The example I usually use is that is January you want to organise a music festival in September; and you ask about the weather the 9th of September. Is impossible to know. You have historical data but the right weather is only accurate one month before. You need to manage that. That’s how estimation works
The key point is that estimation should be constantly reviewed and the given number or date should be include a probability (this is as PERT works). At the beginning you give the number with a low probability as you know little and have a lot of open questions and a lot of things can happen… as you go the probability is higher.
The thing is that the decision about what to do with that number should be taken by business using the probability and assuming the risk… a project with a number below a 95% of probability is suicide.
The example I usually use is that is January you want to organise a music festival in September; and you ask about the weather the 9th of September. Is impossible to know. You have historical data but the right weather is only accurate one month before. You need to manage that. That’s how estimation works
The key point is that estimation should be constantly reviewed and the given number or date should be include a probability (this is as PERT works). At the beginning you give the number with a low probability as you know little and have a lot of open questions and a lot of things can happen… as you go the probability is higher.
The thing is that the decision about what to do with that number should be taken by business using the probability and assuming the risk… a project with a number below a 95% of probability is suicide.
Nice approach to a very common issue in the Tech Industry, by far the best post in this category that I'd read in years!