Identify Your REAL Winners
Not the ones with the most spendβthe candidates that meet your pre-defined evidence and unit-economics criteria.
High spend and platform-attributed revenue do not establish profit or causality. Define a candidate using contribution margin, returns/cancellations, conversion lag, sample adequacy, and an incrementality test where the decision warrants it.
Spend or attributed revenue alone is incomplete. A 2x platform ROAS may be profitable for one business and loss-making for another; calculate the break-even threshold from contribution margin, fulfillment, returns, discounts, and incremental lift.
A real winner is identified by efficiency metrics combined:
| Metric | What It Tells You | Decision Criterion |
|---|---|---|
| CTR | Observed click response, not customer quality | Compare with a concurrent control and uncertainty |
| ROAS | Platform-attributed revenue per ad spend | Above your calculated break-even level; verify incrementality |
| CPA | How much does each customer cost? | Below your target CPA |
| Frequency | Exposure context, not fatigue proof | Interpret with time-series response and audience size |
| Hook Rate (video) | Early-view behavior under the platform definition | Compare like-for-like placements and objectives |
A decision candidate needs an adequate sample, completed outcome lag, a pre-defined primary metric, contribution economics, andβwhere materialβincremental evidence. CTR, ROAS, CPA, and frequency are context, not a universal winner formula.
In the synthetic example, high CTR coexists with weak downstream performance. Do not infer that the hook caused a customer type or return behavior. Investigate funnel quality and compare profit-qualified outcomes with adequate lag.
Break Down WHY It Won
The difference between looking at an ad and ANALYZING an ad.
When a creative works, most brands say "great, make more like that" β but nobody knows what "like that" actually means. This step gives you a precise breakdown so you can replicate success intentionally, not accidentally.
From the video: "We look at which creatives succeeded and what they were able to achieve, and we start to do a reverse-engineer breakdown. We figure out the reasons why it succeeded."
Most people look at a winning ad and think "nice video." A systematic breakdown asks 9 specific questions:
In Part 5, we'll show you how to use AI Vision (Gemini, Claude) to do this breakdown AUTOMATICALLY β you feed it a creative, it answers all 9 questions in seconds. But understanding the questions manually first is essential.
Extract the Repeatable DNA
Not "make another video like that" β specific elements you can replicate.
The DNA is a recipe. Instead of hoping your creative team "gets the vibe," you give them a specific formula: "Use THIS hook type + THIS visual style + THIS emotional trigger." Repeatable. Scalable. No more guessing.
After breaking down a winner, you extract the specific combination of elements that made it work. This is the DNA β the recipe you can replicate across different products, angles, and formats.
This DNA card becomes your creative brief. When you tell your team "make more like Video A," you're no longer being vague β you're handing them a specific formula with every element defined.
Rebuild Better
Use the DNA to craft creatives that are significantly better than the originals.
This turns an observed pattern into a variation system. The variations are new hypothesesβnot guaranteed good adsβand should enter the same controlled testing and human-review process.
From the video: "We figure out the reasons why it succeeded to give you the capability to craft creatives that are significantly better than the ones you made initially, achieving vastly more sales."
With the DNA extracted, you can now:
Each adequately tested creative can add context to your DNA library. Track whether reusing a pattern improves your prospectively defined win rate; do not assume a timeline or a 5-in-10 result.
Reverse-Engineer Your Competitors Too
You don't just learn from YOUR winners. You learn from everyone's.
Public ad libraries reveal visible executions and run dates, not spend, profit, targeting, attribution, or test results. Use them to generate category and differentiation hypothesesβnot to label competitor ads as winners.
The same 4-step framework applies to competitor creatives:
- Find relevant active and historical ads in public libraries. Longevity can reflect many factors and is not evidence of profitability.
- Break down the 9 elements β what hook, visual style, copy angle are they using?
- Extract the DNA β what's the pattern across their top 5 long-running creatives?
- Differentiate and rebuild β don't copy. Take the DNA and apply it with YOUR brand voice, YOUR product, YOUR unique positioning
The goal is not to copy competitor creatives or assume their patterns work. Use public material to form a differentiated, brand-safe hypothesis, clear rights and claims, then validate it with your own authorized data.
Built by @itsmazinzaki β AVAMARTECH