Mercury’s Blog

If you're new here, you may want to get Mercury's Blog by Email or subscribe to my RSS feed. Thanks for visiting!

I’ve recently been reading a classic management book: In Search of Excellence, by Tom Peters and Robert Waterman. First published in 1982, this year is it’s 25th anniversary, and it’s still amazingly relevant today. (I will be the first to admit that a number of the examples have certainly aged, though the principles definitely haven’t.)

One thing that really struck me was a particular acronym/phrase they used a number of times in the book: MBWA (Management By Walking Around). This is something that was driven into me as a young Navy officer; we would be quizzed at the end of every watch what we found as we walked around and inspected our spaces (such as the engine room). What it taught me is that there’s nothing quite like watching, listening, and inspecting to see what’s really going on in an organisation. My bosses did their inspections as well, but they also relied on us because we had more time and opportunity to observe, and looked at different things than they did.

Fast-forward to today. How much touring and inspecting happens in your company? How often does your boss just walk around, seeing for himself or herself what’s going on?

I was thinking of this today during a conversation that touched on building a new data centre. Thinking about the process of building a data centre, there are clearly a huge number of people involved in making it happen. But how many of them have actually been in a data centre, toured one, seen what physically goes into it and makes it work? Has the CIO ever stepped foot in one? Sure, you certainly don’t need to tour a data centre in order to develop a list of requirements and build it, but wouldn’t it help? I didn’t need to walk through the engine room of my boat to know how to do my job; but seeing all of the operations first-hand dramatically affected my understanding of the whole system and certainly affected how well I did my job.

You may be thinking how any of this applies to prediction markets. Well, prediction markets can help fill this gap.

Many managers don’t Manage By Walking Around, despite the benefits. So they fail to pick up on the little details that their subordinates, and their subordinate’s subordinates know about a new product, the effectiveness of marketing, or progress toward a milestone. All of these little insights, if left unused, add up to a massive loss of insight. A prediction market sums all of these little bits of knowledge, of experience, of industry background and tells you with amazing accuracy exactly what you want to know about your business problem.

So, while I firstly advocate Managing By Walking Around, many managers will find that they can gain many of those important insights through a properly-run prediction market.

I’m sorry it’s been a while between posts, but I was knocked down last week by a particularly violent case of food poisoning… I now vow never to eat at the local Chinese takeaway again!

There is a new feature here at Mercury’s Blog, e-mail subscriptions. That’s right! If you don’t understand RSS and feed-reading (which if you do, click here to subscribe), you will instead get an e-mailed copy of my post(s) each day I make them. If this sounds interesting to you at all, please click here.

———–

I’d like to mention a book here that I think could be valuable for everyone that’s not a professional economist (and even those that are!), and wants to understand the nature of markets. It’s called Reinventing the Bazaar: A natural history of markets, written by John McMillan of Stanford’s Graduate School of Business.

Let me say first off that there’s a good reason this book is rated five stars by Amazon readers; it’s absolutely excellent! It’s written for a wide audience, and is able to include both theory and relevant examples that everyone should find familiar. What it aims to do is lay bare the spectrum of modern markets: how they work, why they work, why some don’t work, and the markets’ relationship to government. I recommend it here because it lays a fantastic foundation for why prediction markets are such incredible mechanisms for forecasting and revealing information.

The book opens not on the New York Stock Exchange, but in the Dutch city of Aalsmeer. It is here where a huge flower marketplace is operated, and where millions of dollars worth of flowers are purchased every day in an extremely efficient manner. That the book doesn’t focus solely on standard financial markets such as the NYSE is to its credit. It talks about real-life markets throughout the world-wide economy, such as a Middle East bazaar (hence the name), fishing grounds, the deregulated California energy markets, the famed auction of additional slices of the US frequency spectrum, and many more. More importantly it examines what makes those markets work, and what is required to make those markets work.

I’m also very impressed that there is no ideology here. Some authors writing a book on this topic could tend to write either a “markets can save the world, they just need to be freed from regulation” type book, or a “markets are the cause of all the worlds’ problems” type book. It is neither, and instead discusses how markets are indeed a powerful construct, but can be both well-built and resilient or poorly-built and unfair.

I’d like to quote a few paragraphs here that apply very directly to the prediction markets arena:

-

(In a discussion about central planning and control)

A worker on the production line might observe quality defects that are apparent only on the shop floor, or a machine that is sometimes idle, or a surplus stock of raw materials that could be used. A middle manager might be aware of engineering problems in a new process, or a way of reassigning workers to increase productivity. Salespeople in the field learn about demand for the firms products. Much of the information about demand and costs that the top management needs for planning must come from below. Knowledge that is valuable to an organization is acquired by people - at all levels of the organization, including the lowest - as a by-product of their day-to-day duties.

Why does it matter that the source of the information is separated from the decision-making responsibility? People take advantage of any special knowledge they have acquired. Dispersed information within a hierarchy makes conflicts of interest inevitable. Information becomes distorted because of people’s incentives to exploit any informational advantages they have. “People are reluctant to share their information,” observed the head of a large French company. “Managers in particular seem to think it gives them extra power.” Middle managers “have an interest in husbanding information - and the power that goes with it.

-

For a market to function well, you must be able to trust most of the people most of the time; you must be secure from having your property expropriated; information about what is available where at what quality must flow smoothly; any side effects on third parties must be curtailed; and competition must be at work. A multitude of mechanisms sustain these five key requisites of effective markets. Your trust in your trading partner rests on both the formal device of the law and the informal device of reputation. Your property rights are protected by the law and, in the case of your investments, by regulation. For you to be able to take your business elsewhere, there are channels for the flow of information, so you can locate others to deal with, and there are few impediments to starting up and running firms.

A market’s design, supporting these features, may evolve from below or be imposed from above; usually there is a bit of both. A workable structure provides rewards for good behavior and checks and balances to deter bad behavior, so people act honorably while following their self-interest. When markets are well designed - but only then - we can rely on Adam Smith’s invisible hand to work, harnessing dispersed information, coordinating the economy, and creating gains from trade.

-

So to summarise, this book is an excellent resource for anyone that’s interested in how prediction markets work. Not just the markets themselves, but all of the supporting structures that markets require to operate properly are very well covered in this book.

I highly recommend it for anyone reading this blog!

Companies that are looking to use prediction markets in their business need to build a business case for doing so.  In order to build that case, they need to determine what information can a market efficiently provide, and how can that improve their strategy and day-to-day operations?  I’d like to address three main question areas that prediction markets can solve, and how you can start developing a business case based on the results of the improved forecasting.

The first, and perhaps most powerful, is project management.

Every project has various key milestones, and those on the critical path are even more visible to and closely watched by managers and executives.  Most forecasting to date is largely self-reported by team leaders.  They are largely meant to be as honest as possible in reporting project status.  In reality, this is highly dependent on the company’s culture, and in far too many businesses, project status is a commonly-known lie.  (Unfortunately, this also means that it can be a politically dangerous kind of market to implement, since the project managers that tell these lies usually don’t want to be exposed!)

The business case for a prediction market in project management should be calculated based on the effects of slips in a project’s schedule.  For projects or products where sales are front-loaded (such as films, major software releases, ground-breaking new drugs, and the like) a project’s slip into the next fiscal quarter or next fiscal year could cause a serious impact in potential earnings.  If that information was available weeks or months earlier, how would that affect sales, reputation in the marketplace, and other factors important in your industry?  How much does it cost to plan for a major product roll-out in January, only to have to cancel it all and do it all over again in April?  These are the questions that can build a business case for a project management prediction market.

The second type of market quantifies your industry.

In many industries, this means defining the cost of goods needing to be purchased, the quantity of goods or services to be sold, or the price potential of the goods or services to be sold.  In some industries, this could mean quantifying the user growth of a given product.  (Such as Google’s prediction markets, “How many users of [Google product] will we have at the end of Q3?”)  This is the classic case written about by HP’s research center, when they asked their traders to predict sales figures for various products for the next quarter.  It turned out they were measurably better than the forecasting system they had been using.

Quantifying a business case here is generally more computational, and therefore more straightforward.  What are your costs associated with inventory of products that you can’t sell because you thought there was more demand?  How much less revenue do you earn because you have to discount products to move them out the door?  Alternately, what are your costs associated with paying extra to ramp up capacity because you didn’t realise the demand was as big as it was?  How many lost sales did you experience because of empty shelves?  These numbers are concrete and can hit your company’s bottom line.  They can sometimes be the easiest way to prove your business case.

The third type of prediction market I want to discuss quantifies risk.

Each industry and company has its own measures of risk.  Perhaps it is a measure of exposure to outside influences, such as bad weather for golf courses.  It might even be a measure of product quality, where it is a significant risk to reputation.  (An example could be a market for quantity of high-risk bugs found in the first month of Microsoft Vista deployment.)  Your company likely has key metrics that don’t directly relate to sales, costs, or projects schedules, even though they are key data points for your business.  The results of some of these markets could also be valuable inputs for sales forecasts, and for more advanced companies there is the option of operating combinatorial markets.  (I will be posting more on this later.)

It can be challenging to build a business case for these types of prediction markets, but is certainly still possible.  One method is to take a good-weather/bad-weather approach.  Most businesses plan for good weather, and have contingencies for bad weather.  What if disaster struck, and the actual outcome was worse than even your bad weather scenario?  What would be the financial impact for your company?  Again, how much is that information worth weeks or months in advance?  Those are the calculations that can prove the need for a proper prediction market for these metrics.

In summary, these are three main ways that prediction markets can be implemented in your company.  More importantly, each shows the basics of how to calculate a business case to prove their worth.  Please feel free to leave comments below, or contact me directly with any questions.

Prediction markets are powerful tools in business forecasting. However, their use and the business problems they solve will vary between industries. I’d like to discuss a couple of scenarios where prediction markets can be particularly valuable.

You’ll notice an interesting phenomenon when it comes to early adopters of prediction markets; an inordinate number of pharmaceutical companies are listed. They are the best example of the type of company that needs prediction markets the most: industries where massive research and development costs are spent before any revenue is generated. Pharma’s have a real need to understand which potential drugs and research areas will be successful so they can focus their resources. It’s a similar situation for movie studios, which is why HSX likely has the client list that they do.

Does your company fit that mold? Do you need to make a massive investment before any revenue is generated? Then a prediction market is a good solution for you. Early identification of product popularity and schedule will easily make your business case. If you’re banking on a fall release for a product you expect to be very popular for the back-to-school and holiday market, you need to understand both if it will be ready in time, and if it’s going to be as good as the project manager is reporting. Would you rather know in June that you have problems, or experience the stress of not meeting quarterly earnings targets with a sub-standard product?

That isn’t the only type of business that can be well-served by prediction markets. Mature industries with key variable costs or demand will also see business success through a prediction market tool. As an industry matures and consolidates with fewer players, certain variables can dramatically affect profitability. For example, Hewlett-Packard has used prediction-market-type tools to forecast RAM prices, and has seen forecasting error drop from 4% to 2.5%. They may not seem significant, but multiplied over the sheer volume purchased each year can be a tremendous cost savings. This is exactly what Arcelor has done in the steel industry, as well. Their fairly small prediction market has demonstrated consistently better forecasts in quarterly demand than previous forecasting methods.

These are only two of many different types of businesses that could be well-served by prediction markets. I hope to provide examples of others in the future. As always, please feel free to comment here or contact me with any questions!