Video First Experience for Investigation Discovery
| Objective
Increase total minutes viewed for Investigation Discovery’s apps.
| Team Leads
| Insights
25 – 40 year old female audience that primarily watches alone.
Total minutes viewed on mobile apps exceeded other Discovery brands with comparable audiences but were less on connected TV apps.
Unauthenticated visitors are less likely to convert to viewers than authenticated visitors.
In interviews, Investigation Discovery viewers could not recall the names of shows they watched or any of the actors from the shows. Their orientation is to the network brand and its crime genre.
Most ID connected TV app visitors browsed multiple pages looking for something to watch, then either selected a video from the home page carousel or exited without watching.
The viewers that watched the streaming linear channel had longer sessions than video on demand viewers.
From previous experiments auto-playing video on demand when the page loaded caused most viewers to exit before the video started.
| Hypothesis
Automatic selection and continuous playback of relevant videos for each viewer will reduce the level of effort for them to start watching and keep watching on connected TV’s increasing the total minutes viewed per month.
| User Experience
The Video First experience was co-developed with a customer panel that volunteered for a series of interviews to provide feedback on connected TV wireframes, prototypes, and apps. Remote usability testing sessions were used to validate the panel’s feedback.
The key features of the Video First experience were automatic playback of a personalized list of videos through a player that made it easy to keep watching or find something else to watch without exiting the video player.
Several seconds after opening the app, transitional animations eased the visitor from the home page into watching a video selected just for them by machine learning algorithms. At the end of each video, the viewer was automatically served the next one from their personalized playlist, enabling them to sit back and enjoy “me time.”
To ensure the automated experience did not take control away from the viewer, it was easily stopped by hitting any button on the remote control. With the autoplay stopped the viewer was free to browse and search for something to watch. Manually selecting a video to watch would restart the continuous playback with a new personalized content list.
| Engineering
The simplicity of content in the narrow crime genre enabled an off the shelf recommendations engine to be purchased to speed time to market. The engine received a content catalog feed. It also received impression, scroll, click and playhead consumer behavior events in near real time from a data pipeline. This data was used to iteratively optimize the recommended content for each viewer.
The UX Manager tool provided configurations for all the pages in the apps to a UI Layout service. The layout service was also responsible for fetching recommended content from the engine. It served JSON responses with view and content data to the front end to render the UI’s.
The front end was based on existing hosted web app (HWA) codebases used for XBOX, Samsung CTV’s and websites. Core components, like the shell, video player, analytics and focus manager, were reused to reduce the level of effort. The majority of the development effort focused on styling changes with new interaction designs and UI routes.
| Technologies
React / Node.js
Machine Learning
PHP / MongoDB / REST
Angular / mySQL
| Experimentation
A completely re-imagined experience was made available to 5% of the Samsung connected TV’s and XBOX consumer base starting in January 2019.
The control for testing the experience’s performance was the existing in-market user experience that offered the same content but was heavily curated by editorial staff into a carousel and grid design pattern.
After confirming that the new experience increased engagement metrics, it was made available to 20% of the base to generate sufficient traffic for A/B testing. The team employed ICE (Impact + Confidence + Ease) to prioritize the implementation of solutions that would further increase total minutes. Each test that increased engagement provided the team confidence to increase the audience being served the variant to reduce the time it took to get statistically valid results.
The variants tested through April 2019 included:
◦ The number and type of options as well as the icons in the global navigation.
◦ The number and duration of transitional steps into autoplay and continuous playback.
◦ Locked content presented to unauthenticated visitors.
◦ Location of login prompt in visitor’s journey and its messaging.
◦ Manually curated videos versus machine only.
| Retrospective
◦ Prioritizing the presentation of unlocked on demand videos and locked live channel to unauthenticated visitors generated the highest minutes viewed.
◦ The login prompt increased the authentication rate best as the first step in the journey.
◦ A three-step transition that took 7 seconds from open app to auto playing a video had the lowest abandonment rate.
◦ A global navigation menu with 4 options converted the most visitors to viewers.
◦ Demoing to fans at IDCon was rewarding for the team. Getting their qualitative feedback uncovered the polish necessary to achieve the engagement goal.
| Key Results
More Logins
◦ 19% more visitors authenticated
Higher Streaming Starts
◦ 26% more visitors converted to viewers
Longer Sessions
◦ 31% more minutes watched per viewer
Happier Customers
◦ “I don’t think I’ll ever watch live TV again.”
◦ “The new app has helped me discover new content. Starting with one show I was interested in, snowballing into another and another.”
◦ “Seeing the changes to the app has me watching ID even more. I’m an ID fan for life now.”