A Large-Scale Characterization of User Behaviour in Cable TV

Transcription

1 A Large-Scale Characterization of User Behaviour in Cable TV Diogo Gonçalves, Miguel Costa, Francisco Couto Faculty of Sciences, University of Lisbon RecSysTV 2016, Boston, USA September 15, 2016

2 Today, there are many services from which to choose contents to watch: Live TV Video on Demand (VOD) Catch-up TV Over-the-top (OTT) from 3 rd parties (e.g. Netflix) Understanding how users interact with such services is important to increase: user satisfaction user engagement user consumption (this knowledge helps to enhance the recommendation systems of Cable TV providers) We didn t find any study comparing usage patterns between Live TV, VOD and Catch-up TV in an integrated way from a large scale Cable TV operator. 2

3 Live-TV a client can watch any video content that is being broadcast live (e.g. a live soccer game). VOD (video-on-demand) a client can watch any video content anytime that was pre-recorded and made available, usually a movie or series. Catch-up TV is a type of VOD, where a client can watch any video content that was broadcast live up to a few days before (e.g. up to 7 days). Contents are delivered via Set-Top Boxes (STB) installed in users homes. 3

4 Live Catch-up VOD Total Users 896, , , ,000 Programs 24,000 24,000 15,000 39,000 Episodes 330, ,000 15, ,000 Programs per moment (avg.) 160 6,000 15,000 21,000 Episodes per moment (avg.) ,000 15,000 50,000 Views (>10 min) 617,000,000 56,000,000 9,000, ,000,000 User views (avg.) User views per month (avg.) October to December 2015 (9 weeks) 160 channels available to the user 4

5 Some Results 5

6 # users (thousands) Number of users per day Live TV has much more users than Catch-up and VOD together, but less research Catch-up TV Live TV VOD Days Should we focus more in improving the user experience for Live TV? 6

7 # views (millions) # hours (millions) Number of views per day Number of watched hours per day Catch-up TV Live TV VOD Days Catch-up TV Live TV VOD Days Live TV has much more views and watched hours than VOD & Catch-up 7

8 # views (millions) 1.6 Number of views per hour sleeping working The graph shows the typical work/rest cycle of users for the 3 TV services. Live TV Catch-up TV VOD hour of day 8

9 # views 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% stop watching sooner The average percentage of content watched in Live TV is small (27%) when compared with Catch-up TV (50%) and VOD (60%) most people do not watch the full content % program watched Live TV Catch-up TV VOD 9

10 # views 100% Live TV 80% 60% kids category was more watched 40% 20% 0% adults category was less watched % program watched News TV Series Entertainment Kids Documentaries Sports Movies Adults 10

11 # programs # programs 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% Distribution of program types per day made available by the Cable TV operator Recurring Programs (Repeats) Recurring Programs (New Episodes) New Programs most users watch new episodes of programs already seen Days 100% 80% 60% 40% 20% ~20% of watched episodes are repeated. These are mostly kids programs. Distribution of program types watched by users on live and catch-up TV 0% Recurring Programs, Recurring for User, Repeat Recurring Programs, Recurring for User, New Episode Days Recurring Programs, New for user not having feedback New programs from the specific user11

12 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% # views Views in Catch-up TV per category over the day users watch different categories in different hours hour of day News TV Series Entertainment Kids Documentaries Sports 12

13 # views (millions) 1.8 Views in Catch-up TV hours after broadcast most users watch contents from the last hour 13

14 The characterization of Live TV, Catch-up TV and VOD on a large-scale Cable TV provider shows usage differences between these 3 services. These insights enable to adapt and create better recommendation algorithms depending on the TV service. Live TV receives the majority of views. Still, Catch-up TV and VOD accumulate a large amount of views and hours watched. The 3 services should have recommendations. Users tend to watch Catch-up TV and VOD programs for longer, when compared to Live TV. The implicit feedback provided by the program views should be adjusted to the TV service. Users tend to watch some categories (e.g. Kids) for longer than others (e.g. Adults). The implicit feedback provided by the program views should be adjusted to the content categories. Users prefer to watch different types of programs depending on the hour of day. We should adjust the recommendations for the time period. Users prefer to watch new episodes of programs previously watched by them. Should we recommend the programs that users usually watch or should we suggest new programs? Users prefer to watch programs recently broadcast in Catch-up TV. Should we recommend the most recent contents first? 14

15 Thank you.