A Scrum-style project estimation tool for people doing the actual project work.
This command-line application accepts as input:
- the total amount of work done in past time intervals (for example, story points per sprint)
- the amount of work left in a project
- optionally, an intended rate of work (velocity)
and produces as output:
- the number of time intervals until completion at the intended rate (if given) or at a reliable rate inferred from past intervals
- the percentage chance of actually achieving that, based on past intervals
The percentage is based on simple simulation: randomly picking past intervals until they add up to project completion. After simulating the project many times, the program calculates the percentage of times the target duration was met. This is based on ideas from Evidence-Based Scheduling, but instead of reporting a distribution it reports the success rate for a specific target.
Everything else -- such as picking a work rate if none is provided or converting a work rate into a target completion -- is based on rule-of-thumb assumptions that can be overridden with optional input. For example, it's usually easiest to track the work of a team across all projects. So when a past interval is randomly selected, only a fraction of the work it represents is credited to the project being estimated. The fraction defaults to a rule of thumb for "the big project we're focused on" but can be overridden.
This approach provides the sort of evidence-based second opinion that most occasional estimators are looking for. The inputs and outputs are tailored for someone who might be asked early on what a reasonable timeline is, but who's more likely to be asked later on, "Can you meet the deadline?" and "What if you worked faster?"
If you have Docker and can run amd64 images, that's the easiest way to run the program:
$ docker run bjaress/estimate --help
If an updated version is released, you can bring it into your Docker cache with:
$ docker pull bjaress/estimate
If you can't do that (or would rather not) you can install The Haskell Tool Stack, then build and install the program from source:
$ stack install
$ estimate --help
$ docker run bjaress/estimate 100 '[65, 30, 50, 70, 40, 90]'
success: 97%
target: 4
Meaning that you should target four sprints for the whole project to have a 97% chance of success.
$ docker run bjaress/estimate --target-velocity=50 100 '[65, 30, 50, 70, 40, 90]'
success: 70%
target: 3
Meaning that if you do fifty points per sprint, you will finish the
project in three sprints, but there's a 70% chance of achieving that.
Both --target-velocity and the list of six past sprints refer to all
the work you accomplish, but the program assumes by default that only
most of that work will be on the project being estimated.
Keeping the ambitious intention to work fast and vowing to spend literally all of your work time on the project
$ docker run bjaress/estimate --target-velocity=50 --focus='1%1' 100 '[65, 30, 50, 70, 40, 90]'
success: 72%
target: 2
Meaning that if you work on nothing else and do fifty points per sprint, you will finish the one hundred points in two sprints, but your past history suggests only a 72% chance of achieving that.
The fraction library used by the program separates the numerator and
denominator with a percent sign, so --focus=1%1 is how you specify the
assumption that all your work will go to the project being estimated.
$ docker run bjaress/estimate --focus='1%2' 100 '[65, 30, 50, 70, 40, 90]'
success: 98%
target: 5
Meaning that if you spend half your effort working on the project, you have a 98% chance of being done if five sprints.
$ docker run bjaress/estimate 100 '[70, 60, 50, 100, 88, 120]'
success: 99%
target: 3
Meaning that with a history of higher past velocities, you can target finishing 100 points in three sprints, with a 99% chance of success.