Harnessing the Power of Parallel Universes to Create Realistic Timelines

Talk #2: Use the Power of the universe to creat realistic Timelines - challenge accepted
Lorenzo Santoro & Benjamin Huser-Berta, Hitachi Energy

Are you stuck in lengthy meetings, converting story points to hours, relying on metrics like velocity, and still unable to reliably tell stakeholders when things will be done? If you need to create timelines or have contractual obligations to deliver features by a certain date, you need a reliable way to forecast completion. After trying various approaches, we’ve found a method that provides more accurate results with less effort than estimating in hours, ideal days, or story points. We introduce probabilistic forecasting with Monte Carlo simulations. We will share our journey of moving away from traditional planning towards a system based on managing flow and applying continuous forecasting. We’ll also discuss the impact this shift has had on developers, Product Owners, and stakeholders.

Lorenzo Santoro (Product Owner) und Benjamin Huser-Berta (Scrum Master):
We are Lorenzo Santoro (Product Owner) and Benjamin Huser-Berta (Scrum Master), working at Hitachi Energy. In a traditional organization that operates in a conservative market, we strive to apply modern practices while acknowledging the needs of the organization and its customers. With our teams, we focus on continuous improvement based on data, without being tied to any specific framework, method, or practice—whatever helps us improve is what we should be doing.

Transcript

1. Harnessing the Power of Parallel Universes to Create Realistic Timelines Lorenzo Santoro & Benjamin Huser-Berta
2. About us Lorenzo Santoro (Product Owner) Benjamin Huser-Berta (Scrum Master) About Us
4. The Context Products may end up being loaded in containers and literally be shipped Certain Systems will be running off-shore, with minimal access after it’s being delivered Software & Hardware Containerize & Ship Off-Shore Assembly of Systems “in-house” where they are inspected and “accepted” by the customers
5. What We Tried Traditional Project Management Ideal Days Story Points
6. The Problems You still need to translate “points” into “time” “Capacity Planning” needed There was no correlation between Story Point Estimate and Time Cycle Time: The amoun t of elapsed time between when a work item started and when a work item finished. - The Kanban Guide
7. Monte Carlo Simulations (MCS) Probabilistic Forecast Based on Throughput Runs in Seconds Answers relevant Questions Risk Management No “Guessing” Needed Continuous Forecasting “How Much” and “When”?
8. A forecast is a calculation about the future completion of an item or items that includes both a date range and a probability. Examples: There is a 70% chance that it rains in the next 5 days. There is a 50% chance that it will be done in 21 days or less. Probabilistic Forecast
9. Throughput You already have a Throughput No need to do anything special to “create” it We can base our MCS on the “last X days” Changes in the environment will be automatically included The number of work items finished per unit of time. Note the measurement of throughput is the exact count of work items. - The Kanban Guide Facts
11. How does MCS Work? Example: Forecast how many items we can close in the next 7 days Assume we have 10'000 parallel universes. In each of those, we simulate 14 days of work. 01 To simulate a single day, we take a random day of Throughput from our input. We do this 14 times and sum up the total throughput. 02 Do this for all of your 10'000 universes, and you end up with 10'000 simulations of 14 days. 03 From those 10'000 results, you can read the probability of getting x or more items done 04
12. Answering the real Questions There is a 80% chance that we close 18 or more items in the next 2 weeks When? Release Planning Specific Feature Completion How Many? There is a 70% chance that we finish 25 items until 30th of January or earlier Sprint Planning PI/Quarterly Plannings
13. Continuous Forecasting The real strength of MCS lies in the fact that you can (and should) be doing it continuously Pick a random number 1. Look up a value 2. Sum it up 3. Computers can run this in seconds Simple Math Your Throughput is different every day. You “question” is adjusting every day. Rerunning it quickly and often enables agility. Every Day is Different Inspection without adaptation is considered pointless. - The Scrum Guide Use the data as soon as you learn something new. Take Action
14. Does this Really Work? Monte Carlo Simulations, Accuracy, and Unplanned Work — A Case Study Predictability is something you do! The Flaw of Averages — Comparing Monte Carlo Simulations with Estimates based on Averages MCS will always ‘work’ - but you can take specific action to make it “more accurate” Continuous Flow/Small Batches
15. How does this impact... Developers? No need to context-switch to other activities that are not value- adding. Less Distractions No “Arbitrary Timelines” About what a Story Point now really means... Less Confusion
16. How does this impact... a Product Owner? Even if it would be equally (in)- accurate as other options Less time wasted... Easier to Plan Easier to Communicate Sprints/Features/Releases More accurate and faster Use a shared language (with real dates) to talk to your Stakeholders Actionable Act based on what you learn from the latest forecast
17. Wrap-Up MCS helps you answer the questions "How Many" and "When" Which are relevant questions for our Stakeholders It is answering those questions with a probability Which is great for Risk Management It is based on Throughput Which means you don't need anything extra to create the forecasts It can be run within seconds Which means you can continuously update your plans as you learn new things
18. THANK YOU FOR LISTENING TO US!
19. FAQ
20. RESOURCES https://blog.letpeople.work FOUNDATION OUR BLOG POSTS
21. TOOLS https://www.letpeople.work#lighthouse https://www.letpeople.work/#flowpulse https://letpeople.work/#tools