Online video enters age of Big Data, Predictive Analytics and Machine Learning
As we close the doors on 2016, we are entering into most fascinating times in the history of online video. Over the top (OTT) video, which started as a complementary service to Television has unfolded into a multi-billion dollar SVOD industry, and shows no signs of slowing down. Netflix, Hulu and Amazon have emerged as new age entertainment giants, and Virtual MVPD providers are re-wiring online to bridge any gaps from traditional TV. It is thrilling to witness new events reshaping OTT services each day. Sample this, launch of DirecTVNow; cloud DVR on Sling TV, offline viewing from Netflix and live NFL games on CBS all access, all of these have happened in a period of less than last two weeks. As digital media shifts online, broadcasters and online video distributors are innovating faster than ever. First came the content, then came technology, next was original content production and now it’s the turn of big data, predictive analytics and machine learning to redefine online video.
Big Data Collection
Lack of quality data is the biggest challenge in data analytics, and historically broadcasters had little access to it. Cable and Satellite service providers controlled consumer data, and broadcasters view was best limited to Nielsen ratings from sampled audience data. But it all started to change almost a decade back with launch of VOD streaming services like YouTube, Netflix and others. Broadcasters, fearing threat of being left behind, introduced TV Everywhere services on their part, offering limited content over the top to authenticated subscribers.
Moving content online brought benefits for both subscribers and content providers. While consumers got convenience, content providers got new revenue streams and data to consumer behavior, real insights into how subscribers interact with their content. This enabled providers to capture and store every possible consumer interaction, brightening both the quality and quantity of consumer data that can now be captured. Overtime data collection expanded, and now spans to hundreds of parameters. Viewing history, play clicks, devices used, internet speed, time and duration of viewing, browsing habits, customer support interactions, record scheduling, genre preferences, channel and content switching and much more. The once small tracking clickstream had slowly transformed into big data. It is this data, which is now bringing insights to understand the subscriber. Fast-forward to today and major broadcasters and networks (HBO GO, CBS All Access etc.) are reaching directly to subscribers with full-fledged OTT services. In summary, while early VOD and TV Everywhere services provided glimpse of consumer behavior, direct OTT relationship has opened data floodgates for a 360-degree subscriber view.
Data generated directly on video platform, termed as first party, is primary and most important data for analytics. But a lot of the data is also collected from social media platforms like Facebook and Twitter, capturing real-time audience reactions to content. In addition second and third party data collected from DMP’s provide valuable correlations and enrich the analysis from primary data.
Big Data, Predictive Analytics and Machine Learning in OTT Video
Online video services have been collecting data for several years now, making it a perfect case for big data analytics. Predictive analytics, which uses techniques in data mining, statistical modeling and machine learning, is taking big data analytics to its next logical level. Predictive analysis with help of supervised machine learning algorithms is bringing powerful prescience in predicting future trends. It can help predict what content to create, need of encoding profiles, new device adoption rates, expansion needs for edge servers, geographic growth and mobile consumption trends etc. These predictions are in turn influencing decisions at the very core of online video business. Helping in putting future plans for operations, customer support, content strategy, promotions, personalization and more.
What answers can Data provide?
Famous economist Ronald Coase once quoted, “Data may not contain the answer, but if you torture the data long enough – it will confess“. And rightly so, collecting data is not the end game; neither is reporting or building dashboards. Rather it all starts with a simple question, albeit a hard one. What are we looking for, what question that we want data to answer? And only when we have the right question, we can use predictive analytics to find the answers from available big data.
Online video players like Netflix and Amazon have built extensive data analytics strategies and are leading the way video players are using data to run their business. Netflix decision to enter entertainment production was a result of data analysis, as well as its $100 million decision to outbid top TV channels to earn rights for House of Cards. Media networks and broadcasters although late in the game have extensive consumer data, and it is about time for them to act on their data strategies to get real answers.
Rise of Predictive Analytics in Online Video
Predictive analytics and machine learning is playing an important role in data analytics where traditional tools have hit a roadblock. It brings capabilities to conquer massive amount of data at scale and discover implicit patterns in structured and unstructured data. Below are some of the key factors, which are fueling their growth in online video.
- Access of Big data sources is the key for data analytics. Today, popular online video platforms capture billions of clickstreams of user interaction data every day making it a perfect case for predictive analysis.
- Access to large infrastructure, had perennially been a critical roadblock for big data analysis. Cloud services like AWS and Azure have now made infrastructure scalable, agile and accessible. It has become quite easy to quickly setup and get running a large storage and processing server farm without upfront CAPEX or technology challenges.
- Role of data analysis was traditionally restricted to statisticians. But emergence of software platforms with distributed data processing has now opened data science to software developers. Open source platforms like Hadoop, Spark, H20, TensorFlow with machine learning have bridged the gap with great elegance.
Content creation and distribution is still the core business for online video players. Predictive analytics using machine learning algorithms is helping stakeholders to forecast future trends as well as empower them to take data driven decisions. It is still early days, and we are only scratching the surface in use of knowledge from historical and real-time data to influence operational, marketing and content strategies. Great revelations await us, as true potential of big data and AI gets unlocked in the coming days.