Have you ever had painfull projects not delivering or not according to estimates?
You can avoid that.
PM and EO’s often face painfull experiences when development takes n times more than estimated.
Different estimation techniques do rarely deliver reliable predictions for software projects.
However, using experience can improve predictability and provide answers to difficult questions such as:
“When are we done?” or “How many features can we deliver in this PI?“
If you collect historical story points (estimates) and actual calendar time spent
then you can benefit from the estimation edge of forecast propabilies.
Historical program throughput is a powerful data point.
Your historical data includes all matters often not included in estimates, e.g. waiting time, too high WIP, bottle necks, etc. until done.
And using this actual knowledge (not estimates you want to believe are true) in a probabilistic forecast using Monte Carlo simulations,
will provide a more realistic and predictable insight in how long time it will take to deliver, and
prevent you from the mistake of promising milistones/deliveries to customers that are not realistic.
Call to action:
Grab a list of your historical work, open the attached Excel and be prepared for a more realistic long term plan.
Open the attached Excel.
Put your historic done features/stories/defects (a unit of work) including estimate and actuals into sheet 1 (History).
Check for possible correlation between estimate and work done i nsheet 1 (History) chart. Default data in Excel has no correlation – i.e. estimates have no predictability and can be ignored.
Adjust to your number of history rows in sheet 2 (Simulation).
Adjust number of issues to be done to your case (remove or add rows to the green area and formula’s) in sheet 2 (Simulation).
See the Monte Carlo propabilities in sheet 2 (Simulation).
If your teams can handle more than 3 issues simultaneous, then change the Start Date and End Date section formula’s.
Be aware of the date summaries of simulations to the far right in sheet 2 (Simulation). Date columns will need adjustment so it reflects your future outlook in years or months.
Watch sheet 3 (Results) and observe the End Date probabilities (a range of dates from the most optimistic outcome to the least fortunate), all based on your own history and how caucious you need to be.
This is a simple Excel that reveils the power of the probabilistic approach (although you might not like the results). In real life, you want to adjust the Excel to your situation.
In reality you must also understand your queues and limited resources. This text is not about that.
Carl Starendal, for providing me the idea, at the european 2019 SAFe Summit. It has been a life saver for me since.