Sign In
Technology Suppliers

Asia's Relevance Test For Big Data

Content recommendation engines, increasingly popular in North America and Europe, have been slow to take off in Asia.

Pictured (from left): Aravind Venugopal (moderator); Oliver Tan; Richard Dowling; Sean Li

Systems developed outside the region need to adapt to different scripts and languages, while meta-data – the descriptive tags used to identify and organize different types of content – are relatively scarce for Asia-made movies and TV shows.

Asian OTT platforms such as Hotstar and Iflix, speaking at APOS Tech, advocated human curation and geodemographics as better drivers today of revenue and consumption – the primary goal of recommendation engines – in markets they operate in.

Richard Dowling, VP of products at ThinkAnalytics – a firm specializing in TV and video recommendation software – countered that these workarounds offer short-term and unsustainable benefits.

“What you need from a good recommendations engine is the spike as you put it in – as people find content they enjoy, that they didn’t know was there – and then it should continue to grow, or at least not dip,” Dowling said.

“Over a long period of time, that’s important.”

European pay-TV and broadband major Sky, for example, abandoned alternative statistical approaches in its latest recommendations upgrade, relying instead on content descriptors from meta-data combined with viewing behavior.

“It takes away need to know demographic information, to know anything else, and generates a unique source of information about the customer, because that customer only volunteers that information when they want to consume content,” Dowling said.

The sector continues to build traction in North America and Europe. In September, Netflix – usually reluctant to shed light on its internal performance and machinery – opened up some of its meta-data to Liberty Global, as part of a 30-country set-top integration and distribution deal.

Such moves should help open up content libraries, ratcheting up competition for TV viewing between incumbent services and OTT.

'I don't see that data culture here in Asia'

Nonetheless, data scarcity continues to apply the brakes in Asia on new ways to drive revenue and retention through analytics.

“In the US, they are very generous in the data they share,” noted Oliver Tan, co-founder and CEO of visual search startup ViSenze, speaking on the same panel at APOS Tech.

“That becomes part of the learning process,” Tan added. “I don’t see much of that data culture here in Asia.”

ViSenze, which focuses on AI or machine learning, promises to add another descriptive layer to video, by automatically identifying and cataloging images within the content.

So far, its customers are mainly online retailers, such as Myntra in India and Zalora in Southeast Asia.

The company is also piloting a new service with Asian VOD service Viki, which allows viewers to match up clothes seen in TV shows with merchandize from ViSenze's ecommerce partners.

This approach might also be useful in augmenting content recommendation, something ViSenze is exploring.

“The more signals you bring into your data stack, the better you will be,” Tan said.

Despite challenges around data availability, existing traffic information can also help OTT services, suggested Sean Li, Asia-Pacific & Japan technical services head for Akamai, an online content delivery network.

A clearer view of data traffic can shed light on areas such as video quality and performance.

“Surprisingly, a lot of OTT players have yet to adopt adaptive bitrate technology [which can adjust streaming quality on the fly],” Li said.

“The audience behavior is there, the technology exists," he continued. "Data analysis can provide a lot of insights to help tackle the situation.”

Furthermore, operators that only monitor traffic on their networks have a blinkered view of overall trends, Li argued.

“If you can have all the information over the internet, our customers can be more proactive,” he said.

Please sign in to leave comment or reply