How would you create a probabilistic forecast for 40 stories?

What's really being asked

This question tests your ability to move beyond simplistic, single-point estimates (like velocity) and apply statistical methods to create a more realistic forecast. It assesses your understanding of variability in software development, your skill in using historical data for evidence-based planning, and your ability to communicate uncertainty and risk to non-technical stakeholders like a product manager.

The full answer

A strong answer walks through four key steps. First, you gather the necessary data: the team's historical throughput, which is the number of stories completed per time period (e.g., per week). You need a representative sample, ideally from the last 10-12 weeks. Second, you explain the method: a Monte Carlo simulation. This involves running thousands of computer-simulated forecasts. Third, you describe a single simulation trial: you randomly sample from your weekly throughput data, adding up the stories completed week by week until the total reaches or exceeds 40, and record how many weeks it took. Fourth, you explain how to present the results from thousands of trials: not as one date, but as a probability distribution. For example, 'There is an 85% probability we will finish these 40 stories on or before May 15th' and 'a 50% probability of finishing by April 28th.'

The mistakes people make

The most common mistake is to calculate a simple average. Answering with 'Our average throughput is 5 stories per week, so 40 stories will take 8 weeks' completely misses the point of the question and ignores the reality of delivery variability. Another red flag is getting stuck on data quality, claiming a forecast is impossible without perfect data. A senior candidate works with the data they have, states their assumptions, and provides a forecast with appropriate confidence levels. Finally, avoid confusing throughput (item count) with velocity (story points); the question specifies 'stories', making throughput the correct metric.

What usually comes next

Be ready for 'What if we don't have enough historical data?' A good response is to start collecting it now, use a wider, more conservative range for an initial forecast, and state that the forecast will become more accurate over time. Another likely follow-up is 'How do you handle stories of different sizes?' You would explain that for mixed-size items, you might use story points and velocity, but apply the same Monte Carlo principle to the total points completed per week. Finally, 'The PM is pushing for the 50% date. How do you respond?' Explain that the 50% date is as likely to be missed as it is to be hit, and frame the 85% or 95% date as the one to use for commitments where certainty is important.

A concrete example

Imagine our team's throughput for the last 10 weeks was [3, 5, 4, 6, 2, 5, 4, 3, 5, 7] stories per week. To run one simulation trial for the 40-story backlog, we'd randomly sample from this list. Week 1 might be 4 stories (36 left). Week 2 might be 7 (29 left). We continue this until we've completed at least 40 stories and record the number of weeks it took, say 9 weeks. We then repeat this process 10,000 times. This gives us 10,000 possible outcomes. If we sort these outcomes, the 8,500th value gives us our 85th percentile forecast. If that value is 11 weeks, we can tell the PM, 'There is an 85% chance we will finish within 11 weeks.'