Algorithm and Art: Netflix's Approach



It is no secret that Netflix uses algorithms to keep you hooked. Something that has always been interesting to me is their use of data to decide on movie cover/banner art. They have built a system that tests a set of images on their catalogue and chooses the most compelling (engagement driving) one. Through such tests, Netflix has determined that seeing a certain range of emotions actually compels people to watch a TV show or movie more. 

An example of this is seen in the recent winning image for the second season of Unbreakable Kimmy Schmidt (example from https://uxplanet.org/netflix-binging-on-the-algorithm-a3a74a6c1f59). I feel like without knowing the results, I would decide on the bottom right anyway, but this use of algorithm is a superior way of cutting out human involvement or artist bias in the selection process. 

Is artistic involvement and human intervention always a good thing to minimize? Certainly, for Netflix, it means one less person to pay. It does not actually matter that the 'best looking' banner is picked; audiences may be naturally driven to an 'artistically unsound' image. And does that not, in turn, say something about that art?

Netflix uses this same strategy to target individual users as well. Check out the images below: all different variations of the Stranger Things banner. What you see if impacted by what you have seen. For example, if you are a viewer who recently watched Ghostbusters, or is from the Ghostbusters era (in age), you may be more likely to find the 6th banner on your Netflix feed. 

I would be curious to see what associated behaviors result in the different variations of the 'Stranger Things' banner. 

Lastly, if Netflix uses this small yet specific engagement tool fueled by algorithms to increase engagement, I would certainly speculate that algorithms may play a role in the formation and popularity of some of Netflix's most recent original shows. Although an entertainment venue, Netflix has more data than most. It will be interesting to see what new directions the company will take with this method, and what future art their collected data will fuel. 

-Madeline

Comments