Zappi Ad Predictor

Thanks for your support, PH! Roughly a year ago we launched V1 of the Digital Ad Predictor, with support for video ads only. You gave us feedback, and we've acted on it - now, test images, with a great new integration from EyeQuant that predicts exactly how well you'll capture your audiences' attention. Please, let us know - is this useful? What do you want to see next? Thanks! 😸🔮

Hey PH, Alex here - we’re excited to be back! As a quick recap, our Digital Ad Predictor gives you a free performance prediction report for your digital ads BEFORE you go live with them. Simply upload your ad, choose your ad’s category and get your report within minutes. Since our V1 launch, our biggest feature request was to add support for images (in addition to videos). And that’s exactly what we’re launching today! V2 lets you run the predictor on both your static AND video ads. Static ad reports come with a free visual attention heatmap, powered by EyeQuant - it uses some awesome (and genuinely groundbreaking) A.I. to predict what your audience will actually notice and see! Here’s a quick feature summary - Get a free strength & weakness report for your ad (static & video) - Get an industry benchmark report for your ad (static & video) - Get a free visual attention heatmap for your ad (static only) This insight is designed to be actionable, and help you rapidly improve your ad BEFORE putting it in front of your audience. Here’s an explanation of how it works: The Scores: Zappi uses a gradient-boosting machine learning algorithm to rapidly deliver a free prediction, based on a huge volume of anonymised and aggregated survey and norms data. What this means in layman’s terms is that a load of tags — both manually added and assigned algorithmically — are collected, cleaned, and fed into a decision tree model that then “boost” important components, while suppressing less important ones. This is the XGboost regression model. We’re using a machine learning model that predicts key metrics based on the attributes of the tested ad, e.g. branding, celebrity/human presence, length, etc. The model is trained on ads tested on Zappi Video/Digital and the prediction falls on average within 5% of the observed survey scores. This kind of tech traditionally lives in the Enterprise space — we’re excited to be making it available to everyone! Feel free to pop any questions in the comments 🚀

Love this. Very excited to try it out and take some of the guesswork out of ad design.