476

# Think twice before making a time estimate. Then multiply that by two.

• 7
This can't be said enough...
• 3
• 7
I made a small system for estimating. Give 3 estimates. 1 low-ball (if things are easier than expected). My actual estimate based on intuition. A high estimate (if nothing went right). Then my base estimate is the average of these. From the average I add 1 standard deviation (from the 3 estimates) of time for every degree of significant uncertainty.

This approach has worked out ok for me so far
• 21
I hate when you give an estimate, a release date gets announced to the clients, then you are forced to release unfinished covfefe.
• 3
Then hunk about multiplying it by two twice, and then multiply it by two.
• 10
true...

- When do you think it will be ready?
- In a few days, max 1 week.
...after 3 weeks...
- It's almost ready... I will send you it next week.
• 1
So freaking true!
• 1
And then add 10 or 15%
• 5
I spat out my covfefe holy shit that was funny
• 2
Power that to the eleventh, and add a million and thrithy one seconds
• 1
By 3 is more accurate
• 2
Don't estimate in time. Estimate in story points or t shirt sizes if you are very uncertain. People are.much better at relative estimations than absolute.
• 1
@CodeMasterAlex the idea about t-shirt sizes is great! I haven't heard it before!
• 2
"We won't define a time estimate. We'll leave the time estimate open, but when we reach it, we'll double the time estimate" Dilma Rousseff
• 1
Always!.
• 0
@codetinkery If you estimate time like that would't your actual estimation almost always be the result of average of those?

Ex:
(low estimate + mid estimate + high estimate) / 3 = mid estimate(roughly)

Hrs:
(1+2+3)/3=2
• 2
@Kryptic The base estimate is always the average, yes. However, I calculate the standard deviation based on the 3 estimates and add that to the average once, for every degree of significant uncertainty.

http://imgur.com/ze5uRiE

"low" uncertainty means "use the average value"

mid is 1 deviation.
high is 2 deviations.

and here is the code used to calculate the estimates: https://pastebin.com/0F4y2gRK

Do note that "standard deviations" are more precise the larger your data-set.

I've found however, that my "low estimates" are always too short, so the "average" number is more often than not closer to the "real low" while the high estimate lands somewhere near the uncertainty groups.
• 0
@codetinkery Good explanation!
• 1
Multiplying by π sounds a lot more scientific and is more secure in the long run :)
• 0
@seeker it definitely sounds quite appetising if we name this the "pi coefficient"

welcome to devRant!
• 0