44% Higher Memorability: Creative Analysis Framework for Ecommerce

A creative analysis framework is a repeatable system that breaks ad creative into testable components, turns observations into hypotheses, and measures outcomes against clear KPIs. The result is fewer wasted tests and better creative wins you can reproduce instead of guessing at. This piece walks through the components, the step-by-step process, and the tagging system I use with my team.

TL;DR:

• Effective creative analysis requires aligning objectives with measurable KPIs that connect to specific business outcomes, such as ROAS or customer acquisition cost.
• Tagging creative assets across dimensions like hook type, narrative arc, tone, and visuals is essential for deriving actionable insights and comparing campaign performance.
• Establishing a strict, repeatable research-to-evaluate process ensures tests build on past learnings, avoiding repeated mistakes and enabling incremental improvements.
• Running large-scale tests or key spending shifts should always involve incrementality studies or lift analysis to confirm true impact rather than relying on proxy metrics.
• Smaller teams may not need a formal system unless consistently testing multiple creatives weekly, but structured taxonomy and documentation become critical at higher testing frequencies.

Table of Contents

• Essential components and principles every framework needs
• The step-by-step process from research to scale
• What dimensions to tag on every creative asset
• How to write testable hypotheses and choose the right KPIs
• Building a tagging taxonomy and piping it into measurement
• A practical workflow for creative research and competitive analysis
• Top mistakes that quietly wreck creative tests
• Research-backed takeaways and how I apply them at Cosma
• When to run creative analysis yourself versus bringing in outside help
• How my team at Cosma can help you implement this
• FAQ
• Sources

Essential components and principles every framework needs

A creative analysis framework fails when it is treated as a checklist instead of a system. The components below connect to each other: your objectives shape your metrics, your metrics shape your test design, and your tagging structure determines whether you can learn anything once the test ends.

• Objectives and KPIs: tie every test to a business outcome (CAC, ROAS, incremental revenue), not just an engagement metric.
• Audience insight and creative brief: pull from customer research, reviews, and support tickets so the creative direction starts from a real objection or desire, not a guess.
• Creative variants and test design: isolate one lever per variant (hook, offer, visual style) so you know what caused the lift.
• Testing method and measurement: decide upfront whether you need true incrementality (holdout or lift study) or whether a proxy metric like CTR is acceptable for this stage.
• Tagging taxonomy and metadata: every asset gets structured tags so results are comparable across campaigns and months.
• Iteration and governance: someone owns the tagging rules, the test calendar, and the decision to scale or kill.

Teams that skip the taxonomy piece usually end up with a folder of past ads and no way to answer “what actually worked.” That is the gap a framework closes.

Pro Tip: Write your test’s success metric before you build the creative, not after you see the results.

The step-by-step process from research to scale

A creative analysis framework only works when it runs the same way every time. This is the sequence my team follows on paid social accounts, adapted from research-into-practice cycles rather than ad-hoc brainstorming.

1. Research: collect existing creative, competitor ads, and audience signals (reviews, comments, support tickets) to find patterns worth testing.
2. Deconstruct: tag every asset against your taxonomy (hook, narrative, tone, visual, CTA) so you can compare apples to apples later.
3. Hypothesize: write a falsifiable statement naming the change, the metric, the segment, and the expected direction of movement.
4. Test: design the experiment (A/B, multivariate, or holdout) and set sample size and run time before launch, not mid-flight.
5. Evaluate and decide: check results against your winner criteria and statistical thresholds, then decide to scale, kill, or iterate.
6. Document and iterate: log the hypothesis, the result, and the tag combination in your taxonomy so the next test builds on this one instead of repeating it.

The step most teams shortcut is documentation. Without it, every quarter starts from zero instead of compounding on what the last one taught you.

What dimensions to tag on every creative asset

Deconstruction is where a creative analysis framework becomes usable data instead of opinion. Every asset should carry tags across these dimensions before it ever runs as a test.

• Hook type: visual hook, direct question, bold claim, or social proof in the opening frames.
• Narrative arc and pacing: where the opening beat lands, how many story beats follow, and where the peak sits.
• Tone and emotional levers: humor, suspense, empathy, or urgency driving the message.
• Visual treatment: close-up versus wide shot, product-in-hand framing, color palette, and text overlay density.
• Audio: music presence, silence, voiceover cadence, and the timestamp of the first brand mention.
• Offer and CTA clarity: how explicit the offer is and where the call to action appears.
• Format metadata: aspect ratio, duration, and platform placement.

Each of these maps to a metric you can actually move. Hook type tends to show up in hook rate and thumb-stop ratio; CTA clarity shows up further down funnel in conversion rate. Tag loosely and you lose that connection.

Pro Tip: Tag the first-brand-mention timestamp even on ads that do not feature the brand early. It is one of the clearest predictors of retention drop-off.

How to write testable hypotheses and choose the right KPIs

A hypothesis that cannot fail is not a hypothesis, it is a hope. Use a fixed template: “If [creative change], then [metric] will move by [direction] in [segment] over [period].” For example: “If we move the product reveal to second two instead of second five, then hook rate will increase in the cold-audience segment over a 14-day test window.”

• Primary metrics: pick one metric that matches the funnel stage you are testing (CTR for hook changes, conversion rate or ROAS for offer changes).
• Secondary metrics: track view-through rate or thumb-stop ratio as supporting signal, not as the decision metric.
• Statistical considerations: define your minimum detectable effect and confidence threshold before launch, and correct for the number of simultaneous tests running on the account.
• Proxy versus incrementality: CTR and thumb-stop rate are fine for quick triage; when a result will justify a large budget shift, run a holdout or brand lift study instead.

Memorability is measurable, not just a feel. Research on generated ad variants optimized for memorability found they scored about 44% higher on memorability metrics than the originals, with significantly better recognition accuracy. That is a strong argument for including a memorability proxy in your scoring rubric before a full test, not after.

Building a tagging taxonomy and piping it into measurement

A taxonomy only earns its keep when it connects to your performance data, not when it lives in a separate spreadsheet.

• Controlled vocabulary: fields like creative_id, hook_type, visual_style, offer_type, and test_batch should use a fixed list of values, not free text.
• Pipeline: creative metadata needs a shared key (the creative_id) that joins to spend, conversion, and ROAS data in your analytics platform or dashboard.
• Dashboards and decision rules: set a rule of thumb for early stopping (for example, kill underperformers once retention curves flatten below your baseline) and a separate rule for scaling winners.
• Governance: one person or a small group owns the taxonomy, approves new tag values, and versions changes so historical comparisons stay valid.

Nielsen’s guidance on data taxonomies makes the case directly: standardized tagging is what lets you see cross-account patterns, like which hook types outperform across an entire portfolio, that a single ad review would never surface.

A practical workflow for creative research and competitive analysis

Before you can tag and test, you need a steady stream of creative candidates worth analyzing, and a way to prioritize which ones get built first.

• Sources: platform ad libraries, your own in-account archive, and creative partners or UGC contributors.
• Normalization: trim clips to consistent lengths, standardize aspect ratios, and extract on-screen text (OCR) and spoken dialogue (ASR) so you can tag and search by content, not just by filename.
• Scoring rubric: rank candidates on a memorability proxy, novelty relative to your current library, and relevance to the audience segment you are targeting.
• Tooling by team size: smaller teams can run this manually in a spreadsheet tied to a shared drive, while larger teams benefit from a dedicated creative analytics tool connected to the ad platform API.

If you want a structured starting point for the brief stage specifically, a set of creative brief questions built for scoping creative work is a useful reference before assets go into production.

Top mistakes that quietly wreck creative tests

• Optimizing for CTR instead of conversion: a high-CTR ad can still lose money if it attracts the wrong click. Fix: name the funnel-stage metric as primary before launch.
• Scaling on small samples: a one-day lead on a new variant is noise, not a signal. Fix: hold to your pre-set sample size and run time.
• Ignoring novelty and competitive saturation: an ad that worked last quarter can fatigue fast once competitors copy the format.
• Not logging metadata: untagged winners cannot teach you anything the next time you brief a creative.
• Trusting short-term proxy metrics for big budget calls: run an incrementality or lift check before a major spend shift.

Research-backed takeaways and how I apply them at Cosma

Research on ad memorability backs up what good creative testers already suspect: specific, nameable levers move recall and recognition. The Lancashire study on insight and memorability found that ads built around a moment of insight are recalled and recognized better than standard ads, since the insight experience supports memory consolidation. The WACV 2025 long-term memorability paper adds that memorability can be modeled and even generated, meaning specific visual and structural features reliably correlate with how long an ad sticks.

Think with Google’s ABCD playbook translates this into testable execution rules: multiple shots and a brand cue in the first five seconds, tight framing, and a fast emotional hook.

Standardized tagging turns a pile of individual ad opinions into comparable data you can learn from across accounts.

I build these findings directly into client tagging systems, scoring new creative concepts on insight and pacing before they ever go into a test.

When to run creative analysis yourself versus bringing in outside help

Running this framework well takes a tagging discipline and a test cadence most lean teams do not have bandwidth for alongside daily media buying. If your account runs fewer than a handful of creative tests a month, you likely do not need a dedicated system yet. Once you are testing weekly across multiple audiences, a structured taxonomy stops being optional. My team at Cosma has built this exact pipeline for e-commerce accounts scaling past that threshold; our case studies walk through the outcomes.

— Stefano Mazzei

How my team at Cosma can help you implement this

Building a working creative analysis framework from scratch takes months of trial and error most e-commerce teams cannot afford to lose on a shrinking ad budget. I run creative testing, taxonomy setup, and Meta Ads optimization for DTC and e-commerce brands, combining customer research with a tagging system built from the same memorability and performance principles covered above. If you want a second set of eyes on your current testing process, you can book a discovery call.

FAQ

What is creative analysis?

Creative analysis is the practice of breaking an ad into its component parts, hook, narrative, tone, visuals, and CTA, and connecting each part to performance data so you know what to repeat or change. It turns subjective creative review into a measurable, repeatable process.

What are some frameworks for understanding creativity?

In advertising, the two most cited practical frameworks are Think with Google’s ABCD playbook for video opening structure and emerging memorability modeling research that scores creative on recall and recognition potential. Broader creativity theory frameworks exist in academic psychology, but for ad testing these two are the most directly applicable.

What are the 7 C’s of creativity?

Definitions of the “7 C’s of creativity” vary across sources and are not consistently defined in advertising research, so there is no single authoritative version to cite here. For ad testing purposes, the components that matter more consistently are the ones covered above: hook, narrative, tone, visuals, audio, offer, and format.

What is a marketing analysis framework?

A marketing analysis framework is a structured method for evaluating how marketing decisions, including creative, targeting, and offer, connect to business outcomes like revenue and CAC. A creative analysis framework is a specific subset focused on how ad content itself performs and how to test it systematically.

Sources

• Behavior-in-the-wild
• A playbook for building effective creative on YouTube (Think with Google)
• Long-Term Ad Memorability: Understanding & Generating Memorable Ads (WACV 2025)
• What makes creative advertisements memorable? The role of insight (Shen et al.)
• Why you need a data taxonomy (Nielsen Insights)