Monday, August 27, 2007

Keeping everyone happy (almost)

Back in May, 2006 when we introduced the ability for people to create and manage their own prediction markets, we knew we had a challenge on our hands. Outside of a couple successful outliers in the public domain, a prediction market had traditionally been a fairly difficult concept to grasp in and of itself for the pure layman. Now we were trying to provide the ability for people to ask any question they wanted without intervention from economists, academics, or consultants. Ask anyone who has built software or any consumer product for that matter what their biggest challenge is, and often times it's making things easy and understandable. And even with all the lip service to usability, most times they fail, e.g. enterprise software, smart phones, Windows. :)

Over a year and a few thousand user-created markets later, we are still living that challenge. Most people find it very easy to create a market. They are, in fact, surprised at how easy it is and we appreciate the subsequent fan mail. But as with any application that is widely used, you quickly learn what people are not understanding as they go through a process. You also learn the patterns that arise of the types of markets people run, the requests people have, and hopefully react to them in a productive way.

Conventional wisdom (and user requests) often says/asks to keep adding more functionality. This might be good for your marketing so you can have a huge list of features, but soon you have something that is no longer very usable and a shrinking percentage of your user population is happy with your application. New users are especially bewildered. Instead, we've tried to take the approach of simultaneously having a very high bar of introducing new features in the publisher while actively trying to remove options when we notice usage patterns.

For example, the most widely run market type on both our public marketplace and our client's marketplaces is a simple yes/no question. "Will X happen?" In previous iterations, this market type was created the same way that a question with many answers was created, like: "Who will win the Super Bowl this year?" Thus most people's natural inclination was to ask the question and create two stocks: one for "yes" and one representing "no." The problem with this, however is the way trading works in Inkling, a trader would find themselves in a situation of being asked to go long or short on "no." Having a short position in "no" is like saying "yes," so this was kind of redundant and pretty confusing for traders.

Instead of simply adding a bunch more help text in the latest iteration of our publisher, when we ask our users as the very first step to publishing a market what kind of question they are asking, we explicitly name a "yes/no" type question. Then we automatically set up a single stock for them to trade in and don't allow them to create any more stocks. Removing the flexibility that once existed will surely draw a few emails from folks wanting to do something slightly different for a yes/no question, but in this case we'd rather have people running this type of market correctly. And for the vast majority of people, it's now much quicker and easier to set up this type of market.

This is only one of many changes we've made recently to the market publisher. Let us know what you think and we'll try to successfully keep up the perpetual balance between more functionality and usability.

No comments: