Monday, January 28, 2008

Paul Teich of AMD – Scenario Analysis for Predicting the Future

Paul Teich of AMD describes scenario analysis in developing roadmaps for product development. He also discusses the “Long Nose” concept and his first startup experience.

Tell me about your background?

I earned a CS degree at Texas A&M back when C was still a pretty cool programming language, and much later I earned the joint IC2/UTexas MS in Science and Technology Commercialization. I worked at AMD for pretty much all of the 1990s and then moved to Dell’s short lived New Product Concepts group. As that was disbanded I left to direct product management at a start up that you have never heard of. After that ride I went back to AMD. For the past 6 years I’ve been the business strategist for AMD’s Opteron processor brand.

What was the name of the startup?

It was called Fracta Networks, we were based here in Austin. You probably never heard of it.

You’re right I’ve never heard of it. What did it do?

It was a very early experiment in how people can track information they think is important, store it persistently on the web, and share that knowledge with other folks. The intent was to save snippets of web sites and documents so if the web site or document disappeared you still had that bit that you thought was important captured persistently.

Sounds like a backup or archival system. Is that right?

Actually, look at Google Notebook which is a very little known app. It does much of what the Fracta app did years ago for web capture. We did some things differently, of course, they seem to have a broader audience targeted. They’re Google, after all…

What was the business model?

Back in 2000 there wasn’t a viable advertizing model, so we were looking at subscription-based revenue. Our target was knowledge workers, so the customer, or at least the person who eventually paid our subscription fee, was a company. If we did it again today, we’d use an ad revenue model and skip the step of asking customers to ask their employers to reimburse them.

One of the issues systems like this face is privacy issues. As your system gathers information, at some point it learns significant things about customers. What can you ask them to let you monetize? What can you do and not do?

Facebook faces that same question today. How do you solve it?

The whole privacy thing is very interesting because if you ask your customers for meaningful demographic information, if they see the point in letting you have accurate information, you can start to figure out what groups of similar people do on the web. I believe that people at work trying to be productive will make different privacy tradeoffs with their professional persona than they will outside of work with their private persona. What’s it worth for say the Wall Street Journal to know what authors a specific technical community reads and which ones they do not read? We don’t know how much that is at the moment because no one can do it with any accuracy. Everyone says it’s worth money, but how much? There will eventually be interesting incentives in place to gather high-value personal information.

What happened to Fracta Networks?

It was funded in 1999-2000 primarily by Polaris and G51 just as the market was peaking. As the market headed down we started looking for partners or a buyer. We ended up selling Fracta to Chicago-based Divine Interventures as they changed their business model from VC to that of a product portfolio company. I worked for Divine until shortly before their reorganization in 2002.

Tell me about scenario planning. How does it work?

I do a lot of forward looking market research and forecasting, have been doing it for decades. I particularly like a tool called scenario planning. It’s a very handy tool for both envisioning and mitigating risk.

All of the big new technologies and products for the next five years are already in the market today. The challenge for predicting the next big thing is that they have not yet reached critical mass and are therefore, for practical purposes, invisible. I recently read a paper called the Long Nose of Innovation. It talks about the incubation of technology…immature products need refinement and nurturing. Most people see new products only after the incubation and refinement period, as they cross the chasm into larger, more visible customer bases.

I’ve found that a lot of futurists are on the “lunatic fringe.” They are very excitable, they have incredible detail about stuff I’ve never heard of, and in general they don’t listen very well. They know they are smarter than I am. Some of them have better manners than others and do a very good business consulting, but they come into a business arrangement with the expectation that you’re hiring them because they already know the answer.

If you look around, there are a lot of smart people who have bits of relevant information, knowledge and experience…and they are not only willing to help, they’re happy to help.

It’s the “Bazaar” model as opposed to the “Cathedral” model. Are you familiar with the metaphor?

No. What is that?

There’s a great paper called the “The Cathedral and the Bazaar.” Microsoft and Oracle are examples of companies building isolated, monumental, proprietary pieces of software similar to a ‘cathedral’ while the open source movement treats software more as a ‘bazaar’ where development is a collaborative project among many differently motivated people.

The key to forecasting in general and predicting discontinuities in particular is that individuals are guaranteed to be wrong about major aspects of their predictions.

A group of smart, informed people can collaborate very effectively to envision major features of the future. This is the basis for Prediction Markets, but in scenario planning we’re primarily trying to figure out what’s uncertain about the future, because what’s certain isn’t going to help you mitigate risk.

In scenario planning, we’re looking for events that are fundamentally uncertain for the organization and the overall industry. We have a method for ranking the uncertainties and the ones that pop up on top of the list are events that are inherently unpredictable in timing or in directional vector. In a scenario planning exercise we look for these uncertainties because they are areas of weakness in the industry and you can’t predict what’s going to happen. You then create a set of scenarios that are structurally different based on different resolution of the key uncertainties. They are not a little different, they are a lot different. And they are all plausible.

We can then use the scenarios for contingency planning and risk mitigation. Pick a product direction. The scenarios inform the product specification in the same way that wind tunnels inform aerodynamics. You have a set of specific futures you can fly your product through and ask how well it will do. In product planning the goal is as much to make sure the product doesn’t suck as it is to ensure that you hit a home run. A home run product is very hard to intentionally create. I’ll take a series of doubles (extending the gratuitous sports analogy) any day.

Most of the category-killer software applications in the market today are not the best at everything they do. From productivity suites to databases, there are niche products that compete quite effectively with the market leaders by being better at some relevant performance metric. The trick for most of us is to have a “good enough” product to gain a large audience.

Do you know David Smith of Technology Futures ?

I’ve taken his course. It’s very well grounded, I enjoyed it. He’s very quantitative in his approach. He has very sound mathematical models. It takes a lot of research and dedication to use his models as they should be used. His models can’t predict discontinuities, but no trend-based model can – some events have fundamentally unknowable probabilities of occurring. The trick is to use other methods, like scenario planning, to put bounds on discontinuities and their timing. Forecasting markets and technologies within scenarios makes good use of both techniques. It’s all about making rational assumptions.

So what are you planning to do with scenario analysis?

I’m co-authoring a book on the scenario planning technique I’ve jointly developed with the other authors. Our book will help small organizations – startups, workgroups, virtual teams – make use of scenario planning without spending a lot of money or dedicating huge internal resources. Along with the scenario planning I’ve lead at AMD, I’ve done pro bono scenario planning work for non-profits and I’d like to see the technique more widely adopted.

Best regards,
Hall T.

No comments: