Guest: Munjal Shah - Riya
Host: Robert Scoble - The Scoble Show

Robert Scoble - The Scoble Show
So, what's up?

Munjal Shah - Riya
It's good to see you.

Robert Scoble - The Scoble Show
So, who are you?

Munjal Shah - Riya
So, I'm Munjal Shah. I'm the CEO of Riya.

Robert Scoble - The Scoble Show
Riya?

Munjal Shah - Riya
Yeah.

Robert Scoble - The Scoble Show
The photo -- tell me a little bit about what Riya use to be, because maybe there are a few people out there who didn't see it.

Munjal Shah - Riya
So the original idea with Riya was to use face recognition to make your personal photo searchable. So we'd use face certification to go in, finding out, hey, here's Robert and here's Maryam, and then let you search, and even though you hadn't explicitly labeled or tagged all those photographs, it could find them.

Robert Scoble - The Scoble Show
Yeah, it was really neat because we could put a 1000 photos in, and as soon you found your kid's photo or one of your kid's photo, you mark him and then it goes through and finds all the other photos with the same face basically.

Munjal Shah - Riya
Yeah, I think to date now we've had about ten million photos uploaded to the service. So, Flicker was ten million photos in two years, so this happened in about seven months.

Robert Scoble - The Scoble Show
When it first came out it was pretty rough, it didn't work all the time, it worked like 70% of the time or 80% of the time. It felt like that...

Munjal Shah - Riya
The algorithmically it hasn't improved that dramatically. You have reached the limits of face recognition technology.

Robert Scoble - The Scoble Show
Let me just look at the camera here, and just make sure.

Munjal Shah - Riya
That's about as good as its kind of -- it will continue to improve a little bit, but its not going to leap from 70% to 90% anytime soon.

Robert Scoble - The Scoble Show
Yeah.

Munjal Shah - Riya
Yeah.

Robert Scoble - The Scoble Show
So how does the technology work? If I see a photo of you, what is it looking for, is it looking for -- what does it look on your face for?

Munjal Shah - Riya
Actually we don't know, is the short answer. What these are, these are learning systems, and you feed them lots of examples of, this is a face, this is a face, this is not a face, this is the right guy, this is the wrong guy, and these are different variants of learning systems you can put in there on your network, a hidden markup model, whatever, and it probabilistically just learns, and you don't know what it focused on or what it didn't focus on.

One of the criticisms of these learning systems is, you can't go in and tweak them so of speak, you tweak them by building a nose learning system, an eye learning system, and a mouth learning, but you still don't know exactly what it learned, and you combined them, so then you can say, add more nose and add less mouth. But each individual unit is a black box.

Robert Scoble - The Scoble Show
Yeah. You told me one time, it even looks at your clothes.

Munjal Shah - Riya
Yeah, so one of the units we built with cloth learning system, and it would actually look at your clothes and then it would say, if there's another picture of you, I couldn't quite recognize you, because your face was turned. It would say, you know, I can't quite tell who that is, but I saw a guy two minutes earlier in the same album that had exact same shirt on, nobody else has that shirt on, and we call that contextual recognition. Its kind of the ability to look at more than -- so the core inside underneath Riya was that we looked at everything not just your face.

Robert Scoble - The Scoble Show
Right.

Munjal Shah - Riya
Okay.

Robert Scoble - The Scoble Show
Now, that was Riya1.0, right?

Munjal Shah - Riya
Yeah.

Robert Scoble - The Scoble Show
You're just about to announce a new initiative, so what's the new initiative?

Munjal Shah - Riya
So we're announcing on Wednesday, and we'll be live the Like.com.

Robert Scoble - The Scoble Show
Light?

Munjal Shah - Riya
Like.

Robert Scoble - The Scoble Show
Like.com

Munjal Shah - Riya
Like.

Robert Scoble - The Scoble Show
How did you get that domain? How much VC money did you spend?

Munjal Shah - Riya
Not as much as you might think.

Robert Scoble - The Scoble Show
Right.

Munjal Shah - Riya
Yeah, it wasn't cheap, but it wasn't as much as you might think. There's a long story behind it, which I don't get it too much, but it involved us climbing a few fences and delivering a bottle of wine to a guy, but...

Robert Scoble - The Scoble Show
I got to hear that story. We got time.

Munjal Shah - Riya
All right.

Robert Scoble - The Scoble Show
We've got an hour of tape.

Munjal Shah - Riya
So, we were trying to get this domain name, and I was at the visual search engine that helps you find things that look similar.

Robert Scoble - The Scoble Show
Right.

Munjal Shah - Riya
What a great name for it, right? The guy who had it wasn't returning any of his email. We looked him up on Whois, and couldn't find him, the address just didn't seem to work and so we finally just said, we'll go in there, we'll just going to go his house, because the email just wasn't working.

Then there was a gated fence, there was a gated community in San Francisco where he lived and, "Oh my God! What do we do now?" So, we hopped the fence and then he wasn't there. So we went back, bought a really nice bottle of wine and wrote him a card and left it there, and so I hope this is his house, because lots of times the address can be outdated too, and he called us the next day and we ended up buying it from him.

Robert Scoble - The Scoble Show
Did he even have a concept he wanted to sell it?

Munjal Shah - Riya
No, he had sold the name in 1999 to somebody on an installment plan, and the start up never paid and he never got his money. So he was just like, oh, that's what you often find here and so we bought it.

Robert Scoble - The Scoble Show
You're not going to tell me how much you paid for it?

Munjal Shah - Riya
It's $100,000.

Robert Scoble - The Scoble Show
That's not too bad.

Munjal Shah - Riya
Yeah, better.com wanted 450, and the guy yeah--it's not cheap but...

Robert Scoble - The Scoble Show
Yeah, one word domain names go for a lot of money, and they're worth a lot of money, all right, because it's easy to remember and it's easy to tell your friends, hey, go to Like.com instead of Whois.

Munjal Shah - Riya
Any shop at the top, people search for, I'm looking for things like, Prada shoes, if you don't like you will get some of those in that.

Robert Scoble - The Scoble Show
Excellent.

Munjal Shah - Riya
Yeah

Robert Scoble - The Scoble Show
Well, so what is Like.com?

Munjal Shah - Riya
So Like's a visual search engine. It's the first search engine on the web where you can submit a photo as your query. Think about that, every other search engine today even Image Search, Google Images, Yahoo Images, Live.com images.

Robert Scoble - The Scoble Show
I can take a cell phone into a grocery store, take a picture of a box of Wheaties and I can say find me other boxes of Wheaties.

Munjal Shah - Riya
Find me similar looking items, anything. Now, its constrained in our release, it won't do all products and all categories of things. So, I'll talk more about that, but at a high level.

Robert Scoble - The Scoble Show
Which one will work best for it?

Munjal Shah - Riya
We launched it specifically for products and specifically for aesthetic products. So we focused on shirts and jewelry and shoes and handbags, because those are the things you can't describe in text. You don't need this to find even Wheaties or an iPod, I mean text works just fine. You find what you're looking for. It's the things that have designs, China patterns, how do you describe that China pattern in text?

You don't, right, and so -- or if you do, it's a paragraph, and you're not typing a paragraph into a search box. So when we started thinking about this, both as a technology and as a business, we're like, okay, apply it where it's needed most. Its needed most in what we called aesthetic items; design oriented things.

Robert Scoble - The Scoble Show
So I could be at Macy's, see a -- my wife could see a cool handbag and I'm saying, "I'm not going to pay 500 bucks for that." I take a picture of it, and it will go out and look for other hand bags that are like that bag?

Munjal Shah - Riya
Exactly. So the first piece of this, the first implementation of Like.com is, you come in, you do a search by text if you want, or we have a bunch of celebrity pictures, and you can say I like Kate Moss's watch, you click on it, it does then, what we call a, 'likeness search'. So it's basically a search by appearance, and it find things that look similar. In about a month after launch, we'll enable the upload feature, we haven't enabled that yet. We also enable a tool bar or a browser extension, so you could right click on any photo and submit it for a 'likeness search', as you're surfing a web. So...

Robert Scoble - The Scoble Show
Or you have a mobile cell phones and click on it?

Munjal Shah - Riya
Cell phone version (Voice Overlap)

Robert Scoble - The Scoble Show
Because this is the killer app for mobile phones.

Munjal Shah - Riya
It probably is, especially for aesthetic items, right. So we really focused down on it, so we said, all right, what are the things hard to describe and how can we help you find those things. If you look at traditional comparison shopping, even just search on the web, these items really -- try to actually go and search for red strappy shoes on Shopping.com

Robert Scoble - The Scoble Show
Yeah.

Munjal Shah - Riya
Like, three of them are red out of thirty, four of them are strappy, and not the same ones that are red. I mean searching for these aesthetic items today in text doesn't work. So we said, let's make it work by moving the paradigm from text to an image based system.

Robert Scoble - The Scoble Show
This has some deep implications for all sorts of mobile stuff, because if I'm -- like in San Francisco, you're at a bus stop, right, and you see on the bus stop an ad for a movie, and I'd go, oh, I'd like to see that movie, but you have no idea where it's playing, because its just -- it has like, some movie like Borat, right. It has an ad for Borat, but it doesn't tell you where that movie is playing, what time it is playing and I'd love to be able to take a picture of that sign and have your search engine...

Munjal Shah - Riya
We maybe able to build stuff like that. We're focused more right now on products.

Robert Scoble - The Scoble Show
Well, that's where the money is, right? Because you want to be -- are you making money out of transactions?

Munjal Shah - Riya
To the rep its Riya 1.0, we can figure how out to make money, and so one of our goals was -- we got two pieces of feedback out of that that let us to really be focused. One was, okay, this is really neat, I mean there wasn't a single person to use it, it wasn't like, wow, okay. It doesn't work perfect, but I'm just amazed I can even do it, right. The second part -- but their main pieces of feedback were, I don't search my own photos everyday, but I do search the web everyday. Can you find some way to look inside the photo and do something smart for the web?

So that was one, and then the second feedback we had ourselves, which was, why don't you try to make any money out of this versus shoes, handbags, jewelry, all aesthetic items together are $30 billion, shoes is four billion, online today, on the web, Zaples is 600 million of that four billion. People are buying shoes on the web. You think, oh, it's really hard, I don't know, I can't try it on. Most of these retailers have solved it by just saying, free shipping both ways, like at Zaples, its free shipping both way. So I know people who buy ten shoes from Zaples, try on five, send back five, whichever ones they don't like. So, they've just solved this web purchasing issue through shipping, right through shopping. Very simple solution actually, right?

Robert Scoble - The Scoble Show
Of all the handbags that are on the web, how many do you think you're really going to find? What percentage?

Munjal Shah - Riya
We think we have most of the unique skews. How many unique shoes do you really think are being sold that anyone sees in manufacturing in the entire country. It's like...

Robert Scoble - The Scoble Show
A few hundred, thousands.

Munjal Shah - Riya
Yeah, it's in that order of magnitude. So, we're launching with about 200 merchants and 200,000 unique skews, but we believe we've got really good coverage of all the things available for sale in a given season. Even at eBay, not immediately launched, but within a few weeks, because that will get us last season's and the prior season's and some of the other stuff. Finally, you can just point us out and say, "Show me things that look similar to that based on appearance. It'll look at shape, it'll look at texture, and it'll look at color, and it'll look at 20 other factors that are details, like the sheen on your shoe, it's shiny.

If we see huge color gradient transitions in the middle of what we think is the middle of the shoe, it'll say, okay, probably glossy or shiny in some way, and will look at all those different factors, and those all get converted into the visual signature for the shoe, which actually by the way is bigger. We're actually encoding more information in the shoe. It's about a 10k vector that we use to encode it, for each and every shoe, versus for faces, where it was about 3k vector. So, we're actually extracting more information to make the search better. Then what we do is we compare all those 10k vectors real time.

So our technology for this is really two parts; one, how do you compare two of these vectors and say they're the same. Two, how do you do it in sub-second time, because you have to compare 10,000 numbers just to say if two things are the same, and then if you have even ten million of these items or a 100 million, I mean, so we're launching with a 1000 servers. The Like.com search engine is -- and to be more accurate, since folks listening to this will know better, its really a 1000 cores that we're launching it with, so its 250 servers, four core machines, a thousand cores with 16Gb of RAM in each machine.

Robert Scoble - The Scoble Show
Excellent. So, Intel and AMB probably love you to death.

Munjal Shah - Riya
Oh yeah, we're their new best friends. We're waiting for 16 cores to come out, because this is all about process.

Robert Scoble - The Scoble Show
Or an Intel showed off a prototype of an 80 core machine, right.

Munjal Shah - Riya
Yeah. The powerful playing straight, right, so two core machine is 110% of the powerful plane (ph) of a one core or something like that. So, this data center cost is all driven by power these days, so we -- you couldn't have done Like.com four years ago, before bandwidth cost down, merchant selling online, selling enough shoes and multi-core processors.

Robert Scoble - The Scoble Show
So these 250 machines, so its four cores each. When my request comes in, what happens in that system?

Munjal Shah - Riya
It figures out what you've searched on: shoe, handbag, whatever, and then it round robins to one of the servers, but the thing is, its so computationally intensive to compare even with our optimizations and our algorithms, that fundamentally each server can only handle so much, so you have to have a ton just so that you can serve so many people per second.

Robert Scoble - The Scoble Show
Wow.

Munjal Shah - Riya
I'll show you some of the detailed features we have. We actually let you pull out a feature of a hand bag, like a buckle and say, I really likd this buckle, find me other handbags that have the same buckle, and that's even more computationally intensive, because we have to now take a section of this visual signature, and compare it to the sections of the visual signatures of all the other items.

Robert Scoble - The Scoble Show
Right.

Munjal Shah - Riya
It's just the whole different level of technology. So, it's a very simple app on the surface. It's like click, find similar, it seems very -- almost aesthetic, and People magazine like and just kind of put them through and here's what the stars wear and it's just such a simple thing. Underneath it, technologically, it took us eight month since we launched Riya, 14 PhDs just to develop the core algorithms plus another 20 engineers to make this thing return in sub-second time, plus we had done a lot of the research before, just in our general (Voice Overlap)

Robert Scoble - The Scoble Show
So, you're not coding this in Ruby on Rails though, right?

Munjal Shah - Riya
No, no, no, it's all Java.

Robert Scoble - The Scoble Show
It's all Java.

Munjal Shah - Riya
It's all Java on the back end, and actually its even C++ or Linux. It's all C++ for the core algorithms. We couldn't, every time we tried to convert those math algorithms to Java, it ran six times slower. Somehow, the C++ math libraries are so optimized, they're just way faster. So, we use (Voice Overlap)

Robert Scoble - The Scoble Show
They're closer to the middle.

Munjal Shah - Riya
No.

Robert Scoble - The Scoble Show
There's less stuff between the third.

Munjal Shah - Riya
I mean these compilers are older, and they've been optimized better and the math libraries have really been tweaked. Basically, processing matrix math, matrix math versus linear algebra, which is basically what this stuff is.

Robert Scoble - The Scoble Show
Very cool. What else do we need to know? So, does it cost anything to use this system?

Munjal Shah - Riya
It's totally free. We get paid when you click through and buy the item.

Robert Scoble - The Scoble Show
Okay.

Munjal Shah - Riya
So that's our business model. It's...

Robert Scoble - The Scoble Show
So you're getting a referral fee like I do when you click on my book, on my blog and you buy a book from Amazon. It's actually pays me a dollar, $2.

Munjal Shah - Riya
Or with some cases we get CPC, we get it per click. So, one way or the other, if you like it, it's great. It's...

Robert Scoble - The Scoble Show
So you're playing Google ad kind of things on your pages as well or?

Munjal Shah - Riya
No, no, just the actual item, when you click on it, it goes straight through. We'll look at adding more categories, we'll look at landscape pictures, and leading it to Posters.com or thing -- I mean we'll allow more and more, but we really wanted to just get this beachhead and make it work. Like.com, let me visually search, and let me do it for products first. We'll expand beyond there once this comes out, and everybody goes, yeah, that was good, it really worked, I'm really buying stuff, or at least I found it useful, and I did that, and then we'll enable all the other ways you can search, mobile phone, et cetera.

Robert Scoble - The Scoble Show
How much does the image quality matter to you?

Munjal Shah - Riya
It matters a lot. So when we wrote our crawlers to go in, in there...

Robert Scoble - The Scoble Show
And most cell phones are, particularly when you're inside a store with bad lighting.

Munjal Shah - Riya
It just depends on how close you take it. So, we may give you some pretty strict guidelines like, take it this way, lay flat, don't take on the mannequin, because the shape might impact it, and things like that.

Robert Scoble - The Scoble Show
Okay. Does giving it multiple images help?

Munjal Shah - Riya
In some cases, so one of the key things we did is, when we indexed the web, we went for the highest rez version, because our detail search allows you to highlight a certain pattern and say, find me that pattern. Well, when we first did it, we used the thumbnail. When we used the thumbnails, in some cases the box you drew had like three pixels in it.

We're like, wow, what's wrong with our algorithm, it's not working. It works, it just didn't have pixels to work with and it wasn't going to do anything. So, we ended up getting the highest rez from all the merchants we end up calling. We clicked through, we got the highest rez --- it had multi views. We took the side view and the front view, if they had multi colors, we took every color. So we really kind of crawled it to get us the best data, so that the indexing scheme and the visual signature...

Robert Scoble - The Scoble Show
How much disk space does that index take up?

Munjal Shah - Riya
It's not that much. Even 10 million photos is not that much space, but its 10K per photo. So, it's bigger than the thumbnail, actually in terms of the actual amount of space, but we wrote it pretty efficiently. So our memory footprint is as low as we could kind of get.

Robert Scoble - The Scoble Show
Very cool. Does it key in on brand names at all, because with a lot of these items, brand name, I mean ask Maryam, she has to have the Gucci bag, she doesn't want anything that doesn't say Gucci.

Munjal Shah - Riya
I'm sorry, I'm very sorry. We anticipate we'll be getting much hate mail from spouses of women who find our -- and husbands who find our product, helping them to buy more expensive items, but we absolutely built -- the system has a full text search engine, and a full brand-based drill down sort by merchant, sort by brand, sort by price. So interestingly, it's never worst than your traditional text search engine. That was one of our design criteria.

Visual search is a whole order of magnitude better, but there are some things that are in the text, and that are in the metadata that the merchant has and don't lose that information, and let me search on it, so all of that's there. We've done some clever hybrid things, where if you type red strappy shoes, it converts red into a visual signature search. So, it doesn't just use red to task the text, it converts it naturally, it does a search on the color red, and the histogram in our visual signature for the item.

Robert Scoble - The Scoble Show
Wow!

Munjal Shah - Riya
So it does convert between the linguistic domain and the visual domain as well, just in some keywords. We haven't done it yet with like glossy and -- but later on we'll add more words and convert them to their visual equivalent, so that you can get the things that are glossy, but weren't tagged glossy, even if you do a text search on our site.

Robert Scoble - The Scoble Show
Right. Now obviously, you probably tested this out with at least your friends or a small group, a focus group or whatever. What's the reaction been?

Munjal Shah - Riya
It's really interesting, when we thought we'd do visual search, again, I mean this is one of the core lessons of start up domain, is it's all iterative. You try, test, change directions, or whatever you need to do, and so Riya 1.0 was the macro version of that. On a micro basis, we thought we're going to do face similarity and the shopping similarity. We thought we're going to do two products coming out. Like could work for both. The face similarity actually failed, and the shopping similarity worked. So we had 100 users come in, 5.0 scale, 5 being the best. We got 4.1, 4.2, 4.3 were the scores on shopping in different categories. Jewelry was the worst, I think it was 3.9, because lot of intricate designs, lot of glare, lot of sparkles, things like that, and watches were the best, taking pretty much head on every time.

Face stuff, face similarity, we built a great MySpace search tool that will let you find similar looking people on MySpace, including, find me people on MySpace who look like Angelina Jolie. We just played around with it. It got a 2.5 out of 5, and then my guys said, no, we can make this better. I was like, maybe we can't. So I had this interesting experiment done where we had build perfect results. We had a human sit there and say, this is the face you clicked on, here's the ones that are similar, where we built them by hand. So we said, all right, the best the algorithm will ever get, is as good as you and me do it, right. Then we put those results in, every other one they got a hand one, but they didn't know that they got a hand one versus a machine generated one. That only got a 3.3.

So my -- and the same one that got a five from one person got a one from another. So what we concluded was, you've got a million years of biology, You got, whatever it is, 25 percent of your neo-core text dedicated to just faces. You are a face machine, and there's no computer that's ever really going to compete with that, where as you don't have that for watches. You've only been around a 100 years, biology did not optimize itself for watch similarity, but when it comes to faces, something about it -- and maybe it was that, you looked at the chin and you looked at the eye. I don't know what it was.

But when we started just thinking about this as a company, and a strategy and technology, we said, you know what, lets find this intersection where the user expectations are such that you can succeed, combine with there's a business, combine with there's a need and combine with nobody else is doing it, and it's hard tech, and that came out all in shopping, but it didn't come out in faces. At least all the two of those work for faces, but the user expectations are way too high.

Robert Scoble - The Scoble Show
Wow!

Munjal Shah - Riya
Yeah, so we're the pretty geeky choice

Robert Scoble - The Scoble Show
Does (Voice Overlap) is never going to work for security and for other things?

Munjal Shah - Riya
There you probably can tolerate more errors. At least if you can find its one of these 100 people who may be a terrorist, the guard sitting there at the screen can still narrow it down from a million targets to a hundred, and can quickly scan that, so there's still utility of it as an enterprise automation tool, a labor savings, or an efficiency tool, it doesn't have to be perfect. But for you and me, we're not, we're not going to sit there and say, here's the, well okay, we recognize this person. We think its one of these hundred in your friend circle and click on the one that it is. No, I don't think I'm going to do that.

So we just experimented, and people were really interested in shopping. I mean we brought them into shopping, and the same -- we actually brought my MySpace users in to test that product on the face site. We showed them the shopping went almost serendipitously. We're like, oh, they're here, why don't we just show them this other one, and have them score it too. They wouldn't leave, I mean literally for like an hour, they're like, oh, this is great, I've never been able to search this way for these items, they just wouldn't leave.

So we just said, you know -- I have a saying in the company, I'm like, go where the ball rolls downhill. I'm like, if the ball is rolling down hill in this product stuff, the technology work better, the satisfaction is better, the business model is better, we just headed in that direction.

Robert Scoble - The Scoble Show
Yeah.

Munjal Shah - Riya
So we found. Now, we needed a good name. Riya was great, but really associated with the face stuff, so we realized, it'll be hard to communicate the change at least we have to do it with a new name. Company name is still Riya, and we'll still continue to have the face product up there.

Robert Scoble - The Scoble Show
Is the brand name Riya going to be on Like.com, is it going to say (Voice Overlap)

Munjal Shah - Riya
It will say, Like is a product of Riya.

Robert Scoble - The Scoble Show
Okay. Very cool. You wrote once that you thought blogging was a better marketing strategy or PR strategy than even going and visiting (inaudible) or something like that.

Munjal Shah - Riya
Yeah, yeah

Robert Scoble - The Scoble Show
Do you still believe that?

Munjal Shah - Riya
I do, as long as you're talking about certain audiences.

Robert Scoble - The Scoble Show
Okay.

Munjal Shah - Riya
Yeah. So for example, we went on a press tour just now for the launch of Like, where I went to People magazine, and I did reach out to even more traditional, because -- actually it's very interesting. 95 percent of the people who used Riya 1.0, men.

Robert Scoble - The Scoble Show
Interesting.

Munjal Shah - Riya
So yeah, I think TechCrunch and Scobleizer and others are awesome ways to reach...

Robert Scoble - The Scoble Show
Men, geeky men.

Munjal Shah - Riya
Men.

Robert Scoble - The Scoble Show
Geeky white PC men.

Munjal Shah - Riya
I'm as geeks as the last guy, right. I mean our company is so geeky, we're 45 people of which 38 code, that's how geeky Riya is, and even the ones that don't code, can kind of hack around a little. But we really...

Robert Scoble - The Scoble Show
You should go visit Nina, the model blogger over in Paris.

Munjal Shah - Riya
So what we found is -- so I'm hitting kind of, again fashion magazine, but we found a great number of fashion bloggers.

Robert Scoble - The Scoble Show
Yeah.

Munjal Shah - Riya
You know people like -- there's a blog called SheFinds or Purse Blog, Purse Blog is 1.3 million units a month. They talk about purses.

Robert Scoble - The Scoble Show
Excellent. Don't tell Maryam.

Munjal Shah - Riya
It's too late.

Robert Scoble - The Scoble Show
I'm going to make sure Maryam doesn't get to watch this video.

Munjal Shah - Riya
Purse Blog is huge. What's interesting is, a lot of these folks have started their own sub-blogs. They tell me when their articles get put on -- actually one of them was telling me, when their articles get put on Digg, they may have read a great article about what's the best moisturizer. Digg is really linked to and they get SEO to the top of Google, and then they get put on Digg, they get marked as spam within minutes.

Robert Scoble - The Scoble Show
Its kind of like, there's a spam user.

Munjal Shah - Riya
Its a real article, like I really wrote it, it's really on my blog, it's really serious, and they didn't link to anything. It was just a straight article. It wasn't an editorial or anything. So there is a sub-group, when I met many of these bloggers, the fashion bloggers.

Robert Scoble - The Scoble Show
So we need a fashion Digg?

Munjal Shah - Riya
You do or you need this one for all...

Robert Scoble - The Scoble Show
That's a free piece of business advice for (Inaudible) split off Digg, and make another Digg without the Digg brand name, and maybe even a new brand name for the fashion community.

Munjal Shah - Riya
So the point is that we're launching through these fashion magazines, we're also launching through fashion blogs. So both is actually the answer in this case. Just because the fashion blogs aren't quite big enough yet, where we can just launch to them, and you look at the number of unique visitors, and how much traffic, and how many readers they have on their blog.

There's one great one though, named, I'm not obsessed, because people get obsessed with stars, and they're just like, say, I want to know everything about Lindsay Lohan or whoever it is, and so her blog is called, I'm not obsessed, just kind of denial oriented theme, and it's a great blog. It's just small.

Robert Scoble - The Scoble Show
Yeah.

Munjal Shah - Riya
A whole new world.

Robert Scoble - The Scoble Show
Yeah, yeah that's the fun thing, everybody says there's one blogosphere, and that's not really true, there are thousands of blogospheres.

Munjal Shah - Riya
Yeah, I've run a ton. I found this whole running blogosphere and this whole kind of heart disease, kind of eating healthy blogosphere. You just pick your interest, there's somebody out there blogging about it.

Robert Scoble - The Scoble Show
So what else do we -- is there anything else we need to know about Riya?

Munjal Shah - Riya
No, I like (Voice Overlap) You've got to see it.

Robert Scoble - The Scoble Show
Well, we got done with all that. Thanks for coming out and visiting.

Munjal Shah - Riya
Thanks for your time.

Robert Scoble - The Scoble Show
Thank you.