Jason Liebman, the CEO and co-founder of how-to video producer Howcast, has a lot to be proud of these days. His service is up to 1 billion videos streamed since launching three years ago. And the main site just got a redesign that launched on Wednesday. But get him talking, and what excites him is how the New York City-based company is using data to turn the business of creating reference videos into a smooth science.
The company, like a lot of Internet firms driven by search traffic, is obssessed with mining data to find the best way to select and package its content. But Liebman said Howcast is constantly on the hunt for more and more data to help inform every decision it makes. This is becoming critical now that the company is producing more than 1,000 videos a month for its library of 15,000 videos. And it’s also important because each video is like a long-term investment in an evergreen product that is expected to generate revenue over many years.
One of the biggest decisions is obvious: what videos should Howcast produce? The company ingests all the data it can on the way its videos and other videos perform on its own site and through its distribution partners such as YouTube (s goog). For example, it will look at the drop-off rate of viewers second by second. And it looks at what people are searching for on video sites and on search engines. It draws in as many other signals as it can to figure out what to create.
But since it’s trying to balance high-quality production values with popular content, Howcast agonizes over how well it can predict its chances of getting a positive return on its investment for each video. It will look at how well the video is expected to age and what it can expect to make through advertising. And it combines that with the data it has on production costs, down to which camera it uses, to determine what it can expect to make from a video over time and whether it’s worth it to greenlight a project. To present the content, Howcast also looks to data to determine what the best titles and thumbnails are and how to position them the videos for better viewing.
Liebman said this data-driven approach is helping drive more video viewing and time on Howcast. He said it’s a lesson for other older media companies, who are still just putting up shortened content online but aren’t thinking about how best to utilize their videos.
“Video content is exploding and we’re looking to data to intelligently plug that back into the business,” said Liebman. “Data allows us to scale with quality.”
Howcast has invested in its own data analytics tools to uncover more insights and unify disparate data coming in from various sources. And while it has just one data scientist on its staff of 30, all of the teams at Howcast work with the data on a regular basis to inform their decisions. It’s taken a while to really start harnessing the power of data but now, the work is paying off, said Tom Bender, VP of product development at Howcast.
“We’ve gotten to the point where the production model makes sense now and the quality and ROI is much better,” Bender said.
“Big data” is a popular buzz word these days but what you do with that data is what really matters. Howcast is showing that it makes sense to ingest as much data as you can, but more importantly, put the insights to work. This is obviously important for publishers like Demand Media (s dmd) and the Huffington Post (s aol), which rely on getting up quick content that people are looking to read. But there’s a lot of content owners that have a trove of stuff that can be better utilized if they look to the data.