Feature #3595

Evidence Based Scheduling

Added by George Montana Harkin over 5 years ago. Updated over 1 year ago.

Status:NewStart date:2009-07-07
Priority:NormalDue date:
Assignee:-% Done:

0%

Category:Issues planningEstimated time:40.00 hours
Target version:-
Resolution:

Description

It would be great if Redmine could provide Evidence Based Scheduling similar to how FogBugz does.

See http://www.joelonsoftware.com/items/2007/10/26.html for a description.

Comments? Worth implementing? Problem areas that might come up?

I might end up implementing this for fun if anyone feels it would be useful.

History

#1 Updated by Ludovic Aelbrecht over 5 years ago

This would be an extremely interesting feature - I'd personally love it if you could implement this.

#2 Updated by Eric Davis over 5 years ago

This would be a good feature for a plugin. I believe the existing Plugin API should provide everything needed for this.

#3 Updated by George Montana Harkin over 5 years ago

Eric Davis wrote:

This would be a good feature for a plugin. I believe the existing Plugin API should provide everything needed for this.

I'll focus my efforts towards a plugin then. Thanks.

#4 Updated by Hans Kazan over 5 years ago

George Montana Harkin wrote:

I'll focus my efforts towards a plugin then. Thanks.

+1
That would be great, the ideas and concepts are something we also favor.

His idea is implemented in the Fogbugz tool, I do not know if the concept is copyrighted?
I can imagine that a look a like is feasible however.

#5 Updated by Jesse Katzman almost 4 years ago

It doesn't look like this is being worked on... correct?

#6 Updated by George Montana Harkin almost 4 years ago

This is not being worked on at the moment. Feel free.

#7 Updated by Jesse Katzman almost 4 years ago

I'm about to make some acronyms to reduce writing, but let's see if this is correct. Break it down!

To calculate Velocities (how accurate developers are in estimating their time):
  1. Give developers without history a fake, wildly inconsistent history.
  2. Keep track of Estimated Time (ET). (ET must be estimated by developer.)
  3. Keep track of Actual Time (AT). (AT must include time allocated to doing task, not time actually spent on task.)
  4. If a task is completed, then AT/ET is a Velocity.
  5. If developer has completed a few tasks, delete fake history.
Now we need to to arrive at an Evidence-Based Schedule. We'll calculate 100 possible futures for each developer.
  1. Multiply each of their upcoming ETs by a randomly selected Velocity.
  2. Add up total to arrive at an estimated time for finishing tasks.
  3. Repeat 100 times.
  4. Lay down these estimated times on a calendar that takes into account work schedule, vacations, holidays, etc.

We now have an Evidence-Based Schedule. If 95 of these estimated times takes place before a certain date, then we can say there is a 95% probability the tasks will be completed before then.

I think the hardest part for projects that don't work a 9-5 would be to lay this down on a calendar. A possible solution is to calculate an Evidence-Based Schedule using estimated/actual completion dates instead of estimated/actual hours spent working when necessary.

#8 Updated by Etienne Massip over 3 years ago

  • Category set to Issues planning

#9 Updated by Patrick Cummins almost 2 years ago

Does the backlogs plugin provide this?

George, I see your git repo here: https://github.com/harking/Redmine-Evidence-Based-Scheduling, but github is offline at the moment. Have you made any progress on this?

Thanks

#10 Updated by Terence Mill almost 2 years ago

It's not offline by github, its just no existing (any more).
I wrote Goerge a mail too, to asks for restore.

As far is i know the part with the past, means calcluating velocity is done by redmine_charts2 plugin well.

Missing is still the future prediction via monte carlo method. THis part is more difficult, because you have to think about howto do auto selection of developers historic velocities(which projects data to take). And which user to take, the one assigned or just (a manual selected) team velocity average. There are many way to do it, which shall be supported and configurable.

#11 Updated by Peter Volkov over 1 year ago

howto do auto selection of developers historic velocities (which projects data to take).

I'm not Monte Karlo method expert but as far as I know you just have to take these developer's historic data randmoly (as random as possible :) ). If redmine was written on python I'd delved in this task but now I have to wait for the hero!

Also available in: Atom PDF