In May, more than 180 million U.S. online users watched 36.6 billion videos, according to comScore Inc. That's hours of silly feline antics, TV shows and music videos that captured the precious attention of viewers.
A team from Carnegie Mellon University is in the process of developing a system to help online channels capture and keep more of those viewers as they poke around for something fun to see.
The startup company, neonlabs, eventually could rapidly process tens of thousands of videos for a client and spit out recommendations on the best thumbnail image to use for each one -- the right image to appeal to people's unconscious visual perceptions.
Eventually the company could charge a fraction of a penny per thumbnail recommendation, with the sheer volume of videos being posted on the Internet creating a market that would generate enough revenue to support a viable business.
This isn't exactly where the project originally began.
Initially, researchers were looking at applying how the brain takes visual information and makes choices, specifically as it relates to the design of packages.
Some preferences seem to be intrinsic to being human, said Michael J. Tarr, professor of cognitive neuroscience and co-director of CMU's Center for the Neural Basis of Cognition. People, for example, tend to like shiny things over spiky, rough ones. Other visual reactions to common objects are specific to an individual's experience or influenced by factors such as age, interests and culture. "It may also be that context matters a lot," said Mr. Tarr.
But while those may play a role in shoppers' decision on which teapot or box of cereal to buy, companies already have other kinds of systems to test the best designs and the best marketing campaigns for those kinds of physical products.
In part by using a $50,000 grant from a new National Science Foundation program "Innovation Corps" meant to help commercialize scientific innovations, the team has worked to figure out other ways to use the research.
Companies like YouTube and Hulu were looking for ways to keep users on their sites longer, said Sophie Lebrecht, a postdoctoral fellow with CMU's Tepper School of Business and the neural cognition center.
There's money in keeping an audience and having opportunities to share those viewers with marketers. Tucked in and around all the videos Americans watched in May, they saw more than 10 billion video ads, reported Reston, Va., digital tracking firm comScore.
Online video websites may put up thousands of videos daily and they don't have time to field test each one. Thumbnail images -- the images that viewers use to select a video -- typically represent one frame of hundreds or thousands of possibilities from, say, the latest episode of "The Today Show" or the Modern Mom channel.
"People are making rapid decisions on what to post," said Mr. Tarr.
That led to the idea of building a service that can process a selection of videos on an as-needed basis and charging per thumbnail generated. Under that strategy, neonlabs will build a database that combines information on people's innate preferences as well as data from past behavior. It could also allow fine-tuning for, say, choosing the image most likely to appeal to 25-year-old males, if a client is targeting that audience.
"We're leveraging all the data we can collect beforehand," said Mr. Tarr. "It only works because people are relatively consistent."
The recommended thumbnails only need to improve viewership slightly on the current system to pay off. Even 1 percent or 2 percent better would be a tremendous advantage across the huge number of videos, said Mr. Tarr.
Sites might eventually consider offering the neonlabs process to those posting videos who are willing to pay a little extra for a recommendation to increase click-throughs. "It could be a value-added service that someone like a YouTube offers," he said.
Individuals putting up cute cat videos and scenes from the school play aren't likely to be the audience interested, said Ms. Lebrecht. "It's all the videos that they would monetize."
Other members of the neonlabs team are Babs Carryer, embedded entrepreneur at CMU's Project Olympus, and Thomas Kubilius, president of South Side-based Bright Innovation.
In addition to the National Science Foundation assist, the team has had help from Innovation Works, a South Oakland-based tech incubator, and CMU's Center for Technology Transfer and Enterprise Creation.
At the moment, they are continuing to talk with potential clients about how to craft the right service. Proving they have a viable product also will likely require getting data from companies and demonstrating how their recommendations would work.
The startup hasn't set a deadline for taking on commercial projects. "You often can't predict. It depends on the marketplace," said Mr. Tarr.