5 Steps Towards Simple and Effective Project Estimation

Name: 5 Steps Towards Simple and Effective Project Estimation
Type: Process
Status: final
Version: 2008-07-31
Sources: my practice, Niels Malotaux (see http://www.maloteaux.nl/), www.iit.edu/~it/delphi.html

Gist: Every once in a while we start projects, or are about to start one, and somebody asks "how long will it take? " (The better question would be "how long will it take to have benefit A and B?", but that's another story.) The following simple 5 step process provided sufficient precision while not being too time-consuming. Additional benefit of the process: A group of people did it, most commonly the group that as to work on it anyway.

Entry1: You need to have clear objectives.
Note: If you don't have clear objectives, you will notice in Step 4.

S1: Sit down - alone or better with one colleague - and break the work down to one or two dozens of task-chunks. List them.
Notes:
* If it's a well-known problem go find someone who did it before!
* If it's a rather small project you can break it down to single tasks, but you don't have to. Roughly 10-25 list items.
* Best effort. Don't worry too much about forgetting things or about the uncertainty because you don't know every constraint.

S2: Distribute the list to the people who will work on the project or know enough about the work to be done. Ask for their best estimates for each chunk. Allow for adding items to the list.
Notes:
* It's a good idea to limit the time for the estimations, especially if you have external staff that is paid by time and material.
* Don't go for ultimate precision, you won't achieve it anyway. Variances will average.
* You're much better off with an initial, not so precise estimate and early, constant calibration.

S3: See where estimates differ greatly. Discuss or further define the contents, not the estimates.
Notes:
* Everyone's an expert here. It is not about right or wrong estimates, but about understanding the work to be done.
* In the original Delphi process there's an element of anonymity. I can't follow that advice, for the purpose of this process also is to build consensus.

S4: Iterate S2-3 until sufficient consensus is reached.
Notes:
* Niels says, not more than one or two iterations are necessary.
* I say, it sometimes needs a dictator approach.
* Careful here, make sure to discuss the content of the work. Also helpful to further specify the objectives!

S5: Add up all task-chunk's estimates: that's how much work the project is.
Note: The single most useful approach to fitting the work in a tight schedule is to work hard to NOT do things that later will be superfluous.

Related Posts:
Earned Value Analysis for Feature-Burndown in Iterative Projects
Measuring and Estimating one's Velocity
Calculate Process Efficiency
Quantum Mechanics, Buddhism and Projects