Data is King

Data is the currency on which TV Everywhere players depend but service providers are facing organisational hurdles in managing the wealth of information at their disposal. Adrian Pennington reports.

Data is everywhere but it’s how you use it that counts. Many service providers are only scratching the surface of what is possible. Even where there’s a will to interrogate data and effect rapid response to consumer needs or business models, many multichannel networks face an uphill task to overcome legacy organisational barriers.

That’s in stark contrast to pure-play streamers like Netflix or Amazon, which have built a business on integrated customer service and technical operations. The OTT providers cross-correlate data sources that provide insight into quality of service (QoS), marketing, advertising and content recommendation. These draw on remote control commands to the set-top box sent back over the return-path and server logs that record media player interactions.

With traditional broadcast networks there wasn’t a huge need for data. The equation could be reduced to quality of content and how many people watched. With pay TV the model barely shifted. If shows were packaged at a decent price the subscription rate went up. The need for data wasn’t great since the variables didn’t waver. When TiVO introduced the concept of time-shifted consumption the cracks began to appear. OTT has taken this to another level.

“The internet opened a floodgate of consumer choice,” says Keith Zubchevich, chief strategy officer at OTT video specialist Conviva. “It handed control of the TV from networks to viewers. The data that is needed now is of a fundamentally massive order of difference compared to what has gone before.”

Data types

There are broadly three types of data: consumer viewing behaviour, programme metadata and network performance statistics. The latter has been a factor in network capacity planning for some time but is also starting to be used in other areas, such as content acquisition.

“For a long time QoS monitored bit-rates, how many sessions failed. Increasingly the data is about what devices are being used and how user behaviour differs device to device. How long are average sessions and when do they occur?” says Edgeware’s vice-president, products, Johan Bolin.

Through QoS analytics, a service provider can understand what services are more popular during certain times of day, in certain geographies, or on certain devices, providing insight into holes in offerings. This data can then be used to adjust existing services, or add or remove suppliers in the ecosystem.

“A lot of the focus is still on more technical use-cases but the ability to get data from end-customer devices is driving additional areas such as marketing, customer base management, customer care and service management,” says Per Unell, business development, Agama Technologies.

“Processes like fault detection and localisation, network optimisation and change management are very much a reality,” Unell adds. “We also see significant use in service management such as SLA and overall service performance tracking. Some customers are also using data to support their customer care and customer understanding processes.
In general, it’s more straightforward to see quantifiable benefits in operational processes – fixing problems faster, solving customer issues at first call. At least as much value can be realised by systematically and proactively tracking down issues before they become problems, but it’s harder to quantify beforehand.”

Another component of consumption analytics is device-level data, which is increasingly important as the number of devices grows. “The power of this type of data is granularity,” says ABI Research in a white paper on the topic published by Conviva. It can be used to guide purchasing for service providers. Real-time data can help with monetisation decisions for unexpectedly popular content, and can guide future investments to identify more popular content, states AIB Research. Popularity of live events can be accurately judged, and licensing and delivery decisions can be centred around this.

“You’re beginning to look at being able to measure the way that people interact with a TV system in the same way that they interact with the internet,” says Andy Hooper, vice-president, cloud, solutions and services at TV technology provider Arris.

Scratching the surface

“It is relatively early days for the use of big data for service providers,” says Peter Docherty, founder and CTO at recommendations engine provider ThinkAnalytics. “Data is already captured but is not being taken advantage of as much as it could be. The risk of not using data to drive the business is a lost opportunity. Let’s say only 40% of your VoD catalogue has been watched. If you don’t have data you won’t know that and if you don’t have data about what is watched or being routinely declined then you can do nothing about it.”

It’s not as if operators haven’t recognised the need or that solutions are having no effect. ThinkAnalytics’ research suggests that a few months after integrating its recommendations engine, clients saw their subscribers increase viewing time by 20-50%, and the number of channels watched rise by 25-35%.

Subscriber management provider PayWizard says it has delivered acquisition campaigns that drive conversions up to 25%, and run churn reduction programmes that achieve up to 60% conversion. The most recent figures released by Sky from its 500,000-home Sky Viewing Panel show that channel switching during Sky AdSmart commercials was 48% lower than for standard non-targeted ads. “As viewers cannot distinguish AdSmart commercials from any others, higher viewing levels can only be attributed to customers finding them more interesting or engaging,” observed Jamie West, Sky Media’s deputy managing director.

Pancrazio Auteri, chief technology officer at content personalisation firm ContentWise explains some of the ways it uses [icitspot id=”570902″ template=”box-story”]data to assist customers like Maxdome, Mediaset and Sky Italia. The first is to boost churn prevention by detecting anomalies in behaviour.

“If patterns in behaviour diverge from established patterns this may mean a user is using a competitive service and can be an early predictor of a cancelled subscription,” says Auteri. ContentWise data is also used to make service provider promotions more relevant: “If users are sent irrelevant items then any associated communication from that service provider will also be seen as irrelevant so we match the promotion to user habits or tastes.” Doing this has seen a rise of between 20-40% in subscribers opening and viewing a promotion, ContentWise claims. “The same method can be applied to advertising. If a service provider is sending promotions of different advertisers we look at the profile of the user and lifestyle traits and try to narrow down the promotions while increasing their relevance.”

A third form of data usage, dubbed ARPU Rebuilder, will be offered as a module by ContentWise this autumn. This will include a set of algorithms designed to up-sell related micro-subscriptions to users and will focus on the free trial phase of a service and the first month of subscription to ensure that the new user understands the value of the content offer. “Based on our research the obstacles to large-scale adoption of TV Everywhere are the lack of awareness from consumers that the content they are interested in is available and the fragmentation of the applications, making it difficult for users to find and seamlessly consume this content,” says Auteri.

ContentWise uses viewability tracking to ensure that each user can see a different, uniquely personalised UI. “Given the fragmentation of TV Everywhere applications, metrics and KPIs cannot be computed across all applications owned by content providers,” says Auteri. “The unified discovery provided by the pay TV operator UI is the best place to measure user behaviour, across all touch points – i.e. screens, apps, devices.”

Subscriber management systems developer PayWizard estimates that between 5% and 8% of a pay TV operator’s OPEX is down to subscriber management and that data from SMS can have “a dramatic impact” on the business’s “viability, competitiveness and profitability”. “SMSs are a natural collection point of statistical data regarding a pay TV operator’s business processes and subscriber community,” says Bhavesh Vaghela, PayWizard’s chief marketing officer. “This raw data, when given context and viewed against trends, allows marketers to develop and track sales initiatives. Questions like ‘do our free month offers lead to full subscriptions’ or ‘which device is the most popular for viewing as to impact our application development strategy’ can be answered by a whole host of valuable insights uncovered through reporting and analyses. These answers can help solve short-term issues and improve the longer-term profitably of a pay TV business model.”

Data consolidation

Vendors uniformly contend that, from a technical angle, big data collection and analysis is not the issue. The chief problem is the ability of service providers to handle it.

“We have all the tools and databases to perform analysis. The challenge is joining the data together from a business perspective. Service providers are trying to gather data from different parts of the business but have quite a way to go. There’s a lot of emphasis on the customer acquisition side and on customer retention/churn reduction programmes but not so much focus on the stages in between. For example, when you’ve got an active customer how do get them to spend more? That’s a lot to do with not having data in one place able to serve a more personalised engagement with customers,” says Docherty.

Arris reports that a number of operators have built big data teams operating as an IT service internally with mixed results. “This works well in organisations which understand the value that flows end-to-end from capturing and analysis to taking action from big data insights,” says Hooper. “In other organisations, however, there are too many barriers to collecting and sharing data. Departmental teams tend to keep data within their own group. It’s a classic big organisation type of problem.”

An example, cited by Hooper, is a company deploying a monitor solution into its multiscreen video apps for smartphone and tablet. “Using that solution they were getting a huge amount of data, yet people in charge of the call centre had no view on when a session failed. Despite the operator paying a licence to this vendor it had no oversight on the poor customer experience arising out of buffering video. Either through inefficiency or deliberate obstruction, that information was not being leveraged end to end across the business.”

As Hooper sees it, the customer’s experience with a provider now spans data silos from the call centre to operational teams examining session data and digital marketing teams interrogating intent to purchase. “This is not being done at anywhere near enough scale,” he says. “Customer experience management crosses organisational boundaries. Some service providers are addressing this by installing a chief digital officer or customer experience executive, even at board level, but they need to to do more.”

Big data is an established IT discipline in many industries, but most telco or cable companies retain a legacy of network/operations teams separate from marketing and consumer-facing departments.

Hooper points to home network management as another area barely addressed by pay TV operators. “Trouble shooting of this falls on the responsibility of the service provider, explicitly or implicitly,” he says. “It includes management of the home network, [covering issues such as] whether there’s good home WiFi, or the kids are moaning at Dad because they can’t download game updates. All these things are part of the subscriber experience and typically they will end up talking to different bits of the service provider organisation when it should provide one entire customer experience journey and one digital strategy with big data enabled to deliver insight into this.”

Sharing data sets and personalising offers is hampered by privacy issues. “Targeted advertising is still in its early phases, with just a few operators using it in production. It also requires additional infrastructure and generates privacy and personal integrity requirements on the solutions used,” says Unell.

“Video professionals are struggling to adapt to new business models and must use data analytics to manage and grow their services,” concludes Sam Rosen, vice-president, consumer at ABI Research. “Leveraging a single, unified dataset for the needs of different functions in an organisation, and opening up avenues for data sharing between affiliates jointly responsible for a service, can help align everyone on a common definition of success.”

Edgeware’s Bolin believes a single repository is desirable, and perhaps possible, but not soon. “Given that there are so many different systems and sources of data it would be a challenge to have a centralised data store continuously updated and compliant but as the market matures and business intelligence advances I wouldn’t rule it out,” he says.

“You can spend more time crunching data than analysing it,” agrees Zubchevich. “I don’t think we’ll get to a unified single data service. Publishers and pay TV operators have to break data into chunks and look for key providers of, for example, purchase data, advertising and content recommendation.” Anything to do with the playback subscriber experience should be measured, he says. “Failure to do so means churn and subscriber loss. Consumers are polling publishers online. The consumer will terminate their relationship with a network if that service provider is not proactive.”

Buffering and delivery analysis should also be extended to ads. There is evidence from Conviva that as much as 58% of viewer churn is based on poor online ad experiences.

“The impact on a viewer’s experience from ads is massive,” says Zubchevich. “If I watch a show and it’s riddled with ads I’m getting ad fatigue and I’m beginning to look for content elsewhere. The fundamental next step for service providers is to monitor ad impact.”

Not coincidentally, Conviva is launching an Ad Insight product which expands the capabilities of its video playback monitoring.

“Service providers could do more in this area,” agrees Bolin. “Today, very little is being done partly due to data being sourced from two different parts of the business. There is data from the ad insertion server about the actions of customers interacting with an ad, and data from the ad streaming server about the rendering of that ad. There is a need to cross-corrolate this data. As ad personalisation grows I would assume service providers would see these reports as much more of a requirement.”

Acting on QoE

A study published in March by IneoQuest found that more than half of consumers who watch streaming video have experienced ‘buffer rage’, defined as “a state of uncontrollable fury or violent anger induced by the delayed or interrupted enjoyment of streaming video content from OTT services.” It’s no laughing matter. With cord-cutting on the rise, a better understanding of the implications of buffer rage is essential, IneoQuest argues. The contention is that metrics such as packet loss and delay can be used beyond an interpretation of the performance of services to measure the experience of the user. Measuring subjective user experience with objective data is where Quality of Experience (QoE) comes into play.

“Tracking packet loss can highlight problems in a network, and these problems can be extrapolated to possible user experience difficulties; however, there are a few issues with this,” says ABI Research. “Not all service difficulties always lead to user experience degradation, users vary in their tolerance for network problems, and different content types – i.e. short-form video versus long-form video – are affected more by network difficulties.”

Logging a consistent stream of data – such as stream completion or early exit – and correlating this to user data creates a combination that allows for deeper and more personal data dives. Combining this with dashboards to help with specific filtering, such as geographic location, content source and user ISP, helps analysis and subsequent decision-making. According to vendors, however, most companies are primarily reactive and don’t have the resources or time to proactively look for potential problem.

“What service providers need to do is to proactively poll the infrastructure and network that impact customer experience and match that performance information against fixed, dynamic and baseline thresholds as well as configurable SLAs to identify issues that will impact customer experience before the customer is affected,” says Gregg Hara, vice-president, business development and marketing, Centrica Systems. Centrica’s NetOmnia Cable Assurance polls all customer premise devices as well as network devices and correlates this information to identify issues. “This can then automatically kick of a work order or trouble ticket to initiate a truck roll and get a tech onsite to resolve the problem before the customer even notices an issue,” he says.

Viewers are growing impatient when QoE deteriorates. In its recent OTT: Beyond Entertainment Survey, Conviva concluded that one in five viewers will abandon poor experiences immediately, and that 90% of viewers return to services that deliver a superior experience. After a poor experience, one in five will never return.

This leads Conviva’s Zubchevich to conclude that content is no longer king. “For the first time you can forget quality of content as being the single most valuable metric. The number one currency is QoE. We are just starting to see operators look at marketing the experience of viewing as important as the content itself. This is not something they’ve ever collected in the past and it’s a fundamental shift.”

Publishers still care about the quality of content, of course. Zubchevich’s point is that with so many ways consumer scan get the same content, the key will be the experience they receive from the site serving it. “We need to redefine QoS from basics such as ‘is the stream available at all?’ and ‘can I watch it largely uninterrupted? to questions about resolution. If I watch a HD or UHD TV broadcast and move to an IP-based provider I am not expecting a poor experience in comparison. We need a zero tolerance approach to starting the stream. What used to be acceptable is no longer acceptable and failure to address buffering or bit-rate issues in that moment means you lose the consumer,” he says.

Arris’ Hooper suggests that the mantra ‘content is king’ has been a fall-back excuse for some service providers. “As the market fragments into different content sources you’ll find that consumers will gravitate over time to the site where they’re having least friction. That means pay TV doesn’t just have its business threatened by OTT but by new data pipes which can provide a greater customer experience journey through the content lifecycle. Having exclusive content deals is a defensive mechanism. Enabling a better customer experience will deliver more positive brand benefits in the longer term.”