A common problem when using Planning Poker is some people who are more opinionated or argumentative can dominate the estimation process. This usually happens because the standard method for using Planning Poker is to make independent estimates and discuss until all the estimates converge. Crowd wisdom may be able to offer us a way out.
The idea behind crowd wisdom is that together we are smarter than any one of us. This is the central idea behind crowdsourcing. Any one person's contribution is checked and improved by other people. Consequently, these reviews are just as important as the initial contribution. Wikipedia, open source software, and remix culture are classic examples of this.
In the book "The Wisdom of Crowds," author James Surowiecki gives a number of examples. At a county fair, a prize was offered to the person who could guess the butchered weight of a cow. The average of all the guesses was closer to most individual guesses, including some cattle experts. When a submarine was lost at sea, an assembled team of experts made a number of best guesses for the location of the vessel. Ultimately, the submarine was found within 220 yards of the aggregation of these guesses. Pretty impressive, considering the size of the North Atlantic.
In a nutshell, Planning Poker is used to generate estimates for software development task items. In agile terminology, it's used to allocate story points to User Stories. Typically, the number of different possible values for the estimates is constrained, e.g. 1, 2, 3, 5, 8, 12, 20. Each team member, comes up with their own estimate. All of these estimates are revealed simultaneously using Planning Poker cards. More often than not, these are a variety of estimates. The recommended solution is to discuss these differences in estimates to reveal assumptions. Independent estimates followed by simultaneous sharing is repeated until the group reaches consensus. (For software engineering wonks, this is just a twist on the Delphi method.)
Surowiecki identified four elements that are necessary to use crowd wisdom successfully. These are diversity of opinion, independence, decentralization, and aggregation. To apply crowd wisdom to planning poker, we would need to do the following.
Diversity of opinion means that each person is allowed to form their own opinion, even if it's based on private information or an eccentric interpretation of established facts. I believe that this is already built in to Planning Poker, by having a group create the effort estimates.
Independence means that people's opinions aren't determined by the opinions of others. Planning Poker already allows for this, because individuals come up with their own estimates separately before sharing them with the group.
Decentralization means that people should be allowed/required to draw on specialized or local knowledge. In Planning Poker, you'd have a cross functional team do the estimating. This is a good practice anyways, so should not be a big change.
Aggregation is the mechanism for turning the individual, private judgements/opinions into a collective decision. In Planning Poker, this should be relatively straightforward, since we are working with numbers. A reasonable mechanism would be to take the average of the numbers.
This last element diverges the most from conventional Planning Poker. The rationale behind coming to a consensus estimate is that an average isn't a good estimate for any one person. Just as the average family has 1.7 children, but no one family has 1.7 children, because children only come in whole units. Nevertheless, this is a supposed to be the strength that underlies crowd wisdom.
My guess is a hybrid would probably be best. The discussion after the estimates are shared are invaluable for improving the team's understanding of the software being developed. But the average of the initial estimates is the best guess of all.
I'd be interested in hearing from anyone who has tried using crowd wisdom with Planning Poker. Did it work for you?
No comments:
Post a Comment