Jon Pretty

Theory: Engineers are never “done” getting better at time estimation. As soon as you get “better”, you’re asked to estimate bigger, more complex, more ambiguous projects, and you have to level up all over again.

I have three measure: 1/ done before tomorrow ; 2/ looks like something I already did in less than a week, so something like that ; 3/ not the least idea, do you want that we play dices together to fill your columns?

for compsci, a professor told us that it was absurd to ask estimates to dev: our work is by essence to automate things. Either the thing is already automated, and you don't need an estimate anymore. Or it's a new thing to automate, and you don't know yet how it can be done :)

This is an idea I've been trying to promote for years, partly as justification for my inability to estimate well, and partly because I believe it's true. Automation makes the familiar trivially and quickly repeatable, and what work remains to be done is completely unknown.

But that's why an estimation is an estimation. You can improve estimations based on similar tasks you've done before, they'll gradually become better over time, even if never perfect. What is clearly wrong is understanding those as deadlines imho.