Drunk Agile Podcast: Monte Carlo Simulation and Probabilistic Thinking

Hosts

• Daniel Vicanti
• Guest: Pratik Singh

Drinks

• Pratik's Drink: Wild Turkey Rare Breed Bourbon (58.4% ABV)
• Daniel's Drink: Glendronach Sherry Cask (60% ABV)

Main Topic: Monte Carlo Simulation and Probabilistic Thinking

Introduction to Probabilistic Thinking

• Probabilistic thinking acknowledges multiple future outcomes.
• We can't predict the future with 100% certainty.
• Humans are generally poor at thinking probabilistically.

Monte Carlo Simulation

• Definition: A technique to model and simulate future outcomes multiple times to determine probabilities.
  • E.g., flipping a coin to understand probabilities without doing the math.
• Process:
  • Model the future and simulate it hundreds, thousands, or more times.
  • Determine how often each possible outcome occurs.

Practical Application in Agile

• Helps in forecasting and planning by understanding the risk of different outcomes.
• Used in various fields, with historical significance in projects like the Manhattan Project.

Why Monte Carlo?

• Monte Carlo is useful when calculating probabilities manually is difficult or impossible.
• Provides a way to simulate future outcomes and assess risk.

Alternative Approaches

• Other statistical methods like averages, standard deviations, and curve fitting.
• Discussion to be continued in future episodes on why some methods are less suitable.

How Monte Carlo Works for Teams

• Data Collection: Use historical throughput data (e.g., daily completed items).
• Simulation Example:
  • Predicting how many items a team can complete in the next 30 days.
  • Run simulations (e.g., 10,000 times) to model future throughput.

Results Interpretation

• Histogram: Shows probabilities of completing a certain number of items.
• Percentiles: Used to segment results and understand risk.
  • E.g.,
    • 95th percentile: 95% chance of completing 27 or more items.
    • 50th percentile: Coin flip probability of completing 42 or more items.
• Shifts planning conversation to risk management.

Challenges in Implementation

• Educating stakeholders to understand and interpret probabilistic data.
• Encouraging decision-making based on informed risk assessment.

Continuous Forecasting

• Concept: Update forecasts as new data arrives.
• Example: Change the selected date range and throughput data to reflect the most current conditions.
• Similar to real-time sports forecasting or weather updates.

Tools and Techniques

• Use of dashboards for real-time updates and continuous forecasting.
  • Example: Ultimate Software's continuous forecasting system.

Key Takeaways

1. Think Probabilistically: Acknowledge multiple possible outcomes.
2. Use Tools: Employ tools like Monte Carlo simulations to understand probabilities.
3. Take Action: Adjust strategies when probabilities are not in favor.

Closing

• The importance of adapting to new data and conditions.
• Each episode builds on the last; future topics include more on continuous forecasting and statistical methods.

Final Words

• Pratik summarizes the episode by emphasizing probabilistic thinking and the importance of using tools to inform decisions.

Cheers and goodnight from the Drunk Agile team.