Video operators must learn to take real action based on data analytics insights

Alan Burkitt-Gray
Published on:

Data about the video consumption provides a very rich window into the lifestyles of users that operators are starting to mine to both run their networks efficiently and ensure customers get the experiences they require. However, video content delivery and networks are more complex than other networks so a more proactive approach to video data analytics is required. Co-sponsored feature: Ericsson

Francisco Huerta, Ericsson

Francisco Huerta: We need to close the loop between basic 
video data, business-meaningful information and actions

While there has been a lot of discussion about how valuable insights can be generated by big data analytics applied to network and user data, there has been less exploration of the value that can be extracted from data relating to video consumption.

It is now well understood that big data can be mined and analysed to improve quality of service, network utilisation and, ultimately, operators’ ability to monetise their networks. In addition, the data insights themselves, when aggregated and anonymised can be revenue generators in their own right if they are packaged correctly and sold to third parties that find them relevant.

Video services are just another network service and therefore the application of big data insights has similarly significant value to add.

However, the delivery of video involves a greater level of complexity than other network traffic.

Video is highly sensitive to jitter, packet loss and latency and is often consumed on a wide variety of devices and networks. In addition the content is often sold at a premium by third parties that use operator networks for the delivery.


Operators therefore face a series of multi-layered challenges in delivering video. They must configure their capacity to support peaks and spikes in demand. They must take into account the differences in quality required for a viewer on a high definition wide screen television in their home and a viewer watching over a cellular or wifi network on a mobile device.

In addition, they must ensure the quality of service delivered is as expected by their consumer and in many cases by the content provider, which may be paying them for a premium level of service.

It is fundamental to business models that involve advertising supported or sponsored delivery that what is being paid for is delivered.

For instance, if a content provider remunerates a network operator for delivering high definition video to a mobile device, that experience must be flawless. Equally, if the cost of video content to the user is supported by advertising, those advertisements must be served without error.

However, analytics insights into video traffic are not only about service assurance and network utilisation. The data operators collect, once analysed, can be applied to marketing to enable sale of additional services and content and to context-aware advertising, which is increasingly subsidising user consumption of content.

The greater value for advertisers is in knowing as much demographic information about the likes and dislikes, interests and disposable income of consumers and operators can assemble this in anonymised format from video data analytics.

Rich data

Video data is particularly rich because it enables a lot to be learned about the consumer. That knowledge encompasses what movies they watch, what sports they follow, what news shows they watch and when and how often they consume video.

Ericsson has developed its Integrated Video Insights (IVI) framework to address both the service assurance and the marketing needs of operators, content providers and their partners.

“We’ve created IVI to fulfil the needs of video service providers,” says Francisco Huerta, the head of Video Platforms Consulting and Systems Integration at Ericsson. “Our video analytics framework can process the massive amounts of decontextualized data sitting in their operational environments to extract meaningful insights that provide an angle for marketing and enhanced customer care.

“IVI can also help answering questions such as whether a provider is selling content at the right price,” he adds. “It’s a multi-layered business intelligence tool.”

Video delivery is much more complex than traditional network services delivery. “Operators come from environments that are pretty much under control in terms in terms service quality monitoring, such as cable or satellite delivery of video,” says Huerta. “However, the on-demand, multi-device, multi-network model sees the complexity of networks increase dramatically. In addition, if you expand your operation to the over-the-top market quality becomes of great concern to service providers, including broadcaster that are providing OTT services through the open internet or public CDNs.”

In addition, the video consumption market in terms of users is highly fragmented. Users migrate across many types of device and network and also one account can be used by multiple users. It’s important to be able to segment each account according the user profiles of each viewer. If this doesn’t happen parents will be inflicted with advertising and content offers relating to their child’s favourite cartoon.

IVI then is a way to apply big data fundamentals to video traffic but it goes further than that because it doesn’t stop at simply preparing information from the big data. The framework has been designed to enable video providers to be proactive so they can address issues before they become reality.

“The solution builds on the in-house knowledge and technologies used in other big data applications, but what’s unique is the capability to model key performance indicators and net promoter scores for video operations,” adds Huerta.

“Another difference is that IVI is specifically designed to be integrated with video-related components. We extract the most from IVI deployed at operator infrastructure holistically starting with signal acquisition at the headend all the way down to user devices like set top boxes, smart TVs and mobile devices.

“You can see IVI as a platform that establishes all these different tentacles and extracts information from them,” he explains. “Video structures tend to be complex so, by definition, it’s impossible for these types of solutions to be off the shelf. Aside from a platform-agnostic approach, you need to calibrate the thresholds and the levels of key performance indicators, for example.”

Huerta also emphasises that IVI goes beyond simply gathering results from the analytics. “What we’re trying to do is take the additional step and try to answer questions for an operator that are relevant. A good example is to identify what steps can be taken to reduce churn. What actions can an operator take to prevent a user taking negative action? That might relate to the quality of the network or the quality of the content and it’s important for an operator to understand which.”

He gives the further examples of: working out whether the network is properly dimensioned, whether it has the resources or capabilities required to meet user demand and whether it is running out of content delivery network capacity. Away from the network he says the system can play user-oriented roles by intelligently promoting contents, running micro-campaigns or injecting personalised ads.

“We need to run away from the academic approach and look to real possibilities and value to video service providers,” he adds. “The system can learn patterns and a specific use case we have implemented has taken advantage of the fact the platform knows the number of subscribers and the month of the year or day of the week, can establish patterns with similar dates in the past.

“If the platform detects that last year when you offered the Superbowl you were on the edge of your CDN capabilities it will generate an alarm or, if you have an elastic environment it will react by provisioning more capabilities,” Huerta explains. “The proactive side of iVi is very important because it ties up big data with actions, which opens doors virtually endless scenarios and use cases.”