
At Superads our goal is to make ad creative analysis easier, faster, better.. That’s why we're introducing Superads Scores this week: a simple percentile-based metric that tells you how your ad performance is ranked in your own account. It sounds simple but, under the hood, this simplicity is powered by quite a bit of data science.
Let’s walk you through what’s happening behind the scenes and why we chose this approach.
Treating ads as stochastic random variables
Every ad metric (CTR, engagement rates, thumbstop ratio, hold time, and conversion rate) is treated as a stochastic random variable. This lets us account for the natural variability in performance between ads.
Our research is based on analyzing data from over 15,000 ad accounts and $2.2 billion in ad spend. By treating every ad metric as stochastic variables, we’re able to model the natural variability in ad performance and continuously refine our understanding of what great looks like. As more data flows in, the percentile estimates get sharper, and our system becomes even better at identifying standout creative.
The problem with averages
If you’ve ever looked at ad performance across campaigns, you know how unpredictable it can be. Some ads flop. Some do okay. And once in a while, one just explodes. Getting clicks, engagement, or conversions far beyond the rest. Every brand has that 10x+ outlier ad.
This creates what statisticians call a skewed distribution with fat longtail. In plain terms: most ads cluster toward the lower end of performance, and a few “super-performers” stretch the data way out to the right. It’s what’s known as a long-tail.
In this kind of distribution, averages become misleading. A single outlier can pull the average up so much that your “pretty good” ad suddenly looks underwhelming against the average. Even if it’s outperforming most of the others.
Percentiles, not pitfalls
To avoid reporting misleading scores due to the skewed distribution problem, we built Superads Scores using percentiles.
This means we’re not asking, “How much better or worse is this ad than the average?”.
We’re asking, “How many other ads is this ad outperforming?” (and "specifically at what?)
Concrete examples:
- A Hook Score in the 60th percentile (score = 60) means your hook is better than 60% of ads in the dataset.
- If it’s in the 90th (score = 90), it’s truly top-tier.
This makes the score more stable, less sensitive to outliers, and easier to interpret, especially when you’re comparing performance across multiple ad accounts.
Why green starts at 50%
In a skewed distribution with a fat long tail, the average can be heavily inflated by a few top-performing outliers. That’s why the 50th percentile is a much more honest baseline. it shows your ad is outperforming half the dataset, even if the average sits artificially high.
That’s why we color the score green at and above 50% to reflect where real impact starts to show up.
The scores we’re starting with
In this initial release, we’re calculating Superads Scores for five key creative performance signals:
- Click Score – reflects the strength of click interactions of your ad. Main metric: CTR.
- Engagement Score – reflects the strength of engagement interactions of your ad. We're combining multiple metrics like Post reactions, comments, shares, etc.
- Hook Score – reflects the strength of the first seconds of your ad (only available for video). Main metric: Hook rate.
- Hold Score – reflects the strength of your ad at holding attention over longer periods of time (only available for video). Main metric: Hold rate.
- Conversion Score – reflects the ability of your ad driving Purchases. Main metric: Purchases.
More score types will come as we deepen our models and integrate new data sources.
What this means for you
With Superads Scores, we’re giving you a fast, intuitive way to spot winners and identify creative opportunities, without needing to sift through spreadsheets or worry about whether you’re interpreting the data correctly. Every one can look at the same reports, within knowing the specifics of a certain account, and understand if something is performing well or not, compared to your other ads.





