What are the possible techniques to arrive at a good estimate? We use Delphi estimation technique for estimating tasks. What are the other better ways to do so?
Also what would be the do's and dont's while giving an estimate. Often we miss schedules and target dates because of poor estimation.
The product we work on is a legacy system that has been in customer sites for past 10 years. So its not easy to give estimate just like that.
Hence would like to know some techniques that can benefit us.
On our team at Microsoft (Windows) We've used the Wide Band Delphi method with very good success. I find it very collaborative. Most importantly, its more accurate than anything else I've ever tried.
Collect and analyze metrics. Look at how close your previous estimates have been. Do some statistical analysis. Do you see a smaller standard deviation when you look at estimates for what turned out to be smaller pieces of work vs. larger pieces of work?
At the very least, keep track of this personally. Make an estimate, if only for your own use and practice, before doing any amount of work. See how close you are. If you estimate 4 hours for a task and it takes you 12, and then you do the same thing the next day, you'll start to get the idea that you need to scale your estimates by 3x.
But maybe when you make estimates for larger pieces of work, you're off by a factor of 5x instead. For example, you estimated 4 months and it took 20.
Instead of trying to improve your estimation ability, keep making the same kinds of estimates, and apply these fudge factors.
Also keep in mind that if you give a more realistic estimate, you may still blow the estimate. Sometimes work expands to fill (or overflow) the time allotted for it. When you have more time, you feel like you can afford to make something just a little better. That can be a great way to blow even a realistic estimate.
The Briggs-Chase Law of Program Development: To determine how long it will take to write and debug a program, take your best estimate, multiply that by two, add one, and convert to the next higher units.