Probabilistic Forecasting

What is Probabilistic Forecasting?

IS	Example
Expressing forecasts as ranges with probabilities	“There’s an 85% chance we complete 20–25 items”
Quantifying uncertainty explicitly	Makes risk visible
Using historical data to simulate outcomes	Monte Carlo, throughput-based forecasting
Providing multiple confidence levels	50%, 85%, 95%
Enabling risk-aware decision-making	Tradeoffs based on likelihood
Treating forecasts as distributions, not points	Spread, not single value
Communicating likelihood, not certainty	Language of probability
Updating forecasts as new data arrives	Dynamic, adaptive
Separating commitment from prediction	Forecast ≠ promise
Making variability part of the model	Not noise—signal

IS NOT	Example
Single-point estimation	“We will deliver 23 items”
Deterministic forecasting	Assumes one outcome
Commitment-based planning	Treating estimates as guarantees
Averaging without variability	“We usually do ~20”
Best-case / worst-case guessing	Not grounded in data
False precision	Exact numbers without confidence
Ignoring historical data	Gut feel only
Binary thinking	Done / not done
Static forecasts	Not updated with new info
Hiding uncertainty	Risk is implicit or invisible
Velocity-as-truth	Treating velocity as fixed capacity
Date-driven certainty	Back-solving to meet a date

We can complete [some number] PBIs in the next Sprint with 50% probability, [some number] with an 85% probability, and [some number] with a 95% probability based on historic throughput. ¹

Shared Capacity Illusion
- When capacity forecasts are interpreted without accounting for competing demand and prioritization.
- Capacity is shared, but expectations are interpreted as dedicated.
- Probabilistic forecasting in project-based organizations without prioritization creates false confidence.
- Instead of: “We can complete 21 PBIs” Try: “Given all current priorities, there’s an 85% chance we reach item #18 in the backlog. Project A is #22 thus unlikely this sprint.”