TVision Insights is the data analytics company redefining audience measurement. They were recently named to the Advertising Research Foundation’s Innovation A-List, the ARF’s top award to innovative startups in advertising. But when they started in 2014, while cofounder and CEO Yan Liu was pursuing his MBA at MIT Sloan, it was just Liu and a few PowerPoints, as he tells it. In four short years they’ve grown into a sizable company of 45 full-time employees serving top brands and agencies in the TV and media industries, including ABC/Disney, NBC, and The Weather Channel. While Liu is proud of the rapid growth, he is well aware that innovation is a constant process. “Last year, we improved how we operate to leverage the latest deep learning computation technology,” he says. Integrating cutting-edge approaches to AI and machine learning into their core technology goes hand-in-hand with TVision’s mission: “The end goal is to offer the highest quality, unique data to every stakeholder to help them make better decisions so the entire ecosystem will be more effective. Using our platform is a win for content providers, brands, agencies and ultimately, the consumer.”
In 2017, television advertisement spending totaled $205B, with the U.S. market alone accounting for $72B or 38% of global TV ad expenditures. But in an increasingly fractured media landscape, where the advent of digital is just one aspect of the equation informing media consumption, deeper insight into ad placement is essential. Yet, despite the size of the market and the shifting nature of consumer habits, our tools for gauging these behaviors haven’t changed much in close to 40 years.
At the end of the day, media planners and buyers are really buying audience attention,” says Liu. “We can overlay that information on top of Nielsen ratings to help them make better decisions to increase their ROI.
Liu is well aware of the disparity: “The only data widely available are traditional TV ratings,” he says, “which are used to determine the pricing for ad slots, which ads to run and when—basically, all the important decisions in a massive industry are largely being made using an outdated model.” Namely, the Nielsen ratings, which capture whether or not a television is on and what show or ad is on the channel but not actual user engagement. Nielsen’s traditional people-meter technology does a fantastic job of collecting what is on the TV screen, but it isn’t capable of understanding if people are actually paying attention and what in particular they are engaging with. TVision takes audience measurement several steps further, introducing state-of-the-art technology that collects exactly what is going on in front of the screen.
Their computation software can be easily integrated into the graphic processing unit of any web camera. Once installed, their AI technique tracks how many people are watching, their attention level, even their emotions, all in real time. This is what Liu refers to as the special ingredient of TVision Insight’s technology: eyes-on-screen, passive data collection that accurately identifies viewing patterns in a way the world has never seen before. But what about privacy concerns? “Being transparent and maintaining audience privacy is an essential part of how we operate,” says Liu. Every TVision user voluntarily opts in and is compensated on a monthly basis. “We tell all panelists how the data will be used, do not store any images or videos, and all of the information gathered is processed on the local device in the living room.” In other words, the process is anonymous and personally identifiable information never leaves the home device.
TVision’s second-by-second innovations are targeted at three client segments. For the brand marketer, TVision delivers the ability to understand which commercial is the most effective in terms of audience engagement, without the cost and hassle of traditional focus groups. Media planners also benefit from TVision’s endeavors, as Liu explains: “At the end of the day, media planners and buyers are really buying audience attention,” says Liu. “We can overlay that information on top of Nielsen ratings to help them make better decisions to increase their ROI.” Finally, TV networks and OTT content providers like Hulu, Netflix and Amazon can utilize TVision’s data to sell commercials at higher prices because the quality and accuracy of the data provided by Liu and his team allows them to justify the value of the inventory. For example, while a late-night show might have a low rating, attention level may be quite high, which means the inventory can be sold at a higher price. Additionally, that second-by-second data that pinpoints increased attention level could be used by networks to understand what causes viewers to pay attention and how to maintain high-attention viewership or replicate the success of a particular show by implementing similar strategies into another show.
While Tvision is a relatively small-scale startup at this stage, their continued success and overwhelmingly positive feedback from consumers means that Liu and his team are looking to build on the breadth of their current partnerships. He identifies three industry categories for collaboration. Brands and agencies interested in purchasing data fall into the first group. But in addition to being a media measurement company, TVision is also a deep learning AI company, which means they are interested in collaborations with hardware companies or anyone with a significant interest in AI and its unique applications.
Finally, Liu mentions his interest in connecting with international media research firms who want to bring TVision technology abroad. “Today there are 76 countries around the world, all using the same technology to measure TV ratings. We want to build on our success and expand beyond the U.S. market into Europe, Asia and the rest of the world.” As consumer behavior continues to change, Liu and TVision are confident we’ve reached a point where the industry must embrace a new set of standards for understanding audience engagement. “We want to innovate the entire field,” says Liu. “As a relatively small MIT startup we might not be able to change everything, but we’re confident we can play a critical role by offering unique high-quality data to help inform better decision making, thereby making the entire ecosystem more effective.”