Managing the data analytics opportunity with cloud computing

Media companies have always dealt in data, but the scale of the data available to them, which they are themselves collecting from online and multiscreen engagement, is exploding. For some it is a concern, as they do not know what to do with all the information. Equally, it is a huge opportunity to improve and monetize services.

Channel 4, the UK commercial broadcaster, has put data at the heart of its business strategy. The company is on a mission to understand more about its customers and in February it counted 10 million unique registered users. One of the main purposes of its viewer engagement strategy is to enable more targeted advertising for its VOD, but also to improve its commissioning and programme making.

In an example of how the use of cloud computing can benefit media companies, the broadcaster uses Amazon Elastic MapReduce, a service from Amazon Web Services, to provide the analytics capacity it needs. Sanjeevan Bala, Head of Data Planning and Analytics at Channel 4, says that as analysts can now spin up clusters of compute power based on Intel processors, there is potentially infinite processing power available to them, without the traditional upfront infrastructure costs.

The channel owner can analyze real-time viewer interactions, something that can have direct value in terms of advertising. It has also thrown off the constraints imposed by data sampling, where traditionally you had to use a fraction of your audience, like 1-5%, as a sample of all viewing behaviour. “You might miss the needle in the haystack in terms of signs and signals but now our analysts can run queries and algorithms across the entire data set,” Bala declares.

Adrian Drury, Lead Analyst, Media & Broadcast Technology & Services at the consulting firm Ovum, notes that broadcasters have started processing more and more data in real-time about their audience. Given the number of data sources available and the bursty nature of data in the TV industry, he thinks the only scalable way to handle this is through the cloud.    

When it comes to data crunching, having data storage and archiving on hosted infrastructure, like a datacenter, is a good thing, he reckons. The real-estate is cheaper than in the broadcast facility and the data is not going to get in the way there. Security is a big issue though. Drury says people are “paranoid’ about the danger of malicious attacks on their data so they need people they can trust to manage the data securely.

Bouygues Telecom, the French IPTV provider and quad-play operator, is investigating the benefits of the cloud for data processing as part of a wider initiative to test cloud-powered services with Cisco. According to Hubert Cariou, Director at the company, “We can make use of big data processing in the cloud to better understand the end-user behavior and the family profile of our IPTV customers. This results in better service personalization, better service recommendation, better service satisfaction and potentially more accurate targeting for advertising.”

Amruta Shankar, Manager, Platform Strategy Analytics at Service Provider Video Technology Group, Cisco, lists examples of where big data analytics can help Pay TV operators. With descriptive analytics you can understand what is happening on your services, like the most popular channels or on-demand titles in real-time or historically. Among other things, operators are interested in using better viewing insights when renewing content deals or deciding whether to license content for Android devices as well as iOS.

With the advent of connected devices, you can move to a census based reporting system for audiences and advertising, no longer relying on panel samples. “If you want to manage live data from million of clients simultaneously you need computing and storage that only the cloud makes possible. You could do it on premises but it would be prohibitively expensive,” she argues.

Cisco provides an audience profiling service that ingests and anonymizes huge volumes of subscriber information about demographics and viewing behaviour and uses machine learning to map one against the other. Without requiring user log-ins, it enables a platform operator to build an accurate profile of individual viewers, like the high income 40-50 year-old male who only turns on the TV for news at 6am and 8pm. The information is potentially valuable to advertisers and the solution has been trialed by a European service provider.

‘Big data’ can also be used for diagnostics and QoS insights in a world with more networks and devices to monitor, while the science of prescriptive analytics uses computing to suggest actions. A simple example is whether to pre-cache an on-demand programme at the edge of the network, based on how consistently households are watching it and the density of these regular viewers.

“Pay TV operators are now handling volumes of data they have never dealt with in the past. As far as the cloud is concerned, storage and compute capacity is infinite,” Shankar declares.

More reading:

Videonet report, ‘Making the cloud work for TV’