Visual search and shopping collide

Commentary: Like.com enhances buying experience
By Bambi Francisco, MarketWatch
Last update: 12:01 a.m. EST Nov. 9, 2006

http://www.marketwatch.com/news/story/story.aspx?guid=%7B3F0564EB-5993-42EE-859A-7BE5D88FF2D9%7D

SAN FRANCISCO (MarketWatch) -- Artificial intelligence that can detect faces and images is typically used for security systems. Now, similar technology is being applied to the most practical situations: How to find that piece of jewelry or shoe that's hard to describe.

Like.com, the first visual search engine for products, launched this week. It's a service that helps shoppers find products based on images, rather than rely on SKUs (stock keeping units). It's akin to showing a photo to a salesperson at a store and saying, "Show me all the products that look like this." It sure beats poor attempts at describing the ideal item to a salesperson.
Basically, Like.com is the next evolution of comparison-shopping engines, such as Google's Froogle, eBay's, Shopping.com, EW Scripps' Shopzilla or Become.com and the next evolution in image search offered by the portals and search engines, Google, Yahoo, Microsoft's Live and InterActiveCorp's Ask.com.

Like.com is a product offering from Riya.com, a startup whose CTO and co-founder, Burak Goturk, specialized in computer vision technology and has filed for two dozen facial-recognition patents.

"For the first time you can say, 'I want something to look like that,'" said Riya's other co-founder, Munjal Shah, who is also CEO of the two-year-old company. "Like.com looks at all these other items and compares, shape, color and texture."

For now, Like.com helps you find jewelry, watches, handbags, and shoes. The engine crawls the Web for images, and will soon crawl eBay's merchandise as well. Additionally, Riya has some merchant partner deals. All told, Riya compares 2 million products offered by 200 merchants, including Amazon.com.

The service works this way. The homepage has photos of celebrities. You can highlight their shoes or jewelry and tell the Like.com search engine to find a similar item.

In the boot example, Tyra Banks is shown wearing some slick brown boots. You'll note the "likeness search" box in the image I've captured. A shopper would then click that box to conduct a "likeness" search across Like.com's inventory. The search found 2,207 similar boots. The results are presented in the most similar to least similar order. But even the last page, which supposedly showcases the least similar boots, offered up pretty good matches.

If I buy a product from one of the merchants on Like.com's site, Riya gets paid about 10% of the order. Riya also has some deals whereby the company gets paid a buck or so for leads. Shah said that the four categories - jewelry, shoes, handbags and watches - represent $15 billion in sales in the U.S., with $4 billion worth of shoes sold online.

Search evolution

It's about time we started seeing some advancement in visual search. As Shah sees it, none of the big search engines, such as Google, Yahoo, Microsoft's Live, and InterActiveCorp's Ask.com have made any significant improvements to searching images.

Like.com uses "similarity technology," said Shah. Basically, the technology looks insides images and calculates the pixels and converts the picture into a mathematical equation. The math equation - which represents the photo - is what is used to make comparisons. Similar math equations end up on the same page.

Indeed, most search engines organize images based on the text used to describe the image. It's useful if you're searching for a particular branded product, say Nike running shoes or Callaway Big Bertha golf clubs. It's not so helpful when you're indifferent toward the designer but care only about the shape, pattern or color.

Soon enough Like.com will expand to household goods, such as rugs, and clothing, which I'm extremely excited about since in the past 12 months or so, I've wanted to find this particular white-collared blouse I saw my neighbor wear last year. My neighbor, Heather, said she bought it from a boutique in Paris, but didn't have the name. I'm sure we all have had similar experiences.

At the time, I tried my luck with the comparison-shopping engines. I wasn't successful. I shared my white-blouse search story with Lorrie Norrington, who at the time was head of eBay's comparison-shopping service Shopping.com. Norrington said a service like that would probably be available at Shopping.com soon. No such luck. I guess eBay's got too many other challenges, such as figuring out how to capitalize on the growth opportunities in China.

Of course, I'm not able to find that blouse on Riya just yet. But in about a month, I just might. That is, of course, if a celebrity ended up wearing such a blouse. Soon enough, however, I'm sure I'll be able to upload a photo of that blouse and then conduct a "likeness" search using Riya.


Why startups iterate

I'm always fascinated with the stories that founders share about their original business model and product, and the iterations on top of it.

Riya's first product was a visual search engine that helped people organize photos by essentially sorting them based on images. For example, photos of bridges would be filed in one category. Shah said that the Riya visual search service is still around and has about 10 million photos uploaded. The service, however, isn't a priority any longer, Shah admitted, conceding that he and his team couldn't figure out how to make money off such a service.

Basically, Riya's initial customers preferred a search engine to make sense of the Web - a large fast-growing, disorganized library of digital photos - rather than their own smaller, also growing, very-much disorganized library of digital photos.

I never thought organizing photos would be that big of a market, actually. After all, organizing photos isn't exactly a priority for most people. Many people - like me - didn't do it in the offline world; why would they do it in the online world? And, for the few who do want to organize their photos, they typically wouldn't organize photos based on every bridge they stood in front of, or every photo with a mountain background. People typically categorize their photos based on event, occasion or year. Moreover, homegrown pictures have become commodities. If you lost one, you could just take another.

To that end, Riya's setting its site on something far more useful and in demand. Like.com is addressing a real pain point.

I remember in 2004, when it was painful to share my digital videos with others. In order to share them, I had to make a number of CDs. Two years later, one video of my nephew Bubba snowboarding has been shared more than 500 times on YouTube.

Since last year, I wanted a visual comparison shopping engine to find that white blouse. Maybe next year, I'll find it. In fact, maybe by next Christmas, I'll be doing most of my shopping that way.

Editor's note: Next week, I'll be looking at video search. I'll be touching on the differences among Nexidia, CastTV, Dabble, Pixsy, Blinkx, Optevi.net, Quintura and Truveo. I'm also on the lookout for other video search engines.

Sound off: Who's a winner in the video search world? And, what do you think of Riya? Comment on Bambi Francisco's blog.