Category: Uncategorized

Last fall I started blogging what I hoped might become a series of posts based on reading “Meditations” by Marcus Aurelius. (Check it out here.)

Well, I just finished with a massive project for my MBA (thus the lack of posting recently) and am enjoying a week off before classes start again. So I thought it would be a great time to put out a quick post or two to re-start this series. Here’s another great quote from the book:

Just because you find the work too hard to do, don’t leap to the conclusion that it is humanly impossible; but if the work can and should be done by a man, then consider yourself capable of doing it.

-Meditations, Book Six, #19

I’ve had a number of comments to my recent post on prediction markets and how they approach forecasting differently than other mechanisms, both here and on MidasOracle. I’d like to respond to a number of these comments here.

Comments and Criticism

[Chris Masse]: “Number one, I don’t understand why information aggregation would be a “bottom-up” approach (as opposed to “top-down”). Our traders bring bits of information to the market — but these bits of information were originally produced by the traditional sources (news, political polls, political forecasters, opinion leaders, etc.). I don’t understand why this “bottom-up” metaphor would apply to the prediction markets.”

I have a few comments for this. First, I don’t know where Chris gets the idea that “bits of information were originally produced by the traditional sources.” Corporations aren’t trading on political markets where there are polls and expert opinions. They’re trading on things that matter to their company and to their industry. While what he mentions certainly applies to public prediction markets, it’s virtually irrelevant to corporate markets. Individual traders bring their judgement and the perspective from their place in the company and their personal history, which when combined with other employees in the company is very valuable indeed.

With the “bottom-up” metaphor, I was trying to show that forecasts are built from the views and opinions of individual employees… from the bottom-up. “Top-down” is how much of traditional forecasting is done: put data through a model at corporate HQ and generate a forecast which is then distributed throughout the company… from the top-down.

[Chris Masse]: “[trying to paraphrase me] EPMs are such a novelty, and the corporate forecasters such a bunch of retarded people, that it will take decades before commercial organizations get to adopt the prediction market tool.”

It’s not that corporate forecasters are retarded, just that prediction markets are completely different to anything they’ve ever encountered for forecasting. And like anything different, they’re generally going to be ignored. Note that virtually all of the corporate prediction market trials are NOT initiated by forecasters, they’re initiated by general managers who aren’t so directly tied into a specific forecasting tool world-view.

[Chris Masse]: “If enterprise prediction markets were such a revolutionary and powerful forecasting tool, it would have found a market already — just like the iPod, the iPhone, FaceBook or Twitter did.”

This goes to the heart of my post. The iPod became incredibly popular because people clearly understood what it did: it served the same purpose as a portable CD/tape player, but carried the equivalent of hundreds of CD’s instead! The iPhone is still just a smartphone; it’s just got a significantly better interface. All of these technologies became popular because they did the same things their predecessors did, but better. Prediction markets haven’t become as popular as they could have been- because they do the same thing (forecasting) differently.

[Chris Masse]: “The added accuracy of the enterprise prediction markets is marginal — and anyway does not fill the gap with omniscience (contrary to people’s expectations).”

This is where I think Chris unnecessarily limits himself to examples where there is “added accuracy.” There are a LOT of applications for prediction markets where little or no forecasting is currently done; the example I commonly use is forecasting project management milestones. I think it would be ideal if a management dashboard (RAG status) was created using only the inputs of prediction markets on the probability that a project would meet its next milestones!

Sure, in cases where prediction markets are “competing” with other forecasting mechanisms, such as for demand for products down to the individual SKU level, prediction markets may not be the best tool. The power of prediction markets really comes into play in situations that are difficult to forecast without a market.

[Chris Masse]: “In the context of a Fortune-500 company, which is of course much smaller than a country, the pool of potential active participants whose trading activity is sustained over time is quite tiny.”

I just want to point out again that in my research I found that a group as small as 16 people could generate calibrated forecasts.

[Medemi]: “In order to make good predictions one needs both approaches, bottom-up as well as top-down.”

I completely agree with this. Companies can’t live on prediction markets alone, but neither should they do all of their forecasting without prediction markets!

[Medemi]: “The problem is, this valuable information (from the experts in the field) and how the problems can be solved was not passed on the management. Why? Because they are not interested. The bottom-up approach is simply NON-EXISTENT.”

I disagree slightly. I agree that certain people in an organization are dis-interested, because they’re close enough to a problem that they think they can solve it and don’t want to hear any bad news. But I have also talked to senior managers that are very interested in using prediction markets for exactly this reason; they want to know if their project managers are telling them the truth! Unfortunately, these tend to be fairly senior people in a company, and it’s tough to get in contact with them.

Summary

I hope this clarifies my position; that prediction markets are a completely different way of approaching the problem of business forecasting, and should not be pitched or considered as a replacement technology, but as a powerful but complementary technology.

The field of prediction markets seems to be going through a bit of a crisis of confidence recently. I would personally trace it to the recent election (where other forecasters like Nate Silver made forecasts as good as PM’s), as well as recent press like the Economist article. The general feeling is a questioning of prediction markets: if they’re so good at forecasting, why aren’t they being used much more widely? I think I have the start a good reason why.

Core Issue

Traditional forecasting is done through highly analytical techniques using past data. Statistical measures are used to generate forecasts, with probability ranges. This industry is quite large, and is highly exacting.

Prediction markets take an orthogonal approach to traditional forecasting. Instead of a “top-down” approach where huge data sets are analyzed, prediction markets use a “bottom-up” approach that combine individuals’ forecasts.

The reason prediction markets haven’t been adopted widely is because they are a tool that approaches the forecasting problem from a completely different perspective.

An example — Enterprise Business Intelligence

I’ve recently been looking into the Enterprise Business Intelligence/Business Management industry, and came across what I think is a similar phenomenon. The vast majority of the industry is composed of massive analytical solutions from the likes of SAP, Oracle, IBM, etc. They are massive companies, and implementing a “solution” can easily take a year or more. Their clients design the system from a “top-down” perspective, determining from the outset what the processes and procedures are going to be.

But then there is software like Thingamy. Thingamy is the creation of Sig Rinde, a Norwegian living in the south of France. Instead of looking at enterprise business intelligence from the top-down, he has created software that approaches the problem from the bottom-up. Instead of establishing pre-defined processes (that may not even work or will be changed by the time the software is configured), Thingamy tracks emergent processes as they happen. It can start with a very small, hard-to-define process and then scales up as the business needs it.

While Thingamy has gotten some good press and attention over the years, it’s still a fairly small company. Again, I believe this is because it takes a fundamentally different approach to the problem compared to the rest of the current industry. Hugely different approaches cause cognitive dissonance, which slow adoption.

What does this mean?

There are new types of technologies that approach business problems from entirely different directions. Prediction markets is one of these technologies. Using PM’s means companies have to upset some of their current notions about how power and influence flow in a company, relying on “soft” information from lower-level employees. A different approach also means that in certain situations they’ll be clearly superior, but also that in other situations they won’t be. Traditional methods and thumb-rules for situations just don’t automatically work.

For example, prediction markets where there is a lot of public information (like election markets) may prove to integrate new news and information more quickly, but may not be quite as accurate as other methods in the final analysis. But where information is scarce (like some internal corporate forecasts), a prediction market may be ideal. In general, new ways of thinking have to be established to know when and where to use this new tool effectively. That’s why I believe prediction markets will take quite some time to see any sort of a spike in growth; expect a slow burn for a long time.

Quick note

Just a quick note for you all. I’m curious about how well Google’s AdSense can be used to monetize a blog, so I’m going to be running AdSense on this blog on a one-month trial. If you have any opinions on this, please feel free to e-mail me or comment below.

Amir Nathoo is a Founder and the CEO of WebMynd.com, a really interesting startup company founded a little over a year ago. He spoke on Wednesday this week at an event organized by the Cambridge Network at the offices of Red Gate Software.

Amir spoke about his company and his experiences getting accepted to and going through the Y Combinator program. If you haven’t heard of it, Y Combinator is a really innovative program for software start-up companies. You get a small amount of funding and go through a three-month boot camp of getting your software ideas up and running. The whole time you have incredible mentoring from some of the best web/software advisors in the world, and the program concludes with a Demo Day, where companies show off their software and businesses to press and investors.

It was a great talk, and really brought out the benefits and realities of the Y Combinator program. Matt Schofield, the CEO of the Cambridge Network, wrote a blog post about it here.

Before I forget, you MUST try out WebMynd. They released a brand new version just a few weeks ago and it is AWESOME. You have to have the Firefox browser installed, which I highly recommend. (Go here to download Firefox.) Once you’ve got it, just head to WebMynd’s home page here and click on “Install WebMynd.”

WebMynd does four things:

1. Gives you more (and potentially more useful) results whenever you do a Google search. This is a fantastic feature.
2. Records what web pages you’ve been do, so you can literally go back visually to web pages you’ve seen recently.
3. Keeps a listing of the web pages you’ve been on recently, which lets you easily go back and/or share links.
4. One-click sharing pages with friends via Twitter, Facebook, etc.

There was some interesting talk on the night about Y Combinator. A few of the Cambridge Angels were there, as well as other investors like Laurence John (CEO of Amadeus Capital Parters Seed Fund). It seems like people are interested in the model, but want to create something that is appropriate for Cambridge. While Cambridge doesn’t have the same level of expertise in web technology as Silicon Valley does, there is some really advanced technology being developed here and a latent entrepreneurial spirit.

Laurence has started discussing this a bit on his blog, and I look forward to hearing more about it. Based on some things he said Wednesday night, it makes sense that any program needs to come from a consortium of angels or VC’s. This eliminates any negative connotation if a particular angel or VC chooses not to further invest in a company that was accepted into the program.

Perhaps I might be able to provide some perspective later this year… I’ve applied to this summers’ Y Combinator program. (With a thank you to Amir for his feedback and perspectives which were incredibly helpful!) With so many applicants it may be a long shot, but will know more in just a week and a half.