SELECT * FROM Celko – March 2010

I am sitting here typing in a speaker’s give-away shirt that I got at a trade show. In fact, most of my causal shirts are swag from trade shows. I have mouse pads, corkscrews, pens, clocks, and lot of other things; I will never have to buy a canvas tote bag. Most of the swag is worn out by now. Some were clever and useful; some were useless.

I am looking at an interesting tool that I got for free at a trade show. It is a deck of playing cards from a Visual Studio promotion. This is not the usual Poker or Pinochle deck with a company logo on the back. It is called “ Planning Poker® ” from Mountain Goat Software or from Agile Hardware at who sell the decks on their websites.

The instructions are a bit vague, but I am going to call this tool a game. Each deck contains enough cards for four estimators to each hold cards with the following values: ?, 0, ½, 1, 2, 3, 5, 8, 13, 20, 40, 100, and 8. If you want to buy more sets for more estimators, you can. In fact it might be better with more expert opinions.

Here are the rules from the rules card in the deck:

Each team member is given a set of cards.
One person reads the item to be estimated.
The team and customer discuss the item.
Each team member privately selects a card representing his/her relative estimate.
After all have chosen a card, everyone shows the chosen card.
If all estimates match, that item’s estimate is complete.
If estimates are not the same, the group discusses the differences (focusing on the outlying values).
Repeat until consensus is reached.

If you are old enough to remember when the World Future Society (WFS) was in the news, you will see that this is a version of the “Delphi Method” that WFS made popular. The important part is in step #7. The reason for looking at the outlying values is that experts who are at the extremes probably know something that we mere mortals do not. The idea has been picked up by Agile and scrum programming teams.

The more traditional method is to let a single project leader estimate the effort by himself and then sell it to the customer. The customer would come back with a counter-offer. We were negotiating and not estimating.

The guy who estimates low [0, ½, 1] effort with his card might know that we already have a package in house to do this job. I have an example of a government agency that already has purchased ArcGIS and needs to do redistricting this year with the 2010 Census data. I see a quick solution that involves a training class or a consultant who can set up the package and do the work.

The guy who estimates high or impossible [40, 100, 8] effort with his card might know that this problem is NP-Complete, physically impossible and/or illegal. The traveling salesman problem is the classic NP-Complete in business. I cannot think of a physically impossible project off hand, but I can think of a lot of illegal projects.

I have no idea why that particular set of values is used on the deck. I can figure out the “?” (Don’t know anything about the topic), “0” (we already have done it) and “ 8” (impossible). But why is there is Fibonacci series in it? Was there a reason for this scale? I do not know.

Another website added some changes to the basic rules. Some of their modifications were:

Do not have any business members there, but call in the requester if needed.
Re-estimate until within one card value of each other, or take the median value if there’s a majority.

There is also a blog on follow-up games that can be played with “story cards,” which are forms that give a short version of that which is to be estimated. The story is given a name, the role of the requester, the goal of the project and its purpose (“As a third world dictator, I want to conquer Elbonia to stay in power at home.”).

There is an old movie in which two crooks are playing cards to kill time. The first thug asks, “What do you have?” The second thug lays down his hand and announces, “Full house. What do you have?” The first player lays down his hand and announces, “Gin!” and they stare at each other. The expectation is that most of the estimators have some common background and are playing the same game. The median is probably good measure of central tendency on an ordinal scale that has such irregular steps. But what does it mean if the median is arrived at by a polarized distribution?

What if we have a voting paradox in the estimators opinions? Look at the references at the end of this article for four books on voting paradoxes, but the bottom line is that there is no voting system that satisfies simple criteria for fairness. Consensus is a nice goal, and we can certainly aim for it on a project team. But there is no magic.

References for Voting Theory

MenuMenu

SELECT * FROM Celko – March 2010

Joe Celko

MenuMenu

Share this post

Joe Celko