In Part One of this blog series on data science in media, we looked at the increasing role that data is playing in our industry and the various forms in which it is generated. The challenge then is to derive useful insight from this information source, one that is made more difficult by the scale and growth involved.

This is where the science part plays a critical role and we can leverage a lot of the tools and techniques that have been employed and advanced in other industries that have embraced so called Big Data. In fact we are not just employing the techniques from other industries, we are increasingly employing the people too. Most larger media organizations have established data science teams, bringing in experts from data driven industries such as retail, insurance, financial services as well as academia.

 

What do we mean by data science in this context anyway?

The term broadly defines a range of techniques and systems that analyse data sets (and media content itself) to search, discover and visualize insights that can, in turn, be used by people and machines to drive actions. Examples include the use of predictive analytics to assign probabilities to future events based on past activity, machine learning to automatically build algorithms that find patterns in data and content, including speech recognition and natural language processing algorithms that audit and transcribe dialogue into text based data or provide voice based input controls to interact with devices.

The use of computer science, computational linguistics and artificial intelligence will provide new forms of content discovery such as the Siri Remote included in the forthcoming Apple TV and already used in Amazon’s Fire TV.

 

So what will this actually do for us?

As we become a more data driven industry, the use of data science will increasingly pervade the decisions we make, how we engage our audiences and how we deliver our content. Each of these categories is explored below:

Insight

This area is where much of the effort on data science within broadcasters has been focused to date. Data scientists and other specialists are working with traditional departments to help with audience segmentation, programme commissioning, channel scheduling, marketing, CRM, subscription and ad sales.

Personalisation

Whereas insight helps us understand audiences better, personalization allows us to better engage them. As more viewing takes place on connected devices, the opportunity to personalize what and how content is served becomes a reality. This is where data science will play an increasingly important and crucial role. Human curation of channels works well in a one (programme) to many (viewers) model but personalization attempts to deliver a more individual, many to one, experience and curation at that scale requires data and algorithms. These systems will enable automatic content recommendations, personal linear channels, targeted advertising, intelligent search and discovery and greater real-time contextual awareness of each (for example understanding the weather and events in your area to better target programmes and ads).

Delivery Optimisation

Finally, the quality of viewing experience is only as good as the quality with which the content is delivered and served. Poor quality video, content that pauses and buffers or is slow to begin playback are huge turnoffs for viewers and are regularly cited as the most impactful metrics for audience engagement and retention.

As more content is delivered over IP networks, one consequence of personalization and on-demand viewing, we can again use data science to predict viewing patterns and optimize delivery accordingly (for example pre-populating CDN caches, automatically programming DVR recordings or switching on LTE Broadcast delivery in a cell site). The source of this data can also be real-time including network transport data or Twitter conversation around a show.

Of course taking advantage of data science in all of the above examples presents challenges at a technology, skills, organisational and commercial level. It also creates huge opportunities to innovate, differentiate and grow audience share and revenues. The last blog in this series will look more closely at the challenges and opportunities that face media businesses and the wider industry in this area.

Steve Plunkett, CTO