The Project Estimate

In his research, Karner applies 20 person-hours for each UCP. Doing the same, we would end up with a 1597.2 person-hours (20 × 79.86) for the Remulak project. For a 32-hour-per-week schedule with one person doing the work, this would amount to approximately 50 person-weeks.

Because 5 people are working on this project, we should allow about 10 weeks to finish the job. However, after considering time for unproductive meetings, communication issues, and show-and-tell sessions (you know, the ones that are never scheduled but you find yourself doing a lot of), we will add 4 additional weeks to the schedule. The total for the project, then, is 14 weeks.

Although Karner's approach isn't cast in stone, nor does it profess to provide the magic number, it is an excellent approximation. Others have suggested areas for improvement. In their excellent book Applying Use Cases: A Practical Guide, Geri Schneider and Jason Winters suggest that care should be given to the environmental complexity factor (ECF). She suggests counting the number of factors 1 through 6 that have a rating (not an extended weight) of less than 3 and the number of factors 7 and 8 that have a rating of greater than 3. If the total is 2 or less, use 20 person-hours per UCP. If the total is 3 or 4, use 28 person-hours per UCP. She goes on to indicate that if the total is 5 or more, some attention is needed because the risk for failure is very high. Using her criteria, we are still in good shape with 20 person-hours per UCP.

Use Karner's heuristics; they are the best I have seen up to now for estimating projects using the artifacts from UML (primarily the use-case). Tweak it where necessary, but remember that this approach is designed to prevent the sometimes dangerous rule-of-thumb estimates that we are so quick to provide but quite often overrun and regret.