Here’s an excellent introductory article on the wideband delphi estimation technique. Typically wideband delphi is used to estimate software development efforts, but can be used in almost any domain of work. This method might be applied to estimating effort for items in a work list at either a project level or tasks in an iteration work list. The way it is described, it sounds fairly heavy-weight, potentially taking several hours for a relatively small list of work items. However, it is worthwhile for process facilitators and product owners to be aware of these sorts of methods if problems with estimating occur in a project team. The wideband delphi model is related to the Delphi method.
Agile Work and Scrum use a very different method of estimating work that has one very important and incredible property: people don’t have to get better at estimating for them to get better at committing to work!
The basic idea can be thought of as “commitment velocity”. The method here is to use an arbitrary unit of effort that the team applies to estimating all tasks relative to each other. At the start of an iteration, the team can use any collaborative method (including wideband delphi) to come up with estimates for tasks. The team sums up the estimates for all the work of the iteration. Then, at the end of the iteration the team looks at the estimates for all the remaining undone work (if any) and subtracts that from the start of iteration sum. This is the Commitment Velocity. Now on their next iteration, they do their estimation work again, relative to the tasks in the previous iteration. If the sum of the estimates is larger than the Commitment Velocity, then the team needs to de-scope or find more efficient solutions to their work. This process usually converges after only three iterations (assuming: constant team composition, constant iteration length).
This method is taught in detail in my Agile Project Management courses, and there is a little bit more information here: Planning vs. Commitment. The agile estimation method of Commitment Velocity is an application of the Central Limit Theorem from statistics.
Wikipedia now has an article on Wideband Delphi. The article points out some interesting potential problems with the wideband delphi method including corruption of the group doing the estimation and suppression of minority opinions. I know that in my experience as a coach, I have seen numerous occasions of both types of problem with estimation processes. The fact that wideband delphi is susceptible to these problems is more a factor of the human side of things than the method itself.
Interestingly, the wideband delphi method of estimation is very similar to the “Planning Poker” method of estimation used in many agile and scrum project environments. This method is also covered in the Agile Project Management courses.
I have written a detailed article on Agile Estimation using Planning Poker that describes in detail how best to estimate effort on a Product Backlog. One important detail is that the method requires that a group doing estimation come to a consensus agreement just like Wideband Delphi. Some people who describe Planning Poker allow for non-consensus estimates to be recorded. The repeated iteration until consensus is reached is one of the reasons why both Wideband Delphi and Planning Poker are so powerful.
Other links to wideband delphi: