The Technologies Used for Artwork Personalization and the Challenges

For many years, many streaming companies’ personalized recommendation systems mainly focused on how to employ the right algorithm to predict what the subscriber would be interested in based on their viewing history and preferences. By showing those content, companies believe this could provide efficiency to their subscribers and thus their content could reach better performance. However, since it is impossible to present all the details of that content on the homepage, the title shown is unable to contain enough information to trigger the user to click on that. Instead, the artwork which represents the content plays a significant role in the number of clicks the specific content could receive. Although few companies already realized this unprecedented aspect of the personalized recommendation system and started to work on the development of a certain algorithm using A/B testing and contextual bandits approach to improve the system, there are limited research methods that have been employed and they are still facing many challenges. By examining how the A/B testing and contextual bandits approach actually work and the logic behind these two basic research tools and at the same time dealing with those challenges, companies could come up with more comprehensive research designs to better study the users’ reactions and inclination when provided with different artworks. Thus, fully understanding where those challenges lay could help companies to be sure about the future development direction of the personalized recommendation field.