Search is big these days, no doubt– big in ad dollars, big in the size of the segment players, and big in its influence as a doorway to the Web. But Blinkx hopes to make its mark on search by thinking small. To prove it, the San Francisco-based company recently rolled out a search engine that weighs only 1 megabyte but does the heavy lifting of bringing Web content to users without an explicit search.
The tool, called Blinkx Pico, is a Web download that sits on the call bar of any Windows-enabled document its user is looking at–not just Web pages but Word documents, Windows Media content and Outlook e-mail too. As you scan that content, Pico puts Blinkx’s contextual analysis to work and goes out looking for other content that relates to what the page seems to be about.
These found items can be other Web pages, but they can also be video clips, images, relevant podcasts, Wikipedia articles, blog posts, or other specialized content such as medical articles. When it finds one, Pico lights up the appropriate button on its little toolbar. Click on that button, and a list of the articles found—usually a maximum of six—appears, together with space for pay-per-click text ads at the bottom.
If what you’re reading has a lot of Web links and lights up all the buttons at once, you can hit a hotkey and display all Pico’s categories and findings in a single screen.
It’s a tiny plug-in search tool with a big idea behind it: that explicit, keyword-centered search only turns up a portion of the relevant content that the Internet has to offer.
“Pico fits in very much with what we’ve always been obsessed about: where search is today, and how it has to be improved,” says Suranga Chandratillake, the company’s founder and chief technology officer. “Search is a pretty disruptive process. If you’re reading e-mail or looking at Web pages, you may have a thought that you’d like to find out more about something or other. In that case, you basically have to down tools and switch over to a search engine, whether on a Web site or toolbar. You’ve got to think of some keywords, type those words in, navigate through lists of results and pick out the ones that make the most sense. It’s a whole extra process that you basically have to push yourself through.”
By contrast, Chandratillake says, Pico takes advantage of “the latent Web”, pushing relevant content to users rather than forcing them to go out and look for it.
Anyone can—and has—pushed content to users, and most often with irritating results; the key is to make sure it’s relevant. Pico’s key to relevance is the Blinkx search algorithm, which uses context to figure out what a Web page is about, or in this case, what about a page might be most interesting to you.
Most keyword-based search engines, given the query term “George Bush”, will pull up a number of pages with either “George” or “Bush” on them; they have to be told to treat the words as a Boolean phrase, either with an inserted “and” or with quote marks. But Pico is able to look at the whole context of the document you’re examining and to know that you only want data on George Bush, not George Washington or the Bushmen of the Kalahari Desert. It will serve up the Web pages it deems most relevant to that context, as spidered by its own searchbots.
You can help it along by highlighting the words it should use to bring relevant Web content to you—for example, if you were reading about the president playing cricket during a recent Pakistan road trip and desperately hoping that someone caught that moment on video. Simply highlight “cricket” and the video will be waiting for you on the Pico toolbar.
The tool has some other nice features. It can use Blinkx’ “Smart Folders” to aggregate information about subjects you’re probably interested in and automatically collect content from the Web. Users can also set alerts for new content that comes available after they begin working on a document or reading a Web page.
Undoubtedly cool, the application also sits in the middle of some other search applications that Blinkx offers, notably desktop search and a very robust speech-to-text search engine for Web video and audio content called Blinkx.tv that the company debuted last December. “Desktop search is about indexing a portion of content that people didn’t usually index until three or four years ago,” Chandratillake says. “Blinkx video also represents an area that had not been indexed before. Pico really brings both of those products together and offers a new way to access them.”
So how will Blinkx make money off Pico? With contextual advertising from the Miva ad network displayed below the Web results that the toolbar brings up in a shaded box marked “Sponsored Links”. Pico is actually the first Blinkx property to offer paid listings, although the company has talked about adding ads to its video searches. Chandratillake says Blinkx opted to go with the Miva network because its ads provide more contextual information than those of other networks and therefore seem to promise the best fit for search results.
Pico’s “search-less search” is an area in which it does not have much competition as yet, and that’s a familiar stance for Blinkx. In an industry dominated by Google, Yahoo! and Microsoft, the company needs to innovate where others are not looking just to stay competitive—or some would say, to stay alive.
“Search is getting a lot of attention because of the sheer revenues that Google and Yahoo! have been able to generate,” Chandratillake says. “But because they are so focused on those core businesses that do so well for them, they have turned a blind eye on vast areas of technology and potential that we think are pretty important and interesting.
“We think that a small, fast-moving team with smart technical people who are passionate about improving search can still show what’s possible in this area. We’ll keep our eye out for those areas that are being missed by the big guys. And every time we spot one, we’re going to jump on it and build something good. If you find the right application that works in the right way, it’s amazing how quickly you can become an important part of the map.”
Or how fast your company can get acquired by a big player, of course. Blinkx was heavily touted last summer as a possible purchase by Rupert Murdoch’s News Corp., which was rumored to be looking for a search engine that could handle its library of video content.
Chandratillake won’t comment on why the News Corp. buy never went through, but he isn’t in favor of simply building a company in order to be acquired. He points to all the start-ups that came into being when rumors began circulating that Google wanted to add a Web-based calendar function; now a year later, Google is apparently ready to add that calendar—which it built on its own—and Yahoo! and MSN are following suit, and most of those start-ups are likely to be left in the dust.
“That shows me the potential danger of simply ‘building to flip’,” he says. “It’s great if you can do it with just two guys, because selling for several million is still a lot of money at that level. But the reality is that it will probably take more money than that, you’ll probably need to get some funding, and when you do the sums it’s not a great deal.
“I instill in my guys that we don’t want to do that, that we’re going to be a big one. We’re undermining and changing the model that Google and Yahoo! have used, and that means we really need to do this ourselves to demonstrate its power.”