Interview with Munjal Shah, CEO of Riya

Riya is designed to help search photos. The world is awash in digital photos today. Some three hundred billion photos have been taken by consumers in the last five years and there is no way to search them. Most of us have from five to ten thousand photos saved on our computers that have hard to remember names such as DSC009.jpg, because that?s what the camera saves in the name, and we haven?t changed it. Riya started with the simple idea that it?s time to make photos truly searchable, and do that by looking inside the photo using face recognition and text recognition and other techniques to figure out what?s inside. We are currently focused on personal photos, but we do plan to enable searching of photos on the web as well.

I think there were a couple of them. But first, it was just need. I personally have saved 31,000 photos and I just couldn?t find anything. Frequently my sister and others would be like, ?Did you give me all those photos of my son?? And I?d be like, ?Oh, no, I don?t have the time to do that.? So I think that was an element. The second was that I spent some time in Korea before I started this company and noticed that they have these high-resolution camera phones now ? 2 megapixel, 3 megapixel, and consumers there are taking ten pictures a day because they have these high-resolution camera phones, so I just started realizing, ?Wow, the number of photos people are taking is going to go up, hence the need for a photo search is going to go up even more than the three hundred billion.? So, those were two major reasons. The third, less so initially and more so later, is that I dug in there and I realized, you know, Yahoo images and Google images ? the existing photo technology hasn?t changed in ten years. Nothing. It?s just been looking at some text around the photo, and that doesn?t do a very good job if you?ve ever used those products. So, it?s like, wow, there?s a need that?s growing dramatically, and a need I personally feel, and at the same time, something very few others have invested in to try to make better.

Well, that?s a tough number to identify. The estimate that we have, which is very inaccurate, because as you can imagine, it?s a pretty hard number to come by, is about three hundred billion digital photos. What?s happening is that the numbers are getting bigger each year and growing at a faster rate. People are taking more photos than they were before; they?re higher megapixel. They have more resolution to them. It?s just accelerating. I even look at my own collection and my ?99 folder has three photos, and my 2000 folder has 1,000 photos, and my 2001 folder has 3,000 photos.

Yeah, hahaha, I had a kid recently, so? that?s part of it. Haha.

We started with a very small, private alpha of twenty people, and that?s been going on for about a week now, and some of those folks have posted their reviews on their blogs. And so, we took that, and we?re just kind of learning from that and accentuating the parts they liked and fixing the parts that are broken, stuff like that. So, we really used that as an extended QA process. We?ve had this opportunity for people to request an alpha; we?ve gotten 4,000 requests in the last 7 days. So we?re going to try at least 1,000 of those people in about a week?s time. So, sometime by early next week (mid November), we should let those guys in, and then we?ll see what we learn, and if everything goes smoothly, then I?ll be opening it up to more and more people. If things don?t go well, then we?ll fix them, and then open it up. So, I think that, frankly, we?re less conscientious about the timeline, and more about, ?Is this working properly for people??

Well, to be direct, that?s part of the reason we?re opening up to 1,000 people. You actually don?t get enough feedback when you only have 25 people. They give you some, but you need more, and you need more testing. I use it on this PC with this browser, and this thing broke, and you just need more input. Some people use it and that?s great, but they don?t have time to give up detailed feedback, and so you kind of need a thousand people to get a hundred good feedbacks to get a statistical analysis that says 50% of users are having this problem and 25% are having this problem, etc. That?s some of why we?re opening it up further, so that we can get even more input. We?ve actually taken most of the input we?ve gotten and incorporated it into fixing it. So with the first 25, we didn?t get so much that we actually had to prioritize. We just actually kind of took it all and said, ?We?d better fix it all.? Everybody?s comment was important, but it?s not enough; we would actually like more. So, we?ll open it up to more people so that we can get that. I don?t believe that consumer products ever launch kind of successfully on day one.

Exactly, exactly.

It was less strategic than that, I think. We just put up that, ?If you want to join our alpha, e-mail us here,? on my blog, and we put a few comments that people had written about the initial concept. The first 25 were the first 25 who sent us e-mails requesting it. They weren?t cherry-picked, really. These were the people who were interested. By the very fact that we?d only advertised it on the blog, it kind of self-selected to people who take the time to read blogs.

Yeah. Which is okay. We did deal with some testing of our product in Chicago, with moms ages 45 and 35, pulling a standard demographic, but when you look at most web products today, it is kind of the initial websphere, blogsphere that kind of uses them, and the product gets simplified over time and expands so that people can use it.

We?re not talking about that yet in specific. At eye level, we?re building a search engine to search your own photos, as well as search photos on the web or photos you?ve made public. And the business models for search engines are fairly known and are proven.

Yeah, that?s all it is; it?s a search engine. It just happens that we?re starting with a desktop search while most searches start with a web search first and a desktop search second. We?re just kind of turning it on itss head a little bit, and the photos are mostly on people?s desktops right now.

We could do it all on the desktop, but you lose a lot of the effect and for the user, that means they?re going to have to do a lot more work. To give an example, the power of the system is really in the fact that if any person in the world has taken the time to train up on what a certain person looks like, just by adding their name and e-mail ? I mean, you gotta have both — to the address book, so let?s say I go up and I train what you look like and I say, here?s your name and here?s your e-mail address, and then your sister comes online and she just adds you to her address book automatically. In that case, she won?t have to do any work, and all of a sudden you will be recognized in her photos. Then, if she does do more training or somebody else does, it benefits every single person and gives a more robust ? what we call, visual signature of you ? and it keeps building. And so over time, as everybody in your friend?s circle has joined, Riya ends up learning, and you may come in, upload your photos, do no work, and all of your friends are recognized just by adding them to your address book. It does not try to recognize everybody you have in your photos. You really have to have had them entered to your address book. We don?t even try to look for them. Some people have written, ?Oh, I?m worried about it recognizing me in the background of some picture.? Well, if they don?t know you, the system won?t even try to find you in their pictures and therefore it won?t be successful in recognizing you. The real power of it is that over time, what happens is that each person trains their own signature, and all their friends can leverage it. And that?s a system that was easiest done on the web, rather than on the desktop.

We don?t see this as competitive. You can put the Google search box on your webpage, right? And it?ll search your site and provide a site-specific search to your customers without your having to run search engine software and anything like that. And we?re saying, ?Oh, we?ll make the Riya search box.? You put it on your photo site, and it?ll search all your photos. And let your users search their own photos on the site, so that?s exactly the type of concept that we see being able to bring to all of those ? and for many of them, especially if they make money off of printing ? the real economic benefit?

Exactly. And, oh, by the way, most of the time today, printing only occurs for photos taken within the last 30 days, and printing only occurs in 70% of printing is done by the guy who owns the photos, not by the friend they shared them with. But if I can search for all of the photos of me that are available, I?m far more likely to print more. So we think we have some very good propositions that we can bring that are like, ?Hey, your business model is X, and if we can increase printing 10%, that?s a big deal.?

Ah, I think it?s all of the above. I think a blog can?t be overly focused as a strategy or it loses its authenticity, and so it just kind of has to be you, and it just has to be you with all your spelling mistakes and all your imperfections and it just kind of has to be you having conversations with a lot of different people. I?ll be direct about this and say that this is also kind of a way that we?re marketing the company, and we?re not hiding that, but it?s not something that we?re trying to hide in any way. It?s part of being authentic. So, you know, the blog itself is just me for all intents and purposes, and if that happens to help Riya, great. It?s something I want to continue even beyond Riya for a while, in fact. I kind of feel dumb that I even had the epiphany ? I should have been blogging, what was it, only four or five months ago.

Yeah. I mean, it?s so funny. The old way of launching companies was, I stand up, I beat my chest, and I promote the heck out of my company. And you know, all that did was create this exaggeration, right? Where everybody?s like, ?My products great.? Your products are never great when they first launch. That?s not the way it works. They?re okay, usually. They might be a really neat idea, but they?re okay. Finally, the blogsphere is a mechanism for you to really come out and say it. And because you have a direct voice, I, at least, as a CEO, feel far more comfortable saying it. You know, if I were to say it the old way where I went around and I talked to ten journalists in New York and they wrote articles about the launch of my company, I?d be a little nervous, right, if I said, ?Well, it doesn?t really work that well,? but all of a sudden, the quote comes out in the press to, ?This product doesn?t work? But in the blogsphere, there?s such a rich texture of people who will read more than simply one headline. You can kind of be even more authentic and without as much concern. Even if someone else blogged about what you were saying and gets it wrong, you can go and post a comment saying, ?Hey, thanks for this; this part actually is incorrect,? and they can do the same on your site. I mean, it just goes back and forth.

You know when I? uh? I think that I learned a lot out of Andale. Sometimes it?s hard to remember what you learned where, but it?s ? I think, so for an example, I?ll give you, I once did a presentation on a conference called E-Bay Live in a room of about 500 people about a new set of products Andale was launching. And um, that community and I stood up there and was like, ?Hey, this product?s great, you should really try it. Seeing when we got it, and it?s just super.? And then I got to this one product that we had and I just kind of stood up and I paused and I was like, ?This product?s horrible; don?t buy it,? and it?s really broken, and we?re working on it. And I remember a whole bunch of people coming up to me and going, ?I cannot believe you said that as a CEO about your own products, but, you know, I really appreciate that you did, because I thought that too,? or, ?I hate wasting my money and not getting it.? So I realized that it was an important part of marketing, which was saying what works and what doesn?t. And it never really went to an online medium; it was just kind of an offline thing.

Yeah, I mean, we?re all getting so jaded with kind of too much marketing, so I think that you just have to stay authentic about it, and when it works great, I will say it?s perfect. And you know, I say that to my team as well. They?re like, ?Oh, you wrote I t really wasn?t very good.? I go, ?You used it this weekend; what did you think?? They go, ?Yeah, we need to fix this and that,? and I?m like, ?You know, then let?s just say that. It?s okay.?

I think it really is efficiency and technology. We focused on building very efficient foundations for Riya. We didn?t spend a lot of money over the last year to this point unlike my last company which we started 1999, in the go-go days, where we spent an obscene amount of money getting the product into alpha and beta. And I just learned, hey, you really want to build the foundations of a good profitable business early. I got them all to profitability, but only much later, and this time I said, ?I want to keep this thing very efficient from day one.? This for me is not about a web bubble and riding the bubble. It?s about building a profitable business and it?s just something that I learned how to do. There were some very painful years in 2000, 2001, 2002. And that was probably the first thing. The second thing was when you have to compete against the big boys, create some very differentiated technology. You cannot have a technology advantage forever, but you can have it for a year?s time-frame, and use that year as wisely as you can.

Riya is made up of about six, seven PhD?s in face recognition. It?s a pretty technology-oriented bunch that set out to do something very different. The face recognition technology?s been there, but what we?re using is actually quite different from traditional face recognition, and it?s led to significantly higher accuracy rates than you?d get with traditional face recognition. So, that was my second thing, and the third is, just realizing that you want to then create a long-term barrier from your competitors, which in our case ? if you think about it ? it?s those visual signatures. Once our system has a million trained visual signatures, even if a service that copies us launches, you?re gonna come and put your photos with us, because all of your friends will be recognized without your doing any work, right? So going to a new service, you have to do all the work from scratch. It?s kind of like a social network in the sense that the more people that are in it, the more valuable it is for a newcomer, and for that, we?ll grow and grow and grow. I kind of recognized those three things that we gotta make sure we have. And the last part is that we kind of shower our guys ? we just try to take the best care of them, the best machines, the best things, the best toys in the office in terms of their productivity and their work environment and just try to take care of them. You?ll see me blog about them a lot in the blog and in some cases, in a very detailed way. That?s part of being authentic. People keep thinking companies are made by one person, and it?s the Larry Kergei show, or, you know, the Bill Gates show.

We ask each person to answer one question, and they have to be able to answer it to get the job. We said, ?What?s the thing you?re better at than anybody else in the world?? which is a pretty bold question, actually. And it?s almost impossible for any of us to actually say. But we just wanted to see if people had experience in one area, just three standard deviations above the mean, and if they really, really excelled in an area well. So we put together a group of people who are fundamentally experts each in one thing, and each of those things were made different, as you might imagine, to make a successful team, and we just brought people who had shown that they were just great at something, instead of mediocre at a bunch of things.

You know, most people were flabbergasted when asked the question. They all stumbled for a while because it?s a very hard question to answer, but in most cases, for the guys that we brought on initially, it was things like, ?Look, I know how to make a photo site scale to handle ten million users or one hundred billion users. I know how to do that better than anybody, or at least there?s maybe ten guys in the world who can probably do this.? In the case of the face recognition guy, one of the professors at Berkeley who had given me a list and said, ?Look, what you?re doing is achievable, but I think there are only about twenty people who can really pull this off.? We got about five of those people on our team. Many of the sites you see today, millions of job engineers in the world can write them, right? Their idea may be unique, but the actual building of the website is not that hard. In this case, the face recognition, the mathematics behind it are actually quite complex ? I have a Masters in Computer Science and I was reading the math papers and I couldn?t understand the math. It?s something that?s pretty esoteric.

I don?t think I can yet, actually. I think if I make this successful, then I will have the proof that this statement is true, but for me, it?s around the passion of being an entrepreneur and coming up with unique ideas and building companies around them. When Andale first launched, it was the first auction management service that existed, period, for E-Bay sellers. That was the first.

No, I think you covered most of it. Our main thing is that we really believe that there?s this area of photo search that no one has touched in an intelligent way at all, and there?s too many photos to manually tag them all.

Yeah, we?ve been pleasantly surprised at how many people are like, ?Man, I?ve just been dying for this.?

GMail taught everybody something, and what it taught everybody was, no one wants to organize. They just want to keep and search. So we?re trying to allow that for photos.