Pages

Tuesday, September 16, 2008

Applicability of Decision Market technique in projects

Decision Market a.k.a prediction Market  is one of the techniques being used by many universities, governments, research institutes, finanacial companies to predict the future. These predictions have helped many organizations in avoiding obstacles, making better decisions, and in turn becoming  more profitable.  I am of the opinion that this concept can be used in various aspects of software development too.  For ex:  Iteration Planning, Release Planning, Product Planning, etc

This is how I am envisioning applying prediction market concept for release planning

1. Invite all the stakeholders (from the management team), key people from the project team who have good experience on the project

2. Allow them to go to a hall/room and cast vote on a preselected key concept like "the probability of having the release X on time".

3. Ensure that no one discusses their answers with each other until they come out of the room.  This would ensure that people don't get influenced by others decision.

4. Let each person also give the reason while casting the vote on why there could be a delay in the release. This would provide an opportunity for the stakeholders to correct the mistakes and avoid any unplanned obstacles/impediments.

5. Announce an incentive for the people who are making the accurate predictions. 

At the end of the release, identify all the issues faced during the release, and also the actual release status. Whoever has predicted the right things should be given the promised incentives.

More info about image can be found here

Even though the original prediction market talks about punishing the wrong predictors, I think it may not be a good idea in the software development senario. If  one is betting in an open market, one would not really care if the other person is winning or loosing. However in a closed environment like  a project team, punishing one person over the other, leads to animosities and things like that.  

In fact, Planning Poker based estimation technique could be considered as an avtar of prediction market.  During planning poker, the team is given an opportunity to provide their view on the effort needed in completing the task. View from the majority is taken and a consensus on the estimation is drawn. 

Agile 2008 conference had a session on estimation technique using decision markets. Since I didn't attend the conference, am not sure about the content covered during the conference. 

Even Google uses the decision market technique to make key organizational decisions. 

3 comments:

Anonymous said...

The bottom line behind outsourcing software development is definitely cost cutting. And when one gets the best at a comparatively cheaper rate it is nothing less than icing on the cake. Most advantageous and efficient software development is possible at a cheap rate today because surprisingly they come from some of the poorest countries in the world, as for example, India. After all a few years back who would have fancied that the country with the highest rate of poverty and which is ranked among the third world countries would come up with world class first rate software development companies http://www.infysolutions.com.

Brian G. Keen said...

Great thoughts!! You ideas of using the concept of prediction markets are very insightful. I agree that trying to find ways to channel good ideas will only result in optimal results.

I will definitely add you to my Google reader and check back on your thoughts.

Sincerely,

Brian Keen

Mark Stringer said...

This is a very interesting area. I think the whole business of the psychology around estimation isn't talked about enough in software development. People are often forced into making over-optimistic estimates and then forced to stand by them, which can be psychologically very damaging.

One problem with prediction markets (and planning poker) might be that it is perceived to take a long time to estimate all the work involved on a project. I recently came across the notion of using the geometric mean as a way of producing a rough estimate for almost anything in the book "Guesstimation" by Lawrence Weinstein and John A. Adam.