Practical Ecommerce Offers A Different Way To Find, Sell Products

Munjal Shah is the CEO and co-founder of Riya, the company that owns and operates, the photo-likeness search engine.

PeC: What differentiates Like from the many other search engines available?

Shah: “Like is the only search engine that lets you search by photo. With every other search engine, you have to specify what you are looking for in words. The usual query is always linguistic.

“However, some things are very hard to describe in an unambiguous way. Describe a rug pattern to me: It’s likely that: a) You probably don’t have the right vocabulary; b) Even if you did, you wouldn’t likely type a lengthy description into a search engine; and c) Even if you were able to do so, it’s likely that your attempt to describe the rug would be nothing like mine.

“At Like, we call submitting your query by clicking on a photo a visual search. There is no other visual search engine on the web today.”

PeC: How do consumers benefit from using a visual search engine such as Like?

Shah: “For the first time, consumers don’t have to convert what they are looking for into text. The greatest benefit will be experienced when looking for items that are largely aesthetic and hence hard to describe in words. This is true for things like clothing, vases, carpets, flowers and landscapes.”

PeC: Describe the various ways consumers can search for products at the site.

Shah: “Today you can search our index of items. In the future, you will be able to upload or email a photo — including ones taken with a camera phone — then use our browser toolbar to search for matches for that image.”

PeC: As a business owner, how can I sell products on Like?

Shah: “You just need to email us on the link at the bottom of our site which says, “Add your items to Like.” We will set up a cost-per-click or cost-per-action relationship with you.”

PeC: If my products are identical to another merchant’s products, will the least expensive product be shown as the only sales option? For instance, if Bealls, and Nordstrom each offer the same men’s dress shirt, does a consumer see all three options on screen, or just the least expensive one? Is this an environment where the lowest-priced item gets the more prominent position?

Shah: “Nope. The likeness-search algorithm shows items by similarity rank. If two merchants have the item, both will show up in rankings that reflect how similar each photo is to the one uploaded or used for the customer’s search.”

PeC: It appears that consumers check out at the retailer’s site, not Like’s site. If that’s the case, how do you make your money?

Shah: “We implement a tracking system with the retailer, or use an affiliate manager like Commission Junction, so we know if the person bought something. On average, we get about 15 percent of sales but this varies. There is no listing fee to add your items to Like.”

PeC: Like’s initial launch carried a narrow group of products. Will that expand?

Shah: “We recently expanded the list significantly to including many clothing items [belts, hats, gloves, blouses, blazers, coats] and some home items like rugs. We will continue to add more and more items.”

PeC: Are the products found on Like new or used?

Shah: “Today all are new, but we may add eBay — in which case it will be mixed. We will ask that the items for sale are clearly delineated [as to which is which].”

Practical Ecommerce

Practical Ecommerce

Bio   •   RSS Feed


Sign up for our email newsletter

  1. Legacy User March 16, 2007 Reply

    Oh Boy, is just awesome. I love it !
    Nothing like it ;)

    — *Jenny Paul*

  2. Legacy User March 16, 2007 Reply

    Munjal is wrong when he says there is no other visual search on the web. There is another very strong technology on the web which does visual search and browse. The company is called PIXSTA ( and is London based. I recently met them and was very impressed with the technology they own.

    They have several clients running the application by now. They have also put up two demo sites for shoes, called ChezImelda (, and for bags (

    — *PixLover*

  3. Legacy User March 20, 2007 Reply

    PixLover, where are Pixsta's search and results shown? It appears between the lack of information on its site and your post that it is not a search engine for the public to shop, but a back-end for businesses to implement.

    — *Leffrey*

  4. Legacy User March 20, 2007 Reply

    I think is very unique and and, as a photographer, I like the simplicity of it. I'm not sure if they only list big sites with a lot of inventory and not small ones like mine.

    — *hudsonrivertreasures*

  5. Legacy User June 6, 2007 Reply

    I completely agree with PixLover here that Munjal is wrong when he says there is no other visual search on the web. As far as I am aware that there are tons of peer reviewed published papers on the subject, with new algorithms emerging all the time from the literatures. I can point out perhaps 30 publications or more of which I have come across. There are other commercial vendors who are doing this as well. Image recognition (image retrieval) is a branch of engineering (or computing) called Computer Vision.

    Well , how does anyone think that those F-14 , F-15 , F-117 stealth operate? Yep, they operate via image recognition when tracking their ground targets (mobile & static). The fighter plane scan the ground in its wide view angle, then try to matched the image pattern that it receives with the image patterns in its database. Once a match is pinpointed by the computer, plus data gathered from GPS target locator, then the target tracker is locked on to that image on the ground, where the onboard computer has just matched (high probability tolerance of more than 95%). That is why those fighters can attack at night because pilots can't see at dark, but infra-red plus image recognition are used by the onboard computer to track the target. If the image recognition are not available , then it would be bloody hard to attack at night , since the pilots can't see.

    Microsoft is developing and image retrieval (image recognition ) systems.

    "Text-Search Tricks Speak Volumes in Image Search"

    — *Falafulu Fisi*

  6. Legacy User June 6, 2007 Reply

    If any developer is interested in these sort of algorithms, you can find tons of peer review papers on the subject of Computer Vision (image analysis, image & object recognition) here:

    I have no doubt that "Riya" researchers are reading those publications. I do have a pretty good idea of what algorithms that "Riya" are (or is) using. There are some standard ones, and there are advanced ones. Advanced ones have lower misclassification error. This is something that I have thought of to develop a medical imaging decision support systems to help physicians in their diagnosis. There are already commercial image recognition retrieval systems available today at hospitals & clinics. They do help doctors give them a quick hint of what's the likely disease from X-rays, or MRI in front of him. Instead of the physician having to do the search from the medical imaging database, it just scan the image in , then the retrieval system looks for say, the top 10 similar images that had been recorded from the database. The physicians , then pull those retrieved 10 images, and look what were the diagnosis, treatments or prescriptions given to those patients with similar Xray images from the past.

    — *Falafulu Fisi*