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!

First of all, welcome to the new subscribers of this blog.  If you aren’t a subscriber, it’s very easy to do.  (You can use any RSS reader you like, even Firefox or Safari; just click here.)  Please feel free to comment, or alternately e-mail me with questions you’d like to keep private.

I’ve been thinking about prediction markets recently in terms of their usability.  Alex Kirkland has done some interesting work and research in this space, and I’d recommend his blog if you’re interested in more in-depth analysis.  For a prediction market to be successful, it should be both easy to trade, and easy for traders to do complex operations.  These are two ends of the same scale, and unfortunately, there’s little between them.

Kathy Sierra at Creating Passionate Users talks a lot about user experiences.  One of the things that I’ve picked up from her work is that it’s easy to make something simple for users, and it’s easy to make something powerful for users.  The hard (but key) part is to develop a solution that is easy for beginners to understand and get started with, while also allowing them to gradually discover more and more powerful features and turn them into expert users.  Currently, the prediction market industry does both ends of that spectrum well, but the middle is ill-defined.

Why do I bring this up on a blog that’s focused on using prediction markets in modern business?  Well, it’s because user interface design of prediction markets is really important.  Most people when confronted by the user interface of something like Iowa Electronic Markets (IEM) will simply not trade at all; a reaction like this will certainly hurt the quality of your markets.  However, the market potential and power behind something like the IEM is strong, and your company may want those aspects in its prediction markets solution.  In that case, you’ll simply need to train your employees to use a more complex system, (and likely hold some “fun” markets for them to get used to the ideas and terminology of trading).

Alternately, Inkling is a very easy interface to understand, and this may be your overriding concern.  You likely will need little or no training for employees to use it and understand the concepts.  However, for those employees that can understand the concepts of the trading “book” and limit orders, an interface like Inkling’s may be too basic.  Since most trades and prices are set by a minority of highly active traders, this could prove difficult as they aren’t able to effectively use their more advanced skills.

So where does that leave your organisation?  Well, as always, for a market to be well-designed and therefore useful, it needs to match your company’s culture well.  Is most of your potential trading population too busy to learn trading skills or too scared by complexity?  Or is most of your potential trading population very competitive and math/finance oriented?  Where your organisation falls on that spectrum helps dictate your needs when in comes to choosing a software vendor, but also in the training you may need to make prediction markets successful.

Any of the major software packages could work for any organisation; but good design and user interface is an important element to consider when creating your prediction markets.  You don’t want your employees to not even try the markets because they look complex and don’t understand them, but neither do you want advanced, market-moving traders to get frustrated at getting held back by too-basic interfaces and technology.

In your aim to get a high-performance prediction market, one important aspect is getting as many people trading as you can. (In order to come as close as possible to the so-called “Efficient Market”.) But many companies are rightly concerned about how to get their employees motivated to trade on their internal prediction markets.

Emile Servan-Schreiber made a good post to the Prediction Markets e-mail list last year, and mentioned what he called the “Three R’s: Rewards, Recognition and Relevance.” Some combination of these elements is needed to create the incentive structure that gets people to trade. It is important to remember that each company’s culture influences how much of each “R” is necessary in their specific environment.

I learned as a young Navy submarine officer that competition is a fantastic incentive. The results of any type of competition feed the Recognition incentive factor. Bragging rights will be fought for between colleagues and friends, and will likely be the source of at least a few water-cooler conversations. Some sites, like Hollywood Stock Exchange and The Washington Stock Exchange, explicitly list the ranking and net worth of all or many individual traders. (It’s how I know that I did a bit better in trading election predictions than Professor Justin Wolfers of Wharton.) I highly recommend this feature, as it easily shows each trader’s performance in context to their friends and colleagues, and could serve as a touch-point to initiate conversations about market topics. All of this will contribute to both the market’s and organisation’s success.

Other companies go further in the Recognition factor. One client of NewsFutures has a senior executive present an award to the top trader in their internal markets. Again, I highly recommend this to those companies which have a senior sponsor of a prediction markets project. Small ceremonies and newsletter articles can serve to advertise market prediction successes, further encourage participation, and demonstrate that company leaders value the results.

Relevance is a factor that should not be overlooked. If your employees care about a topic, they will be much more ready to trade in a prediction market on that topic. In the research I completed last summer (see our website for details and the paper) I needed no incentives whatsoever to get participants to trade. The reason for this was that the markets involved an area in which traders were particularly passionate; this passion was so relevant to their life that they needed no promise of rewards or real recognition. (At the time, the Inkling Markets software I used only listed the top 5 traders overall.) How does this apply to your company? Well, if you create a market on something interesting, something that your employees have knowledge of and care about, something that matters in their life, you will have gone a long way to ensuring that you get sufficient participation. There is a balance to be struck between the specificity of a question and the relevance to a broad population.

Finally, Rewards are the most common incentive to traders. This unfortunately can become a touchy subject, particularly in large organisations. It is surprising how little prizes go a long way in the rewards department. Google awards t-shirts and similar gifts to top traders. Sometimes all it takes is pizza and drinks! However, some companies give large cash prizes. (See the comment on this post, hat tip to Chris Masse of Midas Oracle.) What a business gives as rewards to traders really comes down to the company culture. Market project leaders should check their Human Resources policies, and think about how their reward structure can match the spirit of compensation in their company.

As Emile mentioned, each of these elements is important in creating an incentive structure. A well-designed set of incentives will ensure good participation in your company’s prediction markets, which should lead to well-deserved success!

Companies are rightly concerned about getting good results from prediction markets. In particular, some important and strategic decisions have a very limited group of employees that know enough to trade in a prediction market. This is sometimes because the company is small and there aren’t that many employees in the first place, but often the specific issue at risk requires so much background and experience that participation by a wide group is implausible.

For example, product manufacturers keep track of sales of their products and are keenly interested in predicting future sales. However, the data is key to their operations and closely held within the company. Only a small group of people (often in Finance, or top Operations personnel) have access to past and current sales figures, and thus have the requisite background information to forecast future sales. This could be seen as a hopeless situation.

That scenario is in fact far from hopeless. Our pre-conceived notions of an efficient market typically involve screaming traders in a pit of chaos. Many people believe that it takes a large group of traders to get good quality predictions. It turns out that markets can also work on a very small scale. This issue was the focus of the first research project completed by Mercury Research and Consulting. (Find the paper on our website; selected for publication in the Journal of Prediction Markets.) The question we asked was: how low can you go?

As it turns out, pretty low. Our research showed that as long as just 15 traders participate in a market, the results are well-calibrated. This is an important finding because it helps to answer two questions:

* Can a prediction market accurately answer this question?
- and -
* Do I need to alter incentives in order to ensure a sufficient number of traders?

You will likely find that in most situations you can get the fifteen or more traders necessary to make an efficient market, though it may require some creativity in diversity of participants. However, you may need extra incentives in order to get them to trade. For example, if there are only 20 people that can reasonably participate in a given market, different tactics may be required to ensure that most of them participate. (Larger prizes, more significant internal recognition, etc.) If less than fifteen people can address a given area of risk, it may be suitable to find alternate means of forecasting than a direct marketplace scenario.

So does your company need to stick to “safe” topics on which anyone and everyone in the organisation has an opinion? Not at all! Specialist questions on key risk areas can be just as accurate with just a small number of participants, as low as fifteen. The fewer the number of people involved, the more important it is that they are independent of each other, but they are still fully able to accurately predict an outcome.