To analyze ad creative performance data, read each creative down the funnel stage by stage (hook, hold, click, post-install conversion) to localize exactly where it breaks down, weight every change by impact rather than raw percentage so high-spend creatives surface above low-spend noise, and turn each diagnosis into a specific decision about what to test next. Creative analysis is not a scoreboard. It is a diagnostic loop: find the broken stage, form a hypothesis, brief the next round.
Most growth teams drown in dashboards without extracting a single decision. The fix is not more metrics. It is a method that tells you which stage of the funnel failed and what that failure means for production. Below is the approach we use at RocketShip HQ.
For the broader strategic context, see our mobile ad creative strategy guide.
Page Contents
Why read the creative funnel stage by stage?
A creative is rarely bad at everything. It usually fails at one specific stage, and that stage tells you what to fix. Reading the funnel in order localizes the problem instead of leaving you with a vague “this one underperformed.”
Walk each creative through the stages in sequence:
- Hook (the first seconds): Did people stop scrolling and watch past the opening? Weakness here is a visual and opening-line problem, not a messaging problem. The fix lives in the first frame, the pattern interrupt, and the opening overlay.
- Hold (completion / watch-through): Did the people you hooked keep watching? When the hook is strong but hold is weak, you grabbed attention and then lost it. The body of the ad is not paying off the promise of the hook. Look for pacing, an unclear value proposition, or a tonal mismatch between the hook and the rest.
- Click (click-through): Did watchers act? Strong watch-through but weak click-through usually means the ad engaged people without giving them a reason to move, or the call to action and end card were soft.
- Post-install conversion: Did the users who installed actually do what you wanted (subscribe, purchase, retain)? This is the most expensive failure because you paid to acquire people who do not monetize. The creative may be setting the wrong expectation or attracting the wrong intent. That points back to the messaging angle, not the edit.
Reading in this order matters because an early-stage break contaminates everything downstream. A creative with a weak hook never gets a fair read on its post-install behavior, because too few of the right people made it through.
How do you weight changes by impact instead of raw percentage?
The trap in any creative report is reacting to the biggest percentage swing. A dramatic-looking move on a creative that barely spent is usually noise, while a smaller move on a creative carrying real budget is what actually moves the account.
We score anomalies by weighting the percentage change by the square root of spend:
weighted anomaly score = abs(% change) x sqrt(spend)
The square root keeps very high spenders from swamping everything else while still pushing meaningful budget to the top of the list. Sort your creatives by this score and the changes worth your attention rise to the top on their own. A swing on a creative with negligible spend drops down where it belongs, and a steady shift on a high-spend workhorse surfaces even when its raw percentage looks unremarkable.
This is the difference between a report that reacts to whatever number is reddest and one that directs attention to what is actually changing the business.
How do you build a report that supports this analysis?
The goal is one row per creative with the full funnel side by side, so you can read a single creative top to bottom and compare across the set.
- Pull platform engagement signals: from your ad platform, isolate individual creatives (not ad-level aggregates) and capture the early-funnel engagement: scroll-stop, watch-through, and clicks.
- Merge post-install events: join your measurement data (installs and the downstream events you actually care about) to the platform export using the creative or ad ID as the key. Without this, you are optimizing top-of-funnel engagement blind to whether it produces value.
- Express each stage as a transition, not a raw count: what share of impressions became engaged views, what share of those held, what share clicked, what share converted. Stage-to-stage transitions are what let you localize the break.
- Add the weighted anomaly score as a column so the report sorts itself toward what matters.
Automate the refresh. Time spent rebuilding the same export by hand is time not spent on the analysis that actually changes your next brief.
How does the analysis decide what to test next?
Diagnosis is only useful if it changes production. Map each funnel break to the specific lever that fixes it, and let that drive the next brief:
- Hook breaks point to the opening. Test new visual pattern interrupts, opening lines, and first-frame overlays.
- Hold breaks point to the body. Test pacing, structure, and whether the middle delivers on the hook’s promise.
- Click breaks point to the offer and call to action. Test the end card and the clarity of the next step.
- Post-install breaks point to the angle. Rework the messaging and the expectation it sets, not the production.
Then look across your winners for patterns, not coincidences. Tag creatives by concept, format, hook type, emotional angle, and visual style, and group performance by those tags. The goal is to find structural and emotional patterns that have a logical explanation (a clearer angle, a sharper hook, a better-matched audience) rather than a superficial coincidence like “the winners were all blue.”
Two craft frameworks anchor what we look for in the hooks themselves. The 3C Principle asks whether a hook has Context (who is this for), Clarity (what is this about), and Curiosity (an open loop worth resolving). The 4-Layer Hook System asks whether the opening stacks a visual interrupt, an orienting text layer, a verbal layer, and audio that amplifies the emotion. Hooks missing a layer or a C are the first place to look when scroll-stop is weak.
Feed these decisions into a documented plan rather than a one-off reaction. Our creative testing roadmap covers how to turn each cycle’s findings into the next round of tests so learnings compound instead of resetting every week.
Frequently asked questions
What is the single most useful creative analysis technique?
Reading each creative down the funnel in order. Identifying the specific stage where a creative breaks down (hook, hold, click, or post-install conversion) tells you exactly what to fix, which is far more actionable than a single overall verdict of good or bad.
Why weight changes by spend instead of looking at the biggest percentage moves?
Because the biggest percentage moves usually come from creatives with too little spend to mean anything. Weighting by the square root of spend (abs(% change) x sqrt(spend)) keeps your attention on changes large enough, on enough budget, to actually move the account.
How do I separate a winning concept from a winning execution?
Group creatives by emotional angle first, then compare executions within each angle. The spread between the best and worst execution inside one angle tells you how much production quality matters for that concept. Some angles are execution-sensitive, others are concept-sensitive, and you brief them differently.
How do I make sure my analysis compounds?
Turn each cycle into a documented decision, not a one-off reaction. Record what worked, what broke, and one hypothesis for the next round, and feed it into a testing roadmap so each batch builds on the last instead of starting over.
Methodology note
This guide describes the qualitative method we use at RocketShip HQ across mobile creative accounts. It deliberately contains no benchmark numbers (no target hook rates, completion rates, click-through rates, or spend thresholds), because those vary widely by app, category, platform, and objective, and any single figure would mislead more than it helps. The one formula here, the weighted anomaly score (abs(% change) x sqrt(spend)), is a prioritization method, not a benchmark: it ranks where to look, and the right thresholds for your account come from your own data. Read the funnel, weight by impact, and let your account set the bar.
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.
