An Algorithm that Predicts the Impact of Video Ads on Sales

Return on ad spend drives most advertisement decisions and is difficult to predict prior to launching an ad. Previously, content creators relied on engagement metrics (e.g., number of likes, shares, and comments) to gauge ad performance, however, the data remains unclear. The present algorithm aims to address these limitations by accurately predicting the effectiveness of a video ad on the sales of the showcased product. To assess effectiveness, a product engagement metric is determined by evaluating pixel-level engagement across the pixels where a product is presented. The inventors propose the concept of a motion score, or M-Score, to determine the extent to which the product is advertised during the most engaging parts of the video. According to the inventors, a one standard deviation increase in M-Score is associated with a 12% increase in the sales revenue. As a result of this technology, advertisers can more accurately predict the success of an ad prior to its launch and generate targeted sales.