Visually Impaired Search

Sponsored search ads changed the face of search marketing. Popularity-based ranking systems forced a change in how pages are found. And blended search changed how we think of SEO.

Image search has officially gotten a facelift with the recently proposed introduction of VisualRank, but will the new Google image ranking system really change anything? Probably not right away, but the technology and methodology are certainly worth exploring.

The visual ranking concept is simple, yet infinitely complex at the same time. Generically speaking, people are either too stupid or lazy to label images appropriately, so a better mousetrap was in order.

Image Search = Clown Shoes

Much in the same way text labels were not enough to evaluate a Web site’s relevance and popularity, text-based image labels haven’t been adequate. It doesn’t take a Google engineer to realize image search won’t help you unless you really know how to manipulate search queries and you have lots of time on your hands.

The divide between people who build Web sites and those who seek to optimize content has always baffled me. Designers think in terms of brand and experience, yet seem to consistently ignore the requirements for building sites that make sense to search sites’ various crawling and indexing technologies.

The field of SEO wouldn’t exist without people exhibiting a lack of understanding of how to make information on the Web “search appropriate” or “search friendly.” Improving images and visual ties to search activity — thereby increasing relevance — is a good thing, particularly since image spam seems to dominate the search results in all but a precious few query categories.

Keep Optimizing?

Should you stop labeling your images? Not just yet, but you’d have to be foolish not to consider this potential change when “amping up” optimization and popularity gaining efforts.

Among the conclusions drawn in “PageRank for Product Image Search,” a paper by Yushi Jing and Shumeet Baluja, was the need for including “human-coded information,” one of the elements that made popularity-based ranking so effective.

In today’s environment, producing sound technology in the absence of a means to commercially exploit it just isn’t popular with the bean counters. Several research studies have indicated users spend twice (or more) the amount of time with images in search than text results.

Commercially Yours

The early days of search (before the ads started appearing) lacked a defined commercial exploitation method. When the folks at Idealab first began tying ads to keyword search, everybody poked fun at the little company that could with the “less sexy than graphic ads” format.

That is, of course, until the bottom fell out of the online advertising market and advertisers flocked to those boring little text ads.

The connection to blended, product, and other commercially exploitable queries with improving image search should be plain as day at this point. Though few researchers discuss the commercial impact of their research, improved image search will accomplish several goals for Google.

A “VisualRank” system will help move image search away from an easy way to track down Miley Cyrus photos to something that can be an efficient means for locating relevant and commercially valid information. Of course, more relevant images means better search for everyone, so searchers win while Google creates yet another way to help build its financial empire. That’s what we in the business call a smart move, Mr. and Mrs. Shareholder.