Software Estimation: Demystifying the Black Art (Best Practices (Microsoft))

17.6 Improving Your Standardized Procedure

When ad hoc estimation procedures are used, it's difficult to improve your estimates because you're never really sure which estimation practices produced the inaccurate estimates. When you use a standardized procedure, you'll know the steps that produced each estimate so that you can repeat the successes and prevent the failures.

After each project, you should assess the effectiveness of your estimates in several ways:

• How accurate were your estimates? Did your ranges include the final result, as discussed in Section 16.6, "A View of a Well-Estimated Project"?

• Were your ranges wide enough? Could they be made narrower and still account for the variability that you've observed?

• Did your estimates tend to be on the low side or the high side, or was the error tendency neutral?

• Were there sources of bias that affected the estimate?

• Which techniques produced the most accurate estimates? Do those techniques generally produce the most accurate estimates, or did they just happen to produce the best estimates in this case?

• Did you reestimate at the right times? Were there too many reestimates, too few, or the right number?

• Was the estimation process more elaborate than it needed to be? How could you streamline it without sacrificing accuracy?