Developing a Task-Based Relevance Engine: an Interview with Intellext's Al Wasserberger and Jay Budzik
It wasn't until reading through the interview responses from Intellext's Al Wasserberger and Jay Budzik that I really started to get the full impact of what they're doing at their core. Somehow in my question-development research I'd not fully grasped what's most exciting in Intellext's Watson downloadable search tool: Task-Based relevance.

The core concept is simple - Watson runs in the background and gives you easy-to-snag bits of information that enhance whatever task you're doing. Such as blogging. Or writing term papers. Or developing code. Or shopping.

This interview digs into much more than Watson though. In our 3,575 word conversation we investigate the contributions of Intellext's venture firms, Stanford vs. Northwestern, the Chicago start up community, social search + Watson, MySpace and attention data.

I want to thank both Mr. Wasserberger and Mr. Budzik for their time, but I also have to pause and give special thanks to Leigh Winter, Intellext's Director of Marketing Communications. Here's why (from her email):

I sat down with Al and Jay on Monday and we went through all your questions. I thought having it as a conversation with them would get you the most natural responses.

I transcribed the conversation as best I could, and now Al and Jay are just looking over the write-up to see if there’s anything else they want to add.

They loved your warm-up question by the way, and thought it was for real at first when I read it! Thanks for putting such thought into it.

Thanks again to everyone involved, including you Ms. Howard - thanks for reaching out. Let your bosses know you should get a raise.

Quick aside: Loren Baker deserves the best from me - I tend to over promise and under deliver to him though. I'm going to state the intention here to send him excerpts from the wonderful thought shared below, and have scheduled time in my calendar for it tomorrow.

Warm up:
Describe the development process surrounding your core business concept of contextually relevant, double underlined links you add to partner sites' content.

Just kidding. Have there been significant or humorous confusions of Intellext and IntelliTXT in the press or anywhere else?

A – Tell me about it – and it’s not just IntelliTXT. We were just down at the CompTIA Breakaway conference, and they were announcing best of show, and as they said “Intel-“ I started to get out of my seat…but they never said the “-lext.” Seriously, though, people identify us and our brand with Watson, but unfortunately Watson Pharmaceuticals, a slightly larger company than us, owns that domain, so we can’t just go by Watson.

Funding:
Intellext is funded by Motorola Ventures, Palomar Ventures, LaSalle Investments, and University Fund.

How do you manage pressures from such a diverse group of funders?

J – Well, at the end of the day they all have the same goals, so they are all aligned around the business being successful.

Motorola Ventures - focus on mobile investments; what is the mobile future of Watson, and what kinds of mobile pressures do you get from Motorola Ventures?

A – I describe it as more as support than pressure. Mobile platforms until recently haven’t had form factors appropriate for delivering this kind of information in this way.

J – Motorola recognized early on that this would change, and we’ve developed infrastructure that would easily port to different forms. Now that mobile platforms have evolved, don’t be surprised to see an offering in the near future.

A – As we evolve into that space, having Motorola Ventures as a partner in the business has obvious advantages.

Palomar Ventures - in their "about us" they sound very hands on. What have been their significant contributions thus far?

J – The partner we work with there is David Curtis, who was one of the authors of CORBA spec. Having a great technical mind like that gives me (and the rest of the technical team) a great resource.

LaSalle Investments - seems to focus on real estate. Why are they investors? What have been their contributions?

A – LaSalle Investments (not to be confused w/ LaSalle Securities) is the investment company run by legendary Chicago angel investor Bob Geras, (recently inducted into the Chicago Area Entrepreneurship Hall of Fame (and profiled in Inc. last year: http://www.inc.com/magazine/20050701/angels-in-america.html). Bob has invested in more early stage companies than anybody I’ve ever met, he’s constantly a resource for us as we grow. Our regular lunches at Bijon’s always yield good actionable advice, and his rolodex doesn’t suck either.

University Fund - is this a fund from Northwestern? How would you characterize the expectations from this fund?

A – No, it’s not associated with Northwestern, it’s an early stage fund that focuses on investments in technologies that come out of universities. Their expectations are simple – profit.

What in all of the Watson functionality has convinced your investors that they have made a good decision in putting money into your company?

A – All of our investors have one thing in common – they love technologies that have potential to disrupt a large market. Each of them realizes that Watson’s ability to get information for the user without the burden of a search box will change the landscape of the search industry as we know it. Google has proven the scalability of monetizing the search box. Our investors see very clearly the upside to monetizing everything people do outside the search box.

J – And more than just the search industry – the information industry, software industry, computer industry…There are so many areas of impact – publishers, advertisers, and obviously the end users.

What do you invision the core monetization of Watson being? If it's advertising, how do you think Watson could or should change the standard search advertising model?

A – Today we monetize Watson in two ways, advertising or subscription. Most users choose to get Watson free by seeing (and hopefully occasionally clicking) on relevant ads. Business users tend to choose to pay through a subscription or license for the ad-free version, Watson Professional.

J – Advertising will be the core monetization.

A – Jay’s right – in the long run, we predict our model will move more and more to an ad-supported user base. We took a big chance a couple of weeks ago by removing the differences between the ad-supported and subscription versions. Surprisingly, we haven’t seen a big change in the percentages yet – about 10% of our users still opt for the subscription. I think over the next few years, though, as corporations get more comfortable with ad-supported software, that many software packages will be delivered in this fashion, and we’ll eventually have a very small subscription base. The trend is certainly for more applications to be ad-supported; you see it with Microsoft moving their consumer applications in that direction, as well as AOL IM and other products. Ads can be served based on profiles or randomness, and many can be irrelevant and annoying, or they can be served in a way that they are helpful to the user. Our ability to drive contextual ads should make this trend in the industry actually feasible.

A – As for Search advertising, it is exactly that – advertising for when a user decides to search; and it’s only applicable when a user goes to a search website and does a search.

J – Right! This opportunity is huge, because users spend more time not searching than they do searching, and in applications away from the browser. With Watson, that is even more true because they don’t need to search in order to get what they need anymore. Now all of that time can be monetized. Instead of bidding for keywords, advertisers will be bidding for contexts, and the resulting experience for advertisers will be richer because they will be more integrated into what users are doing. And for users the ads will be more valuable and relevant.

A – And advertisers who can be very specific will have very high click and conversion rates when the contextual relevance is high.

What are your measures of success when there's no business model existant?

A – Well, there are business models, we have revenue streams from both our advertising and licensing models that are growing every month. But we do look at more than that – our success metrics are user growth, user retention and buzz. We look at how users interact with the product, we look at the impact of new features on user behavior. We get a pretty steady flow of comments both good and bad from users; our goal right now is to get millions more happy users, all of whom are telling their friends about Watson.

Northwestern:
Northwestern University vs. Stanford. Talk to me. What are the cultural differences? What are the academic differences? Why did you guys choose Northwestern?

J – I chose to go to Northwestern to study with the best in AI and cognitive science and user experience, it was a group of people that I knew I could work with well that were creating something powerful; specifically Kris [Hammond] and Larry [Birnbaum]. Northwestern has a very supportive, engineering-oriented product development environment that encouraged its faculty to have tangible impact on the world instead of just academic impact. The engineering orientation allowed us to focus on building software that worked, instead of just making algorithms that were elegant. I can’t say what the cultural differences are, because I’ve never worked or studied at Stanford. They are certainly very smart, but the practical engineering orientation at Northwestern was particularly helpful in developing the technology and bringing it to market.

What other Chicago tech companies are forming around Northwestern tech talent?

A – There are some great examples. NanoInk and NanoSphere are two good examples of the nanotechnology research being commercialized out of Northwestern. Also, NuCrypt and Nerites have some cool products from NWU research. And no, I don’t think there’s a rule that if you did your research at Northwestern you have to have your company name start with an “N”.

What is the Chicago tech startup community like? How do you feel you have contributed to it?

A – Chicago has a stronger technology community than most people in or outside the city realize. Large companies like Motorola, SSA Global, SPSS, Orbitz, Navteq, and emerging disruptive companies like FeedBurner, 37signals and us. There are also a lot of serial entrepreneurs here. This is the third Chicago-based company I’m building, and there are people like Chris Gladwin who founded Cruise Technologies and MusicNow, and is now building Cleversafe; Gian Fulgoni who built IRI and is now building ComScore; Dick Costolo who built SpyOnIt and is now building FeedBurner. We’ve even organized ourselves unofficially into a regular charity Texas Hold ‘Em game (Crains: The Buy-in After the Bust). Not only is the entrepreneurial spirit strong in Chicago, so is the availability of amazing talent.

Northwestern University, University of Illinois, University of Chicago, IIT, DePaul and others provide a steady flow of brilliant, well-educated technical folks, and business schools like Kellogg, Chicago GSB and DePaul are a great source for leadership talent.

Watson:
how would you characterize the desktop app downloading demographic? How are they different from those who prefer web-based apps?

J – With a downloaded desktop application, the perceived value has to be higher because the level of effort to add it to your desktop is higher, but then after it’s installed, it’s stickier. The bar is higher, we have to add more value than a web-based application. With web-based apps, people can go check it out once and then leave. There are few if any barriers to exit, a service like Google is easy to leave. Google proved that by taking all those AltaVista users, right?

Who are you missing out on by not making this a web-based login app?

J – Well, the obvious answer is all of those people using alternative platforms (Mac, Linux, etc.) that we don’t support yet. They are a vocal bunch! Also, some people can’t install software on their work machines because they don’t have administrative privileges. This functionality couldn’t be delivered in a web-based model, though. People are not spending all their time in web-based applications. Watson lives beyond the browser, helping people in desktop productivity applications as well as browsers.

I like the concept, but I think the SERPs categorization and display lacks vision. I appreciate that it's customizable, but I can't invision a user who could ever create the appropriate categorization there for all of a day's given tasks.

I think a more conceptual overlay would help me more precisely invision the "thought space" of the work I'm doing. Think Ask's Zoom function or Q-Phrase's ConceptQ.

What are your directions for results display?

J – You’re absolutely right – there is no single categorization that applies to every task or every user. We’ve made some progress here by letting users re-arrange the categories in Watson. In the not-too-distant future we’ll also let people set up multiple category “profiles”, so they can switch, for example, between homework, shopping, blogging, and other modes. These types of categorizations cut across information sources, and the content of the document. Instead of organizing the search results based on the contents of the search engine, like you’re suggesting (or like Ask Zoom or ConceptQ), we’d like to see them organized along dimensions that matter for the task you’re performing. This is one of the things we’re working on doing differently, and an area where we currently have a patent application pending. Soon we’ll be able to have the categorization match the task, instead of the type or content of the search results. When a user is shopping, they want to see prices, reviews, alternate products, accessories, people in their network who have bought or blogged about the product. This is different than when they are looking at a contact, where they want to know information about the company that person works for, bios, maybe financial info about the company. The way that information is organized, and even where it’s presented (inside a contact record, or on a webpage, or as a part of the document they’re reading or writing) are all going to evolve in this manner. Don Norman, one of my professors from Northwestern and a noted authority and champion of Human-Centered Design, calls this approach Activity-Centered Design.

Talk about including a human-aided search element. What could this look like?

J – User contributed content is one of the most powerful elements in search these days, but it only makes sense when the tags and ratings can be applied in a context; just because an article is highly rated doesn’t mean it’s relevant to me. We’re adding context to this model.

I see this as having a great impact on your different users - for example the shopping users could notify each other of excellent deals they found. Researchers at a common company could alert each other of relevant or especially useful news.

What other scenarios can you envision?

J – We’ve been working for a while on models based on shared user collaboration. For example, users can leave notes on a context so when others enter that subject area, now they can collaborate based on what they are doing with the people that are relevant to what they are doing. This creates a new channel that’s oriented by content instead of places. We can bring together people with similar and specific interests. Think of it as the next generation of MySpace or LinkedIn – where the people you should meet come to you, instead of you having to find them.

Imagine for me how a more social Watson would change if you made it mobile as well?

J – Just as context can be about what you are working on, context can be represented by a number of different factors, like location. Just as Watson can understand documents, Watson could understand the conversations you are having and help bring you information you need. Or the people near you for that matter – Watson meets Meetro?

I find it hard to understand why start up search companies would focus on a "broadcast" or "mainstream" market.

Your approach is compelling - prefetching the information task performers might need. I'm still not convinced about your intention to appeal "to everyone with a computer."

J – Ok, maybe not everyone with a computer, just those that user their computers to consume or create information.

In the Pirillo interview Budzik mentioned that certain tasks are emerging that Watson seems to compliment especially. What are these tasks?

J – For students, Watson is a great research tool while writing papers. For bloggers, Watson finds great links (other related blog posts, etc.) that help them write richer, more interesting posts. From the start, journalists have told us how Watson finds them great information and new perspectives while writing their articles. And news junkies, simply because Watson does the work of finding and delivering related information from all over so they can devour a topic.

Will you roll out Watson plugins that better target these tasks?

J – Absolutely – we started doing this a few weeks ago with our new add-ons site (http://www.intellext.com/add-ons). For students, we’ve introduced add-ons for SCIRUS and PubMed, plus some social networking ones like MySpace and Friendster; for bloggers we rolled out add-ons for many of the most popular blog search engines (Technorati, Sphere, IceRocket, Feedster, Blogster). Going forward, we’ll be making Watson better at interacting and understanding more of the popular blogging platforms, so Watson can work alongside each and deliver relevant information while the blogger is posting. We’ve got a bunch more add-ons coming, many targeted at students, journalists and bloggers, and some new areas as well.

I think a strong plug in would be an iTunes plugin that enables me to identify MySpace or Friendster members who might make personality matches. Please crush or elaborate on this idea.

J - Great idea, want to see a demo of the prototype?

How did you envision people using the MySpace plug in, and how much of the decision to create the plug in was based on the current popularity of MySpace?

J – We try to develop Watson to deliver the information that people care about. MySpace is a place people are spending a lot of time, consuming a lot of information. We see people using Watson to help them stay connected to MySpace even when they are not on the actual site. Now a user checking out music or movies somewhere else online can look over to Watson and see what the buzz is on MySpace. Watson shifts MySpace from being simply a destination, to being a community users can bring with them as they go about other activities.

Who from MySpace was involved in the process?

We didn’t need to work with anyone there – we just had to build an add-on to search MySpace!

Marketing:
Elaborate on your current blogging strategy and how it fits into your overall business goals.

I wish I had more time to dedicate to it, the good news and bad news is we’re so busy.

As far as the product goes, we’ve found some bloggers that draft posts in Word in order to get additional results from Watson and incorporate them. We plan to develop Watson to work alongside other popular blogging platforms as well, to make this more seamless.

Privacy/Attention Data:
You're very consistent about your respect for privacy. Have you seen privacy concerns diminish since your inception or increase?

Watson will learn quite a bit about what I do during the day. How about you package MY data for me and let me license it if I want to? Why not pay ME for using Watson?

A – At first, there were some users that were really concerned about what Watson could “see” – which is why we have taken such a hard line about privacy. Everything happens on your computer, not our server – so we actually don’t learn anything about what you do during the day – we don’t track what you are looking at, the links you click, the keywords you use, or even who you are. I mean, we can tell that a user somewhere clicked a result that came from MSN or MyScpace, but we have no more insight than that.

J - Some products have to use user profiling to deliver good results, and some use them to target advertising. We don’t have to have any of that data, because Watson knows the full context of what you’re doing. That means we can get you the right information, people, or even ads without violating your privacy.

A – As for letting you sell your profile, it’s a neat idea. Is that a direction that the industry is going to go? Your guess is as good as mine. Do we need to collect info in order to deliver a great product? No. Do we have technology that could do that if we wanted to implement it? Yes. Right now, though, our focus is on driving more value for the user. As soon as we start to blur the line on privacy it opens up a lot of questions; so for now we choose to stay very clearly on the line of not even tracking anything controversial.

Watson would be a perfect partner for AttentionTrust I think. What are your thoughts about that organization's involvement in the attention economy?

A – We are aligned with them in our views on respecting users and privacy. Attention Trust’s primary principle that you own your attention certainly fits with Watson, which doesn’t collect or store any of your attention information. Their model of tracking and letting users monetize their attention is interesting. For now, though, we’ll stay on the side of not storing anything sensitive like attention until we see where this goes.

Update:
Intellext - Dr. Jay Budzik and CEO Al Wasserberger A nice spoken interview from David Dalks.
Garrett French - MarketSmart Interactive | permalink | +marktd |