- Installation
- Prioritize projects
- Estimate project duration without historical data
- Estimate project duration based on historical data
A pretty dumb PM.
The whole philosophy behind dumbpm is that PMs (project managers, program managers, product managers, people managers, pokemon masters, etc.) all do some tasks that could use some automation, but at the same time this automation should be as dumb as possible. There are so many changing factors that influence such tasks that you better hire a PM (or a team of researchers to do the automation) to do the clever stuff, not a random software on the internet. In addition, if we keep it dumb, people can just read the code and understand what is going on, if they really want to.
If you have any suggestions for something (but nothing clever!) that you would like dumbpm to do for you, open an issue and let me know.
pip install dumbpm
Tested on both Linux and Mac OS. Windows might work.
NOTE Python 3.9+ required.
Giving a table of projects defined as below, it outputs a list of projects in order of priority within the optionally specified budget (prioritization as "data problem").
The prioritized list is modelled as the exact solution of a Knapsack Problem with the following value function: norm(value) / (norm(cost) + norm(duration) + norm(risk)). Pretty dumb, indeed.
If --budget isn't specified, the list will simply be sorted by the result of the above value function. Budget is relative to the cost parameter.
If you are expressing cost as cost per unit of duration (e.g. developers per week or sprint), you'll have to specify the --cost-per-duration option, so that total cost = cost * duration and budget is measured against total cost (duration will also disappear from the value function not to count it twice).
By setting the pick parameter you can also "cheat" and tell the solver to always include a project in the solution. This can be useful if you want to see how the solution would change based on that choice.
$ dumbpm prioritize --help
usage: dumbpm prioritize [-h] [--budget [BUDGET]] filename
positional arguments:
filename CSV file with projects definition
optional arguments:
-h, --help show this help message and exit
--budget [BUDGET] Max budget allowed
--cost-per-duration Cost is to be assumed per unit of duration. Budget =
(cost * duration)Project definition happens in a CSV file with the following structure:
Project: [required] name of the projectValue: [required] value of the projectCost: [required] cost of the projectDuration: [optional] duration of the project expressed in unit of timesRisk: [optional] risk of failure of the projectPick: [optional, default = FALSE] force the inclusion of the projectAlternatives: [optional, empty field = empty list] comma separated list of projects that are incompatible with this one (e.g. make lunch vs buy lunch)
There is a bit of slack on the headers of the columns (e.g. Project, Projects, project, etc. are all alright). Notable mentions: alts and alternatives work the same; PQ can be used instead of cost if that's your thing.
| Project | Value | Cost | Duration | Risk | Pick | Alternatives |
|---|---|---|---|---|---|---|
| Buy a better espresso machine | 5 | 4 | 2 | 1 | TRUE | |
| Buy ambient parfume for the back | 1 | 1 | 1 | 1 | ||
| Find and remove source of bad smell | 5 | 2 | 4 | 4 | TRUE | |
| Find better coffee vendors | 4 | 2 | 6 | 3 | ||
| Buy smart component for roaster | 3 | 5 | 2 | 3 | Build in-house roasting notification system | |
| Introduce a periodical suggestion survey for treats | 3 | 2 | 6 | 1 | ||
| Buy a more modern sign | 3 | 3 | 1 | 1 | ||
| Contact a designer to re-think the front | 5 | 5 | 6 | 3 | ||
| Buy beds for powernaps | 1 | 3 | 1 | 1 | ||
| Import treats from France | 2 | 4 | 2 | 2 | ||
| Build in-house roasting notification system | 3 | 5 | 6 | 5 | Buy smart component for roaster |
$ cat projects.csv
Project,Value,Cost,Duration,Pick,Alternatives
Buy a better espresso machine,5,4,2,TRUE,
Buy ambient parfume for the back,1,1,1,,
Find and remove source of bad smell,5,2,4,TRUE,
Find better coffee vendors,4,2,6,,
Buy smart component for roaster,3,5,2,,Build in-house roasting notification system
Introduce a periodical suggestion survey for treats,3,2,6,,
Buy a more modern sign,3,3,1,,
Contact a designer to re-think the front,5,5,6,,
Buy beds for powernaps,1,3,1,,
Import treats from France,2,4,2,,
Build in-house roasting notification system,3,5,6,,Buy smart component for roaster
$ dumbpm prioritize projects.csv
01 Buy ambient parfume for the back
02 Buy a more modern sign
03 Find and remove source of bad smell
04 Buy a better espresso machine
05 Find better coffee vendors
06 Buy beds for powernaps
07 Buy smart component for roaster
08 Import treats from France
09 Introduce a periodical suggestion survey for treats
10 Contact a designer to re-think the front
$ dumbpm prioritize projects.csv --budget 10
01 Buy ambient parfume for the back
02 Buy a more modern sign
03 Find and remove source of bad smell
04 Buy a better espresso machine
Giving a list of tasks or milestones the project can be broken into and (best case, expected, worst case) estimates for the duration of such tasks, it outputs an estimate of the project duration in the form of a probability distribution (median, variance, percentiles). You can use these numbers to formulate guesstimates like "I am 75% confident that we will complete the project in 38 weeks". Instead of using a unit of time (e.g. days, weeks), you can instead specify story points for each tasks and so estimate the project size in story point instead of its duration.
The estimation is based on a Monte Carlo simulation of the following equation: project_duration = sum(duration, over=tasks). Tasks duration for each iteration are derived from a modified-PERT distribution interpolating the (best case, expected, worst case) estimates for the task. Pretty dumb, indeed.
The command is called guesstimate to highlight that the result of this method should be considered with lower confidence than the one produced by the estimate command, which is based on historical data. You can switch to estimate after you conclude a few sprints, while relying on guesstimate for the initial estimates during the planning period.
If it is taking too long to perform the estimation on your computer, set --simulations to something lower than 10000.
$ dumbpm guesstimate --help
usage: dumbpm guesstimate [-h] [--simulations [SIMULATIONS]] filename
positional arguments:
filename CSV file with tasks estimates
optional arguments:
-h, --help show this help message and exit
--simulations [SIMULATIONS]
Number of simulations to runEstimates for tasks or milestones have to be defined in a CSV file with the following structure:
Task: [required] name of the task (Milestoneworks as header, too)Best: [required] best case estimate for the task durationExpected: [required] most likely estimate for the task durationWorst: [required] worst case estimate for the task duration
| Task | Best | Expected | Worst |
|---|---|---|---|
| Task A | 5 | 10 | 20 |
| Task B | 6 | 12 | 40 |
| Task C | 1 | 13 | 24 |
| Task D | 10 | 13 | 15 |
| Task E | 5 | 7 | 12 |
| Task F | 12 | 25 | 34 |
$ cat tasks.csv
Task,Best,Expected,Worst
Task A,5,10,20
Task B,6,12,40
Task C,1,13,24
Task D,10,13,15
Task E,5,7,12
Task F,12,25,34
$ dumbpm guesstimate tasks.csv
Duration
count 10000.000000
mean 7.761430
std 0.993793
min 5.000000
50% 8.000000
5% 8.000000
90% 9.000000
99% 10.000000
max 12.000000
Giving a list of past sprint velocities and (optionally) a list of scope changes for the project in story points defined as below, it outputs an estimate of the project duration in the form of a probability distribution (median, variance, percentiles). You can use these numbers to formulate guesstimates like "I am 75% confident that we will complete the project in 38 weeks".
Please note that the estimate is measured in sprints, so you'll have to multiply that for the duration of your sprint and project that on your working calendar to account for holidays and anything else which could affect your schedule.
The estimation is based on a Monte Carlo simulation of the following inequation: scope + sum(scope_change, over=sprints) <= sum(velocity, over=sprints). Pretty dumb, indeed.
By default, velocity and scope change for each iteration are picked at random following a uniform probability distribution from the provided historical data. If --normal is specified, the input will be modelled as normal distribution from which velocity and scope changes will be derived.
If it is taking too long to perform the estimation on your computer, set --simulations to something lower than 10000.
$ dumbpm estimate --help
usage: dumbpm estimate [-h] [--normal] [--simulations [SIMULATIONS]] filename scope
positional arguments:
filename CSV file with velocity and scope change datapoints
scope Remaining scope in story points for the project
optional arguments:
-h, --help show this help message and exit
--normal Use a normal distribution for the input data
--simulations [SIMULATIONS]
Number of simulations to runHistorical data has to be defined in a CSV file with the following structure:
Velocity: [required] velocity for each past sprintChange: [optional] project scope change for each past sprint (zero and negative values are allowed)
| Velocity | Change |
|---|---|
| 17 | 5 |
| 19 | 1 |
| 10 | 0 |
| 12 | 0 |
| 21 | 1 |
| 7 | -3 |
| 15 | -2 |
| 12 | 5 |
| 12 | 0 |
| 14 | 2 |
| 18 | -4 |
$ cat sprints.csv
Velocity,Change
17,5
19,1
10,0
12,0
21,1
7,-3
15,-2
12,5
12,0
14,2
18,-4
$ dumbpm estimate sprints.csv 100
Duration
count 10000.000000
mean 7.761430
std 0.993793
min 5.000000
50% 8.000000
75% 8.000000
90% 9.000000
99% 10.000000
max 12.000000
$ dumbpm prioritize projects.csv 100 --normal
Duration
count 10000.00000
mean 7.75637
std 0.99810
min 4.00000
50% 8.00000
75% 8.00000
90% 9.00000
99% 10.00000
max 13.00000