Estimate Quality Metrics

While doing some Google search on metrics, I have stumbled upon a metrics called Estimate Quality Factor, also known as EQF. Now, as a first guess, I thought that this might have something to do with the confidence on the estimate. I was partly right. Though I was thinking in terms of the effort and how confident was one on the effort, the EQF actually is on the delivery date, i.e. the schedule. It is a really simple idea. One needs to plot the estimated end date for any phase of delivery against the calendar date. For instance, if you estimate that the Tech Spec phase of a project will be completed by say 20^th May, you plot 20^th May in the y-axis with today’s date on x-axis. Similarly, whenever the estimated delivery date changes due to any factor, the next estimated end date is plotted. If there are no changes also, one can just plot the estimated end date with respect to the calendar date on a weekly basis. For example, say next week, we estimate that the Tech Specs will be completed by say 23^rd May. So, we plot that. The idea is to see if all the dates plotted form a kind of straight line parallel to x-axis, which will signify that the deviation from planned delivery date is not much. The more there are ups and downs in the delivery date, means that the EQF was not good. I think that the EQF can also be plotted for the effort estimate also. And an effort estimate with a high confidence will have a line parallel to the x-axis.

If estimates keep changing too often, which is the case mostly, the line will never be parallel to the x-axis. Again, only if there is a process of estimation built into the development process and the estimates are recorded somewhere, pulling out this data on a weekly basis should not be much of a challenge.

Now coming to the part where I need to decide if this metric will be of use to any one? I am sure that we will not be able to get a parallel line to x-axis in real life projects since almost all the projects will have change requests coming up in the course of the project, which will change the estimate effort and may be the end dates also. Only in a scenario where there are no change requests, we can probably use this metric to prove the quality of estimates.

Instead of doing all these stuff, a simple metric like the effort variance should be able to tell me how good the estimate was.