Bringing Web Video Into the World of Contextual Advertising

Text and display ads on the Web are a $17 billion business. The bulk of that revenue comes from “contextual” ads targeted to each page’s text content, like the Google AdWords ads for sleep medications that might turn up in the margin of a blog entry about insomnia. But while text is still king on the Web, audio and video files make up an increasing portion of its content, and text-based ad-placement systems have a hard time identifying this content and matching it with relevant advertisements. That means a lot of Web content is going unmonetized.

Cambridge startup EveryZing is out to close that gap. A spinoff of renowned government contractor BBN Technologies, the company uses advanced speech-to-text software developed at BBN to make rough text transcriptions of audio and video files. Its system analyzes these transcriptions for key words and uses those terms to supplement the existing titles, tags, and other metadata for each file. That means ad-placement software and search engines can effectively “look inside” multimedia content and deal with it more appropriately.

“It would be ludicrous to search just the titles and tags of a text document, yet that’s the state of the art with video,” says Tom Wilde, who joined EveryZing as CEO four months ago. “We take multimedia content, create a text output of near-transcript quality, run our natural language processing software against that, extract the key concepts, and put them into a friendly package. Now you’ve made all of this multimedia content discoverable to search engines and brought it squarely into the world of contextual targeting.”

The startup, which was launched in January 2006 under the name Podzinger, is the first spinoff from BBN since Verizon sold the lab to Cambridge-based private-equity firm General Catalyst and Palo Alto venture-capital firm Accel Partners in 2004. Its first product was a Web-based engine for searching the content of podcasts. But this June, General Catalyst, Accel, and Fairhaven Capital committed a second funding round of $10 million, and the company changed its name to EveryZing.

The podcast search service “was a good proof of concept but not a great business,” says Wilde. “Podcasts are very useful containers for content, but they are not growing at nearly the rate of consumer Internet video.”

Wilde predicts that within five years, Internet users will spend as much time watching video as they do using e-mail and search engines. Yet analysts say video-related ads on the Net will bring in only $4 billion in revenues by 2011, compared to $30 billion for text-related ads. “That gives you a sense of how much revenue is being left on the table,” says Wilde. He hopes content owners will hire EveryZing to help wring more advertising proceeds out of their multimedia content, processing not just their current content but their archives. “If you make content more discoverable, you are by definition creating more inventory [to sell ads against]. The effect is multiplicative.”

Already, Wilde says, local news portal is using EveryZing to add multimedia results to its standard search results, and information service Factiva is using the system to make more content from its parent company, Dow Jones, available to subscribers. The company also operates a multimedia portal at, featuring recent and popular videos and podcasts on news, sports, health, and the like. But Wilde calls the portal “a good demonstration—it does drive some traffic to our partners, but it’s really our sandbox rather than our primary business. It’s where we try out new features before we bring them to our partners.”

Wade Roush is a freelance science and technology journalist and the producer and host of the podcast Soonish. Follow @soonishpodcast

Trending on Xconomy