Almost a year ago, Gartner Inc.’s annual Hype Cycle for Emerging Technologies sparkled with newcomers like big...
data, consumerization and natural language question answering (in the vein of IBM’s Watson). But another new recruit, appearing on the cycle’s initial slope, received considerably less fanfare: video analytics for customer service.
Businesses have been analyzing Internet videos for years, seeking ways to determine who’s watching what. Think Nielsen ratings for YouTube. Like traditional Web analytics, businesses are attempting to serve up targeted advertisements and keep tabs on conversion rates or the number of times a viewer clicks though links to make a purchase or provide personal information.
“It’s been around a long time,” said Joe Stanhope, senior analyst for Forrester Research Inc. in Cambridge, Mass. “As long as there’s been video on the Internet.”
But these days, video analytics is more than dissecting demographics and determining how viewers consume the content. The definition has broadened to include analysis of the content itself and even the capability of image recognition.
Gartner’s hype cycle, for example, describes video analytics in content-based rather than viewer-based terms: “Video analytics is mostly being deployed for security purposes (for example, tracking suspicious behavior or trespassing). The retail industry is the biggest user of video analytics -- outside of security -- to identify and analyze foot traffic, and identify customer events.”
“There are no hard and fast rules on the definitions,” said Hung LeHong, a research vice president for Gartner Inc., based in Stamford, Conn., and co-author of the 2011 emerging technologies report.
Like other industry-related terms, how video analytics is defined can change from one person to the next: While Stanhope refers to viewers, LeHong talks about content. But both analysts agree that analyzing video is on the upswing. Albeit slowly: Compared with other 1,900 technologies in the hype cycle, Gartner lists video analytics having the smallest benefit.
Analyzing video content
LeHong said video content can be analyzed in two ways: historically and in real time. For historical analysis, the user can tailor the software to sift through footage and collect data on a type of event that has already happened.
“This is an application that’s very popular in retail,” said Zvika Ashani, chief technology officer for Agent Vi, a video content analytics software provider based in Rosh Ha’ayin, Israel.
Store owners can, for example, keep counts on customers entering or exiting the store through the use of security-camera video, Ashani said. The data can be used to discover peaks and lulls during certain hours of the day or days of the week. Or it can help anticipate staffing needs or calculate a brick-and-mortar conversion rate.
But store owners can also use video to create heat maps, in this case tracking how customers move through the store and where they spend the most time, Ashani said.
On a more sophisticated level, LeHong said, users can program the software to identify items such as a logo or a product. The software can then tag and timestamp the item each time it appears.
“We call that tagging essentially metadata,” LeHong said. “It’s descriptive data about the video itself.”
More on metadata management
Learn about metadata security and preventing leakage of sensitive information
Read up on the difference between metadata and sub-metadata
Discover a few metadata basics
Building metadata enables a business to create a CliffsNotes version of specific items, making them easier to access and search, LeHong said. But metadata can also cut down on cost. Because video footage tends to eat up a lot of memory, metadata can help businesses distinguish between relevant footage, which it should keep, from irrelevant footage, which it can discard. According to Ashani, indexing specific items can be done while footage is being recorded or, for the unanticipated, after the footage has been recorded.
More sophisticated still is real-time reporting, which requires events to be processed as they’re happening.
“That takes a lot more firepower, more sophisticated algorithms, and it’s just a lot harder to do,” LeHong said.
The technology can be programmed to flag events that break predetermined rules and then trigger an alert. For Agent Vi, these events or rules refer to certain behavior defined by the customer such as a person traveling in a certain direction or a vehicle traveling at a certain speed.
“That’s mostly a security-type of application where the security officer or whoever is responsible for maintaining the [security cameras],” Ashani said.
LeHong agrees. Beyond security or surveillance, most businesses aren’t using real-time video analytics even though the technology exists, he said. Although he doesn’t discount the possibility of a future implementation to target specific shoppers based on store location or product interest, LeHong said there is no solid business case today.
Traditionally, most analysts place video analytics into one of two categories: Analyzing viewer behavior as they watch the content or analyzing the content itself.
“There is a third class,” LeHong said. “Some might call it video analytics; others won’t. I like to call it recognition.”
LeHong is referring to facial or product recognition. While Gartner considers recognition technology to be separate from video analytics, LeHong said the loose nature of how the term is defined means there are still some gray areas.
“We have it as a separate thing, but the notion is the same,” LeHong said. “It’s taking static shots or video and it’s recognizing something because it’s using analytics in the background.”
That could mean the big IT vendors like IBM, SAP and Oracle may be competing with leading-edge technology companies such as Google, Apple and Amazon for more than just storage or social media capabilities.
“It’s the class of really early adopters, the guys who set the marketplace, if you will, who are acquiring businesses [with this kind of technology],” LeHong said. “It’s another proof point that the guys like IBM and SAP are just behind, at least with respect to something like this.”