There is a persistent belief in Meta advertising that the key to better performance is better targeting. Build tighter lookalike audiences. Stack more interest layers. Exclude more demographics. Test Broad vs. 1% Lookalike vs. 3% Lookalike. Advertisers spend hours every week in Audience Manager, tweaking parameters, building custom audiences from pixel events, and constructing elaborate exclusion rules — all in pursuit of showing their ad to exactly the right people.
This approach made sense in 2020. It is actively counterproductive in 2026. The single biggest shift in Meta advertising over the past three years is the inversion of the targeting-creative hierarchy, and most advertisers haven't caught up.
The Old Model: Targeting First, Creative Second
For nearly a decade, the mental model of Meta advertising was straightforward. Find your audience first. Build detailed targeting parameters using demographics, interests, behaviors, custom audiences from your CRM, lookalike audiences from your best customers, and website visitor retargeting pools. Once you had a well-defined audience, you created ads designed to convert that specific group. The creative's job was to be compelling within a pre-defined audience.
Media buyers were evaluated primarily on their targeting skill. The ones who could build the most precise audiences, find the most responsive lookalike seed lists, and layer the right exclusions were considered the best operators. Creative was important, but it was downstream of targeting in both workflow and strategic priority.
This model worked because Meta's delivery system at the time was fundamentally an audience-matching engine. You told the algorithm who to target, and it showed your ads to people within those parameters. The algorithm's optimization was limited to selecting the best individuals within your defined audience — it couldn't override your targeting decisions, even when better options existed outside your parameters.
What Andromeda Changed
Meta's Andromeda algorithm, which has undergone several major iterations since late 2023, fundamentally restructured how ad delivery works. Instead of matching ads to pre-defined audiences, Andromeda uses a retrieval-based ranking architecture that processes creative signals to discover audiences dynamically. The algorithm doesn't just find the best people within your targeting — it uses your creative as a signal to identify the best possible audience across Meta's entire user base.
When you upload an ad creative, Andromeda generates a multidimensional representation of it — what's commonly called an Entity ID — that encodes its visual characteristics, copy patterns, emotional signals, and structural properties. This Entity ID is then used in a two-stage retrieval process: first, a broad pass that identifies thousands of candidate users whose engagement patterns align with the creative's signal profile, and second, a precision ranking that scores those candidates on predicted conversion likelihood.
The critical implication: your creative doesn't just communicate a message. It acts as a targeting mechanism. A lifestyle image with warm tones and benefit-focused copy attracts a fundamentally different audience than a studio product shot with feature-focused copy — even when both ads use identical targeting parameters. The creative's signal patterns are telling Andromeda which users to find.
Why Tight Targeting Now Hurts Performance
If Andromeda uses creative signals to build audiences dynamically, then constraining targeting actually limits the algorithm's ability to find your best customers. Here's the mechanism.
When you set narrow targeting parameters — a 1% lookalike audience from purchasers, filtered by age 25 to 45, in specific metro areas — you're telling the algorithm: "Only search within this pool of roughly 200,000 to 500,000 people." Andromeda can still use your creative signals for ranking within that pool, but it cannot discover responsive users outside it. Your best potential customer might be a 52-year-old in a smaller city who doesn't match your lookalike profile but responds powerfully to your creative's hook type. With narrow targeting, Andromeda never gets to test that hypothesis.
The data consistently supports this. Multiple independent studies and Meta's own published performance benchmarks show that Advantage+ campaigns — their broadest targeting option — using diverse creative sets outperform narrowly targeted campaigns on CPA by 15 to 30% across e-commerce verticals. The broader you allow Andromeda's search space, the better it can leverage creative signals for audience discovery.
There is a psychological barrier here that many experienced media buyers struggle with. Broad targeting feels irresponsible. It feels like spraying money into the void and hoping the algorithm figures it out. But that intuition is calibrated to the old system. In the Andromeda era, broad targeting plus diverse creatives isn't irresponsible — it's giving the most sophisticated audience discovery engine in history the maximum data and space to work with.
Creative Signals: The New Targeting Dimensions
If creative is now your primary targeting lever, then understanding what signals your creative sends is essential. Based on observable delivery patterns and Meta's published research, creative signals operate across five primary dimensions.
**Visual composition** includes layout structure, product placement and sizing, use of whitespace, and background complexity. A flat-lay product arrangement signals a different user preference than a lifestyle in-context shot. Andromeda clusters users by which visual structures they historically engage with.
**Color and tone** encompasses warm versus cool palettes, high contrast versus muted treatment, and background types. Color psychology in advertising is well-established, but in the Andromeda context, color treatment doesn't just influence perception — it actively routes your ad to users with matching visual preferences in their engagement history.
**Text density and structure** covers headline size, body copy volume, CTA placement and prominence, and the ratio of text to image. Heavy-text ads attract users who engage with informational content. Minimal-text ads attract impulse-driven users. The same product advertised with different text treatments reaches genuinely different audience segments.
**Hook type** is the psychological trigger in the opening element — whether that's curiosity, fear of missing out, social proof, a direct benefit claim, or a pattern interrupt. Different hooks activate different motivational frameworks in users, and Andromeda has learned which hooks resonate with which user clusters through billions of historical data points.
**Emotional register** reflects whether the ad's overall tone is aspirational, urgent, educational, humorous, or empathetic. This dimension interacts with all the others — a humorous UGC-style ad with warm colors sends a categorically different signal than a serious, data-driven infographic with cool tones, even if both promote the same product.
The Diversity Imperative
Understanding creative signals explains why creative diversity is so critical in 2026. Each distinct creative sends a different set of signals, which means each distinct creative targets a different audience segment. Running five similar-looking ads is like targeting five overlapping audiences — you're covering the same ground five times. Running five genuinely diverse ads is like targeting five different audience segments simultaneously, letting Andromeda discover which segments convert best.
This is the mechanism behind the frequently cited recommendation of running 15 to 25 active creatives. It's not about volume for volume's sake. It's about signal coverage. More diverse creatives mean more distinct signal combinations, which means Andromeda is exploring more of the addressable market. With only 3 to 5 creatives, you might be reaching 20% of your potential audience. With 20 diverse creatives, you might be reaching 60 to 70%.
The key word is "diverse." Two ads that look visually different to you — different product images, different background colors — can have highly similar Entity IDs if they share the same layout structure, text positioning, and hook type. Andromeda clusters these as redundant and treats them as a single signal. Genuine diversity requires variation across all five signal dimensions: different layouts, different color treatments, different text structures, different hooks, and different emotional registers.
The New Media Buying Workflow
If targeting is no longer the primary performance lever, how should media buyers spend their time? Here's the workflow that reflects how Meta advertising actually works in 2026.
**Strategic time allocation.** In the old model, media buyers might spend 60% of their time on audience construction and 40% on creative. The inverted allocation is closer to what works now: 20% on targeting (mostly setting up Advantage+ campaigns with broad parameters) and 80% on creative strategy, production, and analysis.
**Creative strategy replaces audience strategy.** Instead of building audience hypotheses ("people interested in yoga who bought from competitor X"), build creative signal hypotheses ("a UGC-style testimonial with warm tones will reach aspirational buyers, while a data-comparison graphic with cool tones will reach analytical buyers"). Your creative brief is your targeting brief.
**Testing means creative testing, not audience testing.** When performance plateaus, the correct response is to launch new creative variations — not to create new audience segments. New creatives send new signals, which unlock new audience pockets. New audiences with the same creatives just constrain the algorithm's search space in a different way.
**Analysis shifts from audience reports to creative reports.** The most useful data in 2026 is creative-level performance data: which creatives drive the best CPA, how quickly they decay, which signal patterns (visual style, hook type, copy angle) consistently outperform. This data informs your next round of creative production. Audience demographic reports, which many media buyers still study obsessively, tell you who the algorithm chose to show your ads to — not who you should target.
Addressing the "But My Retargeting Works" Objection
The most common pushback on this thesis is retargeting. "My website visitor retargeting campaigns have a 5x ROAS. How can you say targeting doesn't matter?" Retargeting is an exception — but a narrower one than most people think.
Retargeting works because the audience is pre-qualified by their own behavior (visiting your site, adding to cart). You're not building a targeting hypothesis; you're responding to demonstrated intent. This is fundamentally different from prospecting, where you're trying to find new customers who haven't discovered you yet.
However, even retargeting is less differentiated than it was. Advantage+ campaigns now include retargeting audiences automatically in their delivery optimization. Meta's data shows that in many cases, letting Advantage+ handle the balance between prospecting and retargeting produces better overall ROAS than manually separating the two into distinct campaigns. The algorithm is better at determining the optimal retargeting frequency and timing than most manual setups.
What This Means for Your Budget
If you're currently spending $2,000 per month on a media buying agency or specialist whose primary value-add is audience construction and targeting optimization, you should redirect that investment toward creative production. The ROI of another 20 diverse creatives per month dramatically exceeds the ROI of further audience refinement in almost every scenario tested.
This doesn't mean media buying expertise is obsolete. Campaign structure, budget allocation, bid strategies, and performance analysis still require skill and judgment. But the center of gravity has shifted. The competitive advantage in Meta advertising today comes from the ability to produce diverse, high-quality creatives at a pace that keeps up with Andromeda's appetite for fresh signals. Advertisers who generate creatives efficiently — whether through AI tools, in-house design teams, or streamlined production workflows — consistently outperform those who invest primarily in targeting precision.
The targeting obsession is a legacy of a system that no longer exists. The sooner you shift your strategic energy from finding the right audience to creating the right signals, the sooner your Meta ads performance will reflect what's actually possible in 2026.
Related Reading
Understand Meta's Andromeda algorithm and how creative classification works at the system level. Learn about Entity ID diversity and why 15 to 25 creatives is the minimum for effective signal coverage. See the creative testing framework for how to test creative signals systematically. And explore how AdRiseLab generates diverse creatives from any URL to build the creative volume your account needs.