The first pass to produce a videogram grid/card is done entirely via the Videogram Video Indexer and Machine Learning. The videogram is then embedded in website and/or apps. The platform starts tracking user behavior, tracks frames that are getting clicked, shared, commented. Based on this tracking, algorithms can identify which parts of the video are getting traction. It also tracks objects and speech of those frames that are important to the consumer. This data is used to draw inference on consumer content preference. Machine learning is used to understand similarity of content preference between one to many. The output of this data collection is used to provide user content recommendation & personalization.