Probabilistic Release Forecasting (Monte Carlo): How It Works
This page explains how Release Management, Roadmaps & Product Portfolio for Jira forecasts release dates using Monte Carlo simulation on your team's real historical throughput. It is a neutral, factual reference for how the capability works and how to read its output.
What probabilistic forecasting is
A probabilistic forecast replaces a single committed date with a range of possible completion dates, each carrying a confidence level. Instead of "we will ship on 15 September," you get "there is an 85% chance we ship on or before 22 September." The forecast is generated from how much work your team has actually completed in the past, not from manual estimates.
The inputs the forecast uses
Throughput history — the number of issues your team has completed per unit of time, read directly from Jira's issue and status history.
Remaining scope — the issues assigned to the selected Fix Version that are not yet done.
Team capacity — working days adjusted for PTO and holidays, so the simulation does not assume a team works every calendar day.
How the simulation runs
The app runs many thousands of simulated "futures." In each run it samples from your historical throughput to estimate how long the remaining scope will take. Aggregating the results produces a distribution of completion dates. The percentile markers are read from that distribution:
p50 — the date you have a 50% chance of meeting (an even bet).
p85 — a commonly used planning date; 85% of simulated outcomes finish on or before it.
p95 — a conservative date suitable for external commitments.
How to read the output
Use the percentile that matches the stakes of the commitment. Internal targets often use p50–p85; dates shared with customers or executives typically use p85–p95. Re-run the forecast as work completes — the range narrows as the remaining scope shrinks and the throughput sample grows.
The forecast is a decision aid, not a performance target. Its purpose is to make uncertainty visible so scope, capacity, and date can be traded off deliberately — not to pressure a team into guaranteeing a percentile.
Where it runs
Forecasting runs natively inside Jira Cloud on Atlassian Forge. No data is exported to an external spreadsheet or third-party service to produce the forecast.