Introduction

Movielens is a platform that allows users to get non-commercial, personilized movie recommendations. The platform has a huge database of movies that users can browse through, review the community-applied tags, or apply their own tags. One aspect of the platform is that, it offers it’s users to tune the matching algorithm so the results are more personalized. It is completely free and does not require any subscription. Movielens is part of Grouplens which is an organization part of Social Computing Research at the University of Minnesota.

Scenario Design

Target Users

My assumption in terms of the target users for Movielens is consumers ages between 18-35, college students and possibly web and digital savy individuals. Consumers are within different ethnic groups but none of them are international (outside of United States). In terms of type of users, they tend to subscribe netflix, hulu and possible other streaming services, that love watching movies and dont want to waste their time searching of movies prior to renting or streaming it.

Business Goals

High level business objective is to help users find the movies to watch and collect user data and insight information along the way. These data can further be used on research articles in the field of Computer Science, psycology, sociology and medicine by Grouplens. The main objective of the product is to provide personalized movie recommendations with an highly accurate prediction algorithm. If the movielens product is build based on target user requirements, providing accurate movie recommendation to the users, more consumers will sign up and this will result with more data collection for Grouplens for further research.

Evaluation

Movielens has an easy to navigate, simple interface, however it lacts of enhance user experience design. For example, margins between the content components, text and image placement are not accurate accross the site which creates bad UX. As expected, there are no adverstising content that antices user engagement to the content and further sign up through out the site.

Overall, it is a functional website where no intention on user acqusition (such as email sign up, newsletter and etc…), engagement (such as extensive explanatory content) or conversion (multiple sign up forms)

To be able add custom tagging to the movies and being able to select commercialized tags for personalized movie recommendation is a great feature however the content that goes with this feature does not make sense in certain places.

Current Model

Based on my research, currently movielens uses multiple different algoritms that predicts the recommend movie for the user based on the filtering and selection of the user. The Algorithm leverages collaborative filtering where it collects many users movie feature preference and maps the same opinions as a recommendation to particular user. It also uses item-item collaborative filtering which recommends based on similar movies against the users ratings. It also uses Singular Value Decomposition.

Currently the algorithm defines a target user based on their input in terms of star ratings and tags for the movies. Further maps these preferences with other users. Algorithm then takes the movies that the target user never provided input on, and creates a prediction of star rating and tags. Then, algorithm provides the recommended movie accordingly.

Recommendation

The recommendation engine requires user input and sometimes this can become too much work for the user. For example; in order for the algorithm to provide accurate recommendation, the user needs to provide star rating, custom or community tags for the movie. If there was a possibility to partner with streaming network provider and tie user accounts together, it might improve the user experience, as the algorithm would also know what user watched in the past.

Another improvement would be to display all the community tags for each movie for the user to select rather than providing what the algorithm thinks it is right community tag for that particular movie. This would improve the filtering process.

Overall website user experience and content (such as links to imdb, adding more variables besides Genres) would help improve the user engagement and movie recommendation.