tl;dr
The performance of most brands depends on just a few creatives — when those fatigue, the whole business tanks
Stop trying to replicate why one creative worked (micro learnings) and instead learn what works across your entire library (macro learnings)
The tool that makes macro learnings possible is a rigid naming convention that lets you analyse performance by creative characteristics
Once you understand what works, you can build a diverse creative library that speaks to different audiences and keeps scaling
The problem: creative concentration risk
Most ad accounts look the same. You might have 50 or even 100 ads running, but the top five creatives account for 80%+ of spend. That's a concentration problem.
When a handful of creatives carry your entire budget, your business performance depends on those creatives continuing to work. And creatives stop working — that's just how it goes. When they do, your ROAS drops, your CAC spikes, and your growth stalls. You're exposed because you didn't diversify.
There are two things every brand should be doing to solve this:
Increase your win rate — so more of the new creatives you launch become winners
Create a more diverse creative library — so you're not dependent on one style, hook, or audience
From micro learnings to macro learnings
When a creative performs well, the natural instinct is to figure out why and replicate it. Maybe you try a new hook on the same format. Maybe you get the same creator to record a follow-up. This is micro learning — trying to reverse-engineer one specific creative.
The problem: why a particular ad worked is often random. If you re-uploaded that same winning ad today, it probably wouldn't perform the same way. Ads are noisy. Attributing success to a specific element of one creative and trying to copy it is usually a dead end.
What works instead is macro learning: understanding what characteristics perform best across multiple creatives. If a pattern shows up in five different ads that all worked, that's a signal worth acting on. A single data point is noise — a pattern across your library is insight.
The shift from micro to macro is the foundation of a scalable creative process.
The tool: naming conventions
To generate macro learnings, you need to be able to group and compare creatives by their characteristics — hook type, content style, target avatar, benefit, format, and so on. That requires clean, structured data attached to every ad. Naming conventions are how you create that data.
Most ad accounts fail at this in one of three ways:
No structure at all — ads named whatever, no pattern
Inconsistent structure — some follow a system, others don't (common when you work with an agency that adds their prefix to their creatives but nothing else follows suit)
Too limited — some structure exists but it only captures one or two dimensions, nowhere near enough to generate real learnings
What to include in your naming convention
There's no single universal template — it depends on your business — but a good starting point covers:
Creative name — a short human-readable label so you know what the ad is without watching it
Creative ID — a unique identifier
Variation — if you're running multiple versions of the same idea
Content style — UGC, brand content, static, etc.
Hook — what's used to grab attention in the first few seconds (can be split into written hook, visual hook, audio hook)
Offer — what you're promoting (taster pack, hero product, bundle, etc.)
Avatar / target group — who this creative is aimed at
Benefit — what you're leading with (ingredients, results, price, lifestyle, etc.)
Format — video, image, carousel
Creator — who's in the ad (if applicable)
Posting page — which page it's running from
Landing page type — PDP, advertorial, collection page, etc.
Exact landing page URL — the specific destination
A concrete example: hair.video-call.ID001.v1.UGC.secret-knowledge.taster-pack.woman-lifestyle.ingredients.video.holy-page.promo-page.taster-pack
Key rules for naming conventions that actually work
Use a clear, unique separator. A dot (.) works well because people almost never use dots in place of spaces. Dashes and underscores are common separators but people also use them within words, which makes parsing ambiguous. Pick one separator and make it sacred — it can only ever divide characteristics, never appear inside one.
Use dropdowns, not free text. For fields like avatar, benefit, offer, hook, and content style, enforce a fixed list of options. If team members type these freehand, you'll end up with "Mum", "mummy", "Mom", "mother" as four separate categories that are actually the same thing. Dropdowns eliminate this. Creative names can stay free text, but everything else should be a pick-list — even if that list keeps growing.
Always keep the structure, even when a field doesn't apply. If a creative doesn't use a specific creator, don't skip the field — put 0 or NA. Skipping fields breaks the structure and makes it impossible to reliably parse later.
Use auto-tagging to inject the ad name into your tracking parameters. Most ad platforms let you dynamically insert the ad name into UTM parameters or other tracking fields. Set this up so you can match ad names back to performance data downstream without manual work.
This takes time — stick with it
Naming conventions feel like overhead, especially at the start. There's no immediate payoff. The learnings only become meaningful after a few months, once you have enough volume to spot patterns.
The failure mode is starting the system and then letting it slip when things get busy. That wipes out the compounding value. Force consistency across your team. The payoff is real — it just takes time to materialise.
How to generate macro learnings from naming conventions
Once you have a few months of consistently named ads, you can start to slice your performance data by creative characteristics.
A simple example: group all your ads by avatar and compare performance across groups. You might find that you've created a lot of ads targeting the "mummy" avatar, but the return isn't there — while the fitness niche has fewer ads but much better performance and lower spend. That's a macro learning: deprioritise mummy, invest more in fitness.
Then you go deeper: what hook type works best for the fitness avatar? What content style? What benefit? You can cross-reference any dimension against any other and start to build genuine best practices that are grounded in data, not guesswork. The goal is to always be starting new creatives from the combination of characteristics most likely to work — that's how you improve your win rate.
Why diversity matters
A side effect of only running micro learnings is that your entire creative library starts to look the same. Different hooks, but similar formats, similar messaging, similar personas.
The best-performing ad libraries right now are highly diverse — and that's not an accident. Meta has become extremely good at matching ads to the people most likely to respond to them. The catch is that different people respond to fundamentally different creative styles. If all your ads look the same, you're only reaching the slice of your audience that responds to that one style.
A good example: PipDex sells educational card decks at a price point of $100–$570. Despite that high price point, they're reportedly doing around $15M in annual revenue. Their ad library is extremely varied: educational content, viral and mystery-style videos, two-person dialogue format, post-it note aesthetics, different products, different funnel stages. The bottom-of-funnel ads are mostly static; mid-funnel and awareness content is video-heavy with diverse formats and personas.
The reason for that diversity is strategic. As you scale, you eventually saturate a target audience — you've reached most of the people in it who were likely to buy. Growth then comes from conquering new audiences. New audiences respond to different messaging and different creative styles. A library built around one format won't scale across multiple audience segments.
Even within a single target group, a more diverse creative set typically outperforms a homogeneous one. Macro learnings tell you what works across the board — diversity ensures you're reaching the full range of people those learnings apply to.