Creative Diversity on Meta: How Engagement-Based Delivery Changes Creative Strategy

A common pattern in Meta advertising goes like this: an ad performs well, budget is increased, and within a couple of weeks performance declines. The usual response is to iterate on the ad, adjusting headlines, visuals, or hooks, with mixed results.

This pattern is often interpreted as a creative quality problem. A more useful interpretation is structural. The way Meta matches ads to people has changed, and many creative strategies were designed for the system that came before it.

From interest categories to behavioural signals

For much of the last decade, Meta's targeting relied heavily on interest categories. Users were grouped based on pages they followed and content they interacted with, and advertisers selected the categories they believed matched their customers. In that model, targeting defined the audience and creative was primarily responsible for conversion.

Meta's delivery has since moved toward a model driven by engagement behaviour: what a person watches, how long they watch it, what they save, share, comment on, or scroll past. These signals are continuously updated and reflect not just what someone is interested in, but how they tend to respond to different kinds of content. Meta's recent ad retrieval infrastructure, publicly referred to as Andromeda, places considerable weight on the characteristics of the creative itself when deciding which users an ad is likely relevant to.

The practical implication is that creative now functions as a targeting input. The message, format, tone, and visual language of an ad provide signals that help the system determine who it should be shown to.

Why broad audiences are really collections of segments

Interest categories are coarse by design. A category such as "health and wellness" contains people with very different motivations, levels of knowledge, and reasons for buying. Under interest-based targeting, those differences were largely invisible, so a single message had to serve the whole group.

Behavioural signals make those differences more visible to the delivery system. This raises a question that many advertisers haven't explicitly addressed: if the system can distinguish between different kinds of buyers, does the creative in the account give it enough variety to match each of them?

Consider a brand selling a mushroom-based coffee alternative. A conventional persona might be "adults interested in coffee and wellness." In practice, the buyer base is likely to include several distinct groups, for example:

• People reducing caffeine for comfort reasons. They value the ritual of a morning drink but experience jitteriness or anxiety from coffee. Messaging about calm energy and continuity of routine is likely to resonate, often in a relatable, lifestyle-oriented format.
• People optimising for focus and productivity. Their interest is sustained concentration without an afternoon energy dip. Mechanism explanations, ingredient detail, and direct comparisons with coffee are more likely to be persuasive.
• People focused on digestive health. They tend to evaluate sourcing, formulation, and expert input carefully, and respond to transparency and detailed evidence.

Each group has a different relationship with the underlying problem and finds different kinds of proof credible. An account whose creative consistently communicates a single message, such as "focus without the crash," is likely to be matched primarily with the group that message addresses. The other groups remain reachable in principle, but the account contains little that the system would identify as relevant to them.

Defining creative diversity

In this context, creative diversity refers to variation at the level of concept rather than execution. Changing colours, button text, or headline wording produces variations that are likely to be interpreted as broadly similar signals.

Conceptual diversity involves differences along several dimensions:

• Angle: the problem, motivation, or belief the ad addresses
• Audience: the type of buyer the ad is evidently intended for
• Evidence: the form of proof used, such as testimonials, data, demonstrations, or expert endorsement
• Format and tone: for example UGC video, static comparison, founder-led explanation, or educational content

A simple heuristic is to describe two ads in a single sentence each. If the descriptions suggest different intended audiences or motivations, the ads are conceptually distinct. If they differ mainly in wording, they are variations of the same concept.

Creative fatigue as audience saturation

Greater specificity has a trade-off. When a concept is well matched to a particular segment, it can perform strongly within that segment, but the segment is smaller than the broad audience a generic ad would once have been shown to. As a result, the concept may reach a large share of its relevant audience relatively quickly.

Much of what is described as creative fatigue can be understood in these terms. Declining performance often reflects saturation of the segment the concept was matched to rather than a decline in the ad's intrinsic quality. Rising frequency and increasing costs per result are consistent with this interpretation.

This suggests that creative volume, meaning the ongoing introduction of new concepts, plays a different role than it once did. Its purpose is less about finding a single long-lived winner and more about maintaining coverage across segments as individual concepts reach their natural limits.

How diversity and volume interact

Diversity and volume address different constraints:

• Diversity determines the breadth of the market the account can effectively reach.
• Volume determines how long that reach can be sustained as individual concepts saturate.

When diversity is present without volume, an account may expand into new segments but struggle to sustain performance once those concepts fatigue. When volume is present without diversity, production effort is concentrated on concepts that compete for the same segment. Sustained scaling generally depends on both.

Evaluating an account at the portfolio level

This perspective shifts evaluation from individual ads to the account as a whole. Relevant questions include how many distinct concepts are active, which buyer segments they address, and how spend is distributed across them.

Within such a portfolio, ads serve different functions. Some provide stable performance over longer periods. Others perform strongly for a shorter time within a specific segment. Some are exploratory, testing whether a new angle or audience is viable. Short lifespans are not necessarily an indication of failure when viewed in this way.

Several practices follow from this approach:

1. Derive segments from customer language. Reviews, support conversations, and post-purchase surveys often reveal distinct buying motivations more reliably than internally defined personas.
2. Analyse spend by concept. Grouping active ads by angle and audience shows whether spend is concentrated in a small number of ideas.
3. Plan creative around coverage gaps. Identifying which segments or motivations lack dedicated creative can guide briefing more effectively than iterating only on current top performers.
4. Review the competitive landscape. Public ad libraries indicate which angles are heavily used in a category and which remain underexplored.
5. Maintain a concept pipeline. Developing new concepts ahead of expected saturation reduces the performance gaps that occur when replacements are created reactively.

Indicators of limited creative diversity

The following patterns are often associated with insufficient conceptual diversity:

• Performance declines consistently as spend increases
• Well-performing ads lose effectiveness within one to two weeks
• Cost per acquisition rises despite frequent testing
• Most active ads communicate essentially the same message

These patterns can have other causes, including offer, landing page, or seasonal factors, but a concept-level review of the account is a useful starting point.

Conclusion

Meta's delivery system has become increasingly capable of identifying which people are likely to respond to a given ad. It depends, however, on the range of creative it is given. Accounts built around a single message will tend to reach a single segment, regardless of how well that message is executed.

Treating a market as a set of distinct buyer groups, developing creative that addresses each of them, and maintaining a steady flow of new concepts as others saturate offers a more durable foundation for scaling than searching for individual winning ads.