Recommendation systems have become integral to our daily lives, with apps suggesting what we might want to watch, eat, read, or invest our time and energy in. The average American consumer uses daily at least 3 apps involving some sort of recommendation system (Medium, 2021) and spend around $200 monthly on subscription systems (Yahoo Finance, 2024), 60% of YouTube searches come from recommendations and 40% of the apps present on Google Play offer in their services a recommendation system (GoogleDevelopers, 2024). However, 75% of users complain that these apps do not actually reflect their taste (The New York Post, 2024), rather leading them to be overwhelmed by the sheer volume of products offered (CNET, 2024). In particular, in online media markets, streaming platforms like Netflix rely on recommendation systems for 75% of its revenue (Medium, 2019) while companies like Amazon attribute approximately 35% of their e-commerce revenue to product recommendations (Quartz, 2018) and yet viewers lament a complete inadequacy of their algorithms (HBR, 2024). How is this misalignment becoming possible and how do we fix it? In our research, we hypothesize that integrating the current algorithms with consumers’ time and energy availability estimations might improve recommendation systems, increasing both consumers’ satisfaction levels and companies’ revenues.
Specifically, we propose a model where effort and time estimation can improve both the choice of content users select to watch – and their satisfaction with it – and the recommendation systems output. By factoring in the effort and time a user is willing to invest, the recommendation system can tailor content that aligns more closely with the user’s current state, thereby improving both the selection process and overall satisfaction.…Read More