At Pencil, we use advanced machine-learning models to provide predictions for your ads. This guide will walk you through the process in simple terms, so you can understand how our predictions are calculated.
Pencil Score
The Pencil score evaluates your ad creatives by comparing them to our model, which is trained on a dataset of over $1 billion in ad spend.
To calculate the score, your creatives are divided into smaller components (images, texts, contrasts between layers, and other features). These components are then compared to similar fragments in our dataset using our model, which generates a score.
This score is a percentile ranking within your sector. For example, a score of 91 means your ad is predicted to perform better than 91% of all ads for the selected metric, placing it in the top 9% of performers.
If you don't have ad accounts connected in your workspace, your score will be based on our common pool dataset for your sector. However, if you have connected your Meta account, your Pencil score will be heavily weighted to your own ad performance over the past 12-52 weeks.
Explainability
To help you understand your ad’s score, we use advanced techniques to break down the contribution of fragments of your ad (like text visibility or image clutter) to the Pencil score.
It's important to note that the explanations we provide are meant to illustrate what factors contribute to your current score on average. Applying these may not necessarily change the overall score, as they are simply insights into what is considered.
The explanations are sorted according to what contributes the most to the current score.
FAQ
Is the score going to change over time?
At Pencil, we commit to constantly improving our predictions and explainability models over time.
This will result in changes over time, as both our models and dataset grow.
Why does the model generate a score on incomplete creatives?
Pencil Score will always provide a value as long as the canvas has content on it.
The score will not work reliably if the content of the canvas is of really bad quality or incomplete, but it will keep providing value.
The closer your ad is to being finished, the more accurate our score will be.