Mike Cohn, of Mountain Goat Software, has a new agile book out that focuses upon planning and estimating. He has made some of it downloadable. In the Techniques For Estimating chapter he describes why estimates are usually way off the mark. I found the accuracy curve kind of interesting. The accuracy of the estimate increases as the effort put forth to come up with the estimate increases (I know... no sh&* Sherlock). But the unexpected part was that accuracy peaks and then declines with increased effort. I guess it is like over studying for a test kind of a thing. So the author came up with a fun way to increase the effort to the sweet spot that results in a more accurate estimate -- Planning Poker.

Collaborative estimates being more accurate, you begin by bringing together a team of subject matter experts from various parts of the application. These experts should range in job roles covering all aspects from business analysts, developers, testers, database administrators etc. Each player will need their own deck of cards. The cards have numbers (1, 2, 3, 4...) written on one side. The dealer will explain to the players what the number means. For instance it may represent 1 hour or 1 day or 1 week or on a scale of 1 to N with a simple (1) to complex (N) effort etc.They hold the deck face down and listen to the dealer. The dealer reads a brief but succinct description of the problem. Each player then selects a card from their deck that represents the estimate. Each player places the cards face down in front of them until all players have selected the card they feel best represents the estimate. Then each player flips the card over. The dealer examines the values. The dealer will ask the lowest and highest estimate player to explain why they feel that way. The dealer will make use of a 2-minute sand timer to keep the answers short. Then the cards are returned to the deck and the process is repeated until the players have clumped around some value. He says 2 or 3 rounds per problem is about right.

The author explains that by getting estimates from the various view points to converge you get a highly accurate estimate. And by using a fun team activity like playing Planning Poker the effort is not too light and not too heavy ~ just right!