Comments on: Random Numbers Out of Triangular Distribution http://rule-of-thumb.net/2008/09/06/random-numbers-out-of-triangular-distribution/ A project manager's log book Mon, 09 Nov 2009 21:23:52 -0700 hourly 1 http://wordpress.org/?v=3.0 By: Sašo http://rule-of-thumb.net/2008/09/06/random-numbers-out-of-triangular-distribution/comment-page-1/#comment-44 Sašo Fri, 26 Sep 2008 08:29:48 +0000 http://rule-of-thumb.net/?p=51#comment-44 Timm, of course I agree with that domain knowledge improves estimates, but at different stages of the project you can expect different levels of expertise. At the initial concept of a project you don't know yet, whether a task of hitting a golf ball will be included in a plan or not. Example. You're planning Olympic Games and you need to fire the olympic flame in the opening ceremony. It might be by fire golf ball hit by Tiger Woods or it might be in any other way. Before you come to this level of detail, there will be hundred checkpoints where different level of estimates will be required. When you apply for the host of the Olympics, you need to provide a cost estimate. You don't ask Tiger Woods how far can he drive the ball at this stage. Steve McConnell talks about <a href="http://www.construx.com/Page.aspx?hid=1648" rel="nofollow">The Cone of Uncertainty</a>. It charts the level of uncertainty depending on the time / phase into the project. The bottom line is that every estimate should be made by three points. Uncertainty involved in the estimate should be defined by the width of the range. Much more important than the width of the estimate is the probability, that the outcome will be inside the range. Using percentile estimates expands the ranges so that they at least to some extent cover for "out of range" errors in estimates. Timm, of course I agree with that domain knowledge improves estimates, but at different stages of the project you can expect different levels of expertise. At the initial concept of a project you don’t know yet, whether a task of hitting a golf ball will be included in a plan or not.

Example. You’re planning Olympic Games and you need to fire the olympic flame in the opening ceremony. It might be by fire golf ball hit by Tiger Woods or it might be in any other way. Before you come to this level of detail, there will be hundred checkpoints where different level of estimates will be required. When you apply for the host of the Olympics, you need to provide a cost estimate. You don’t ask Tiger Woods how far can he drive the ball at this stage.

Steve McConnell talks about The Cone of Uncertainty. It charts the level of uncertainty depending on the time / phase into the project.

The bottom line is that every estimate should be made by three points. Uncertainty involved in the estimate should be defined by the width of the range.

Much more important than the width of the estimate is the probability, that the outcome will be inside the range. Using percentile estimates expands the ranges so that they at least to some extent cover for “out of range” errors in estimates.

]]> By: R. Timm http://rule-of-thumb.net/2008/09/06/random-numbers-out-of-triangular-distribution/comment-page-1/#comment-42 R. Timm Thu, 25 Sep 2008 19:04:53 +0000 http://rule-of-thumb.net/?p=51#comment-42 Great Blog. I ran across this looking for VBA Macros to Export Project Data to Excel. I disagree with the example claiming that folks are bad cost estimators / PMs if they cannot create an upper and lower bound around information they know nothing about. You have to do the research first by communicating with the engineers / experts. If you ask a golfer to put a 3 point estimate around how far Tiger Woods will drive the ball you will likely get a decent range. If you ask someone who has never heard of the game they aren't going to provide a very good estimate. Great Blog. I ran across this looking for VBA Macros to Export Project Data to Excel. I disagree with the example claiming that folks are bad cost estimators / PMs if they cannot create an upper and lower bound around information they know nothing about. You have to do the research first by communicating with the engineers / experts.

If you ask a golfer to put a 3 point estimate around how far Tiger Woods will drive the ball you will likely get a decent range. If you ask someone who has never heard of the game they aren’t going to provide a very good estimate.

]]>