How to analyze ad creative performance data

Published by Shamanth at March 9, 2026

Step 1: Define Your Metric Hierarchy Before Touching Any Data

Not all metrics are created equal, and analyzing creative performance without a clear hierarchy leads to conflicting conclusions. Establish a primary KPI (usually ROAS or CPI at target), then map supporting metrics in order of proximity to revenue. The hierarchy we use at RocketShip HQ: ROAS > CPI > CVR (install rate) > CTR > Hold Rate > Hook Rate.

Separate efficiency metrics from volume metrics

Efficiency metrics (ROAS, CPI, CVR) tell you how well a creative converts. Volume metrics (impressions, spend, installs) tell you how much the algorithm trusts it. A creative with a $2 CPI on $50 spend is not proven. You need both efficiency and volume to declare a winner.

Set minimum spend thresholds for statistical relevance

We typically require a creative to have spent at least 3x your target CPI (or $100, whichever is higher) before drawing any conclusions. For ROAS analysis, you need at least 50 installs per creative to account for variance in post-install behavior.

Assign metric weights by business impact

Use RocketShip HQ's Weighted Anomaly Scoring approach: weight metric changes by business impact using the formula abs(% change) x sqrt(spend). A 15% ROAS drop on $5K/day spend scores far higher than a 40% drop on $200/day spend. This eliminates 70%+ of false alarms and keeps your attention on changes that actually move the business.

If your team argues about whether a creative is 'good' or 'bad,' you haven't defined your hierarchy clearly enough. Write it down and share it before any analysis session.

Step 2: Build Your Creative Performance Report

Pull data from your ad platform and MMP into a single view. The goal is one row per creative asset with all relevant metrics side by side. This sounds basic, but most teams either analyze platform data in isolation (missing downstream events) or rely solely on MMP data (missing top-of-funnel engagement signals).

Export platform-level data for engagement metrics

From Meta Ads Manager or TikTok Ads, export: creative ID/name, impressions, 3-second video views (for hook rate), ThruPlays or average watch time (for hold rate), clicks, CTR, and spend. Use the 'Breakdown by Dynamic Creative' or 'Asset' view to isolate individual creatives rather than ad-level aggregates.

Merge MMP data for conversion and revenue metrics

From AppsFlyer, Adjust, or your MMP, pull: installs, cost, CPI, and any post-install events (purchases, subscriptions, D7 ROAS). Join this data to your platform export using creative ID or ad ID as the key. This gives you the full funnel in one sheet.

Calculate derived metrics

Hook Rate = 3-second video views / impressions (benchmark: 25-40% is average, 40%+ is strong). Hold Rate = ThruPlays / 3-second views (benchmark: 15-25% is typical). CVR = installs / clicks. Add a column for 'cost per ThruPlay' as a proxy for creative quality that the algorithm rewards.

Automate this report. We use a simple Google Sheets + API connector setup that refreshes daily. If you're spending more than 30 minutes building this report manually, you're losing time that should go toward analysis.

Step 3: Diagnose Creative Performance Using the Funnel Drop-Off Method

Once your report is built, the most powerful analytical technique is identifying where each creative breaks down in the funnel. A creative is rarely bad at everything. It usually fails at one specific stage, and that diagnosis tells you exactly what to fix.

Flag creatives with low hook rate (below 25%)

If a creative can't stop the scroll, the opening frame or first 0.5 seconds needs work. This is a visual and headline problem, not a messaging problem. At RocketShip HQ, we apply the 4-Layer Hook System here: check whether the creative stacks a visual pattern break (like a 0.3-0.8s zoom), a text overlay under 15 words, a verbal/voiceover element, and audio that amplifies emotion. Hooks missing any of these layers typically underperform. For more on this, see our guide on how to write ad hooks that stop the scroll.

Flag creatives with strong hook rate but low hold rate

This means you grabbed attention but lost it. The creative's body (seconds 3-15) isn't delivering on the promise of the hook. Check for pacing issues, unclear value propositions, or a mismatch between the hook's tone and the rest of the ad.

Flag creatives with strong engagement but low CVR

People watched and clicked but didn't install. This usually indicates a disconnect between the ad and the app store page, or that the creative attracted the wrong audience. It can also mean the end card or CTA was weak.

Flag creatives with strong installs but poor ROAS

This is the most expensive failure. You're acquiring users who don't monetize. The creative may be setting wrong expectations or attracting deal-seekers rather than high-intent users. Revisit the messaging angle entirely.

Create a conditional formatting system in your sheet: green for top 25%, yellow for middle 50%, red for bottom 25% on each metric. This makes funnel drop-offs instantly visible across dozens of creatives.

Step 4: Identify Patterns Across Your Top Performers

Individual creative analysis is useful, but the real leverage comes from pattern recognition across your winner pool. Tag every creative with structural attributes so you can filter and aggregate. This is where creative analytics shifts from reporting to strategy.

Tag creatives by concept, format, hook type, and visual style

Build a taxonomy. For concept: problem-solution, testimonial, UGC, gameplay, story-driven, feature demo. For format: static, video under 15s, video 15-30s, video 30s+. For hook type: question, bold claim, social proof, emotional trigger. For visual style: bright/colorful, dark/cinematic, text-heavy, live action, animated. Every creative gets multiple tags.

Aggregate performance by tag

Pivot your data by each tag dimension. You'll see things like: 'problem-solution concepts have 22% lower CPI than feature demos' or 'hooks that open with a question have 35% higher hook rates.' These are the insights that shape your next creative brief.

Look for psychological patterns in top performers

As Bastian Bergmann of Solsten discussed on the Mobile User Acquisition Show, psychology-based creative changes can outperform algorithmic optimization alone. For Solitaire Klondike, shifting copy from 'train your brain' to 'hardest solitaire game' based on psychological profiling improved IPM from 0.97 to 2.4. Look for the emotional or identity-level patterns in your winners, not just structural ones.

At RocketShip HQ, we apply the 3C Principle when analyzing winning hooks: every top performer has Context (who is this for?), Clarity (what is this about?), and Curiosity (an open loop that compels continued viewing). When we tag hooks against these three dimensions, creatives with all three Cs consistently outperform those missing even one.

Step 5: Separate Emotional Angles from Execution Quality

One of the most common analysis errors is conflating a winning concept with a winning execution. A beautifully produced ad with the wrong emotional angle will lose to an ugly ad with the right one. Your analysis must separate these two variables to inform future production correctly.

Group creatives by emotional angle

Gonzalo Fasanella, CMO at Tactile Games, shared on the Mobile User Acquisition Show how Lily's Garden explored sadness, anger, and anxiety as emotional angles when 90% of competitive ads relied on 'funny or cute.' This counterintuitive approach drove massive performance because it resonated on a deeper level. Group your creatives by the core emotion they trigger and compare performance across emotion categories.

Within each angle, compare execution variations

If your 'aspiration' angle has 5 creatives, rank them against each other. The spread between best and worst execution within the same angle tells you how much production quality matters for that specific concept. Some angles are execution-sensitive (UGC testimonials), while others are concept-sensitive (story-driven ads where the narrative matters more than polish).

Fasanella's team showed creative teams only 2 KPIs to prevent analytics-driven bias from stifling creative exploration. Consider limiting which metrics your creative team sees to avoid them optimizing for CTR at the expense of emotional resonance.

Step 6: Audit Your Testing Structure to Ensure Clean Data

Before trusting your analysis, verify that your campaign structure isn't contaminating your data. Asset stuffing, where you place all creatives in a single ad set without thematic separation, is one of the most common culprits. It prevents the algorithm from identifying appropriate audience segments for each creative, making your performance data unreliable.

Check for spend concentration

If one creative in an ad set is getting 80%+ of spend, the other creatives in that set never got a fair test. Pull a spend distribution chart for each ad set. Healthy distribution shows at least 3-4 creatives receiving meaningful spend before the algorithm picks favorites.

Separate creatives thematically

Group creatives by concept or audience theme into distinct ad sets. A UGC testimonial and a cinematic gameplay trailer appeal to fundamentally different audience segments. Mixing them in one ad set forces the algorithm to choose, and usually one concept dominates before the other gets adequate data.

Account for creative fatigue in your analysis

Pull performance by week for your top creatives. Most mobile creatives experience meaningful decay within 2-4 weeks on Meta. If you're analyzing a 30-day window, a creative that was amazing in week 1 but fatigued by week 4 will look mediocre in aggregate. Segment by time period to catch this.

If you're scaling AI-generated creative volume, remember that more output requires proportionally larger test budgets. As we discussed in our piece on using AI to generate ad creatives, the hidden cost of testing is one of the most overlooked pitfalls. Producing 100 AI creatives that each get $20 in spend teaches you nothing.

Step 7: Build a Creative Insights Loop That Compounds Over Time

The final and most important step is turning individual analyses into a system that makes every future creative better. The best mobile growth teams don't just analyze performance. They build institutional knowledge that compounds.

Maintain a living 'Creative Playbook' document

After each analysis cycle (weekly or biweekly), add 2-3 bullet points to a shared document: what worked, what didn't, and one hypothesis for the next round. Over 6 months, this becomes your most valuable strategic asset. Include screenshot examples of top and bottom performers.

Create a 'Winner DNA' template for your creative team

Distill your pattern analysis into a one-page brief that describes the structural elements of your top 10% creatives: typical hook type, emotional angle, pacing, length, CTA style, and visual treatment. Update this quarterly as new patterns emerge.

Run quarterly 'exploration sprints' to avoid local maxima

If you only iterate on past winners, you'll plateau. Dedicate 20-30% of your creative testing budget to concepts that break your current patterns. Some of these will fail, but the ones that succeed often become your next generation of winners and prevent the stagnation that comes from over-optimizing on what already works.

The most dangerous trap in creative analysis is only iterating on what already works. Your data will always tell you to make more of what's winning today. But the breakthroughs come from testing concepts your data can't predict. Build that exploration discipline into your process from day one.

Common Mistakes to Avoid

Analyzing creatives with insufficient spend: Drawing conclusions from a creative that spent $30 is like flipping a coin 5 times and calling the result statistically significant. Set minimum spend thresholds (at least 3x your target CPI) before any creative earns a 'winner' or 'loser' label.
Treating all metrics equally: A creative with an 80% hook rate and terrible ROAS is not a good creative. It's an expensive scroll-stopper. Always anchor analysis to your primary business KPI (usually ROAS or CPI) and use engagement metrics as diagnostic tools, not success criteria.
Ignoring creative fatigue in aggregate data: Averaging performance over 30 days masks the reality that most creatives decay significantly after 2-3 weeks. A creative that looked phenomenal in week 1 but fatigued by week 3 will appear mediocre in a monthly report. Always segment by time period.
Asset stuffing all creatives into one ad set: Without thematic separation, the algorithm picks one winner fast and starves the rest. Your data for 'losing' creatives in that set is essentially meaningless because they never got a fair chance with the right audience. Separate by concept or audience theme.
Confusing correlation with causation in pattern analysis: Just because your top 5 creatives all use blue backgrounds doesn't mean blue backgrounds drive performance. Look for patterns that have logical, psychological explanations (emotional angle, messaging clarity, audience targeting) rather than superficial visual coincidences.

Analyzing ad creative performance data is both a science and a craft. Start by defining your metric hierarchy and building a clean, merged report. Then diagnose where each creative breaks down in the funnel, identify patterns across winners using structural and emotional tags, and ensure your testing structure produces trustworthy data. Most importantly, build a system that compounds your learnings over time. At RocketShip HQ, we've found that teams who invest in rigorous creative analysis consistently outperform those who simply produce more volume. The data is there. Your job is to turn it into your next winning creative brief.

Looking to scale your mobile app growth with performance creative that delivers results? Talk to RocketShip HQ to learn how our frameworks can work for your app.

Not ready yet? Get strategies and tips from the leading edge of mobile growth in a generative AI world: subscribe to our newsletter.

Shamanth

Shamanth Rao is the founder of RocketShip HQ, a performance creative and growth marketing agency helping mobile apps scale through ad creatives, experimentation, and data-driven marketing systems. With over a decade in mobile user acquisition, he has managed growth for apps with hundreds of millions of installs across Meta, TikTok, Apple Search Ads, and Google. He hosts the Mobile User Acquisition Show podcast and has spoken at MAU, Pocket Gamer Connects, and App Promotion Summit.