How Netflix is Using Data to Improve Quality of Experience

We recently reported on how Netflix is using big data algorithms to power its recommendation system; now, a Netflix blog post highlights other ways in which the company is using big data behind the scenes to enrich its service. They’re putting the data to work largely to improve Quality of Experience (QoE)- the experience the user has after they hit play.

The blog post, written by Director of Streaming Science & Algorithms Nirmal Govind, walks us through several ways in which data is enhancing the quality of experience. The first approach outlined is looking at “the algorithms that run in real-time or near real-time once playback has started, which determine what bitrate should be served, what server to download that content from, etc.”

“With vast amounts of data, the mapping function discussed above can be used to further improve the experience for our members at the aggregate level, and even personalize the streaming experience based on what the function might look like based on each member’s ‘QoE preference'”, he continues. “Personalization can also be based on a member’s network characteristics, device, location, etc.” For instance, a member using a computer with a high-bandwidth connection will probably have very different quality expectations to someone watching on their mobile using a low-bandwidth network. They’re also looking into algorithms to optimise content location so it’s as close to end user (in terms of network hops) as possible.

Although in our last report we mentioned Netflix weren’t going to use their data when producing the content, but there is one way in which content is affected by data- in the field of subtitles, audio and closed captions. Given the amount of countries Netflix is now available in, subtitles and captions can have a huge impact on user’s opinion of the quality of the service. The first port of call when checking the quality of captions is user feedback- but relevant feedback is mired in an ocean of feedback that’s not related to content (e.g. network issues), non-issues, or comments that only relate to a user’s tastes and preferences. As Govind states, “identifying issues that are truly content quality related amounts to finding the proverbial needle in a haystack”.

Netflix’s solution is a model which predicts which content might have quality issues. The model detects patterns such a sharp drop-off points at a particular time in a show, and then couples this with feedback on that show to identify problems. They’re also using natural language processing and text mining techniques to improve the quality of captions and subtitles before they go live- a tool which will prove vital as Netflix continues to expand internationally.

The recommendation system is certainly the most well-known of Netflix’s data-based applications. But getting the user to hit play is only half the battle; if the quality doesn’t meet expectations in terms of speed, image quality and the accuracy of captions, users are going to click off. The techniques outlined above are trying to ensure that data improves the user experience right up until the credits roll.