Hacker News Reader

Comment by 0xffff2 | Hacker News Reader

10x is excessive, but I'd also watch closely the relationship between estimates and actual time taken. My estimates are probably 2-3x longer than they where when I first started out because I realized that I was consistently underestimating tasks. Despite being warned about it consistently in college, I still find that virtually everyone, when asked to estimate a task, gives an estimate based on the best case scenario where they encounter absolutely no issues while executing the task.

View on HN

I remember reading a research (don't have a link, sorry) that when one employee gave a one-week estimate and another employee gave a two-weeks estimate for the same task, and each of them delivered the output in two weeks, most managers perceived the first employee as more competent. Which, if true, would explain why most software projects are underestimated -- people instinctively perceive this, and want to be seen as competent.

An interesting experience is to estimate things in your own life. You plan to do something, write down when you expect yourself to finish it. It is quite humiliating to notice the effort you put into convincing yourself that you are more productive than you actually are. Unlike with job, you can't make the excuse that impressing other people is a good strategy. Yet, it can take a lot of work to stop doing that.

The golden rule of estimating: double it and add 20%. The sooner you start doing that, the sooner you start underpromising and overdelivering instead of the other way around.

Starting out I was told double it and move up a unit.

2hrs === 4 days.

A week == 2mths.

Anything past that would never get done.

He was a cynical bastard but not a bad guy and sadly he was in a lot of dysfunctional environments not that far of the mark.

Why add 20%? That's upside precision for something so inaccurate. 2.2x is a good way to set a budget for someone doing a task after it's already been done once by someone else.

The golden rule of estimating is to measure past estimates and use them as a multiplier in future.

The platinum rule is to stop estimating because it's useless.

Double it, then add 20% make a 2.4 modifier.

  20*2*1.2 = 48  
  20*2.2 = 44 // I, also, was confused myself at this point  
  2 *1.2 = 2.4 // not 2.2

Right. Estimating before you even know what you _don't_ know is tea leaf reading. It's better to ask the person who wants it done "how much is this worth to you", or "what's my budget to get this done", and then work from there.

The golden rule of estimation is to give a 1-10 confidence level and if your confidence level is below a threshold (7-8 for example) then also define the risks. Finally, use that analysis to give a reasonable upper and lower bound (or be honest and say "I don't know).

Reply on news.ycombinator.com