's algorithm needed explicit instructions about who to target. It no longer does. The silos you built to separate audience types are now largely redundant because Meta's AI is making those distinctions dynamically at auction time, often more accurately than manual segmentation allows.
What this means in practice: consolidating campaigns into fewer, broader structures typically improves performance because it gives the algorithm more budget to work with and more conversion events to learn from. Fragmented campaigns with small budgets per ad set produce slow learning phases and inconsistent results.
Consolidation is not the same as abandoning structure entirely. There are still legitimate reasons to maintain separation.
Keep separate: brand campaigns (to maintain visibility and messaging control), campaigns targeting specific geographic markets with distinct creative, and any campaign where performance measurement genuinely requires isolated budget tracking.
Consolidate: prospecting and retargeting into Advantage+ campaigns where conversion volume supports it (50 or more conversions per week is the commonly cited threshold), multiple interest-based ad sets targeting overlapping audiences, and lookalike audiences at different percentage ranges.
The practical test is simple: if two campaigns are targeting audiences that would plausibly overlap in an open auction, the algorithm is already managing that overlap internally. Separating them does not give you more control; it just limits the data available to each campaign's optimisation process.
This phrase has become a cliché, but it is accurate. In a broad-targeting environment, your creative does the work that audience segmentation used to do. The algorithm uses creative engagement signals to identify who responds to what. A video that resonates with a 35 to 50-year-old B2B decision-maker will, over time, train the algorithm to find more people like them, even if you never specified that demographic.
The implications for creative strategy are significant. Volume matters: more creative variations give the algorithm more signal to work with. Differentiation matters: creative that looks and sounds like everything else in the feed will not produce distinct enough engagement signals to guide optimisation effectively. Speed matters: winning creative degrades over time as audiences saturate, so a systematic production and testing pipeline is now a core operational requirement, not an optional extra.
First-party data is your primary lever for steering Meta's AI in a broad-targeting world. The three most important inputs are:
Customer Match uploads: uploading your CRM list tells Meta who your existing customers are. This both informs the lookalike model and enables you to cap spend on existing customers if you are focused on acquisition.
Pixel events: a properly configured Meta Pixel firing on view content, add to cart, initiate checkout, and purchase events gives the algorithm a full picture of your funnel. Missing events create gaps in the model.
Custom conversions: if your business has a longer sales cycle, defining intermediate conversion events (such as lead qualification stages or content downloads) gives the algorithm signal to optimise toward in the absence of sufficient purchase data.
The shift to broad targeting changes what you should be testing. Audience testing is largely obsolete; the algorithm will find the right audience if the creative and the conversion signal are strong. Creative testing is the primary lever.
A practical testing framework: run a consistent testing budget (roughly one times your target CPA per day, per test) for seven days per creative variant. Use the campaign-level Advantage+ structure so the algorithm can allocate toward winners. Identify the creative variable you are testing per round: hook, format, value proposition, or call to action. Scale winning variants into your main campaign and replace low performers rather than simply pausing them.
Meta's default attribution window is a seven-day click and one-day view. GA4 uses data-driven attribution by default, which distributes credit across touchpoints. These two models will rarely agree, and the discrepancy is not a sign that one is wrong; it reflects the fact that different tools are measuring different things.
For B2B or considered-purchase advertisers, view-through conversions are particularly significant. Meta will report conversions where someone saw an ad but did not click, then converted later through another channel. GA4 will typically attribute that conversion to the last click. Neither is the complete picture.
For larger budgets (broadly, above £100k per month in paid social), Media Mix Modelling provides a more robust framework for understanding Meta’s incremental contribution. For smaller budgets, a disciplined UTM structure and consistent GA4 event taxonomy will get you most of the way there.
The Viaduct Generation Intelligence service covers data infrastructure and reporting frameworks that make cross-channel attribution practical. If you are working through how to restructure your Meta campaigns around these principles, the Blueprint service is a useful starting point for strategic planning.
