—
BLOG
FORECAST · GUIDE · 7 MIN READ
Monte Carlo from ADO throughput.
Forecast release dates from actual completion history, no story points required. Output is a probability band, not a single date — which is the only honest answer to "when will it ship?"
—
WHY
Why story points fail at forecasting.
Story points were never designed for cross-team comparison. A team's velocity is meaningful internally, but using it to project release dates assumes future work resembles past work in size — which is rarely true at the granularity that matters for forecasting.
Monte Carlo sidesteps this. It only asks how many items the team has completed historically and how many remain.
—
HOW
The mechanics.
- 01 · Pull last 12 weeks throughputITEMS COMPLETED PER WEEK
- 02 · Count remaining backlogITEMS LEFT IN SCOPE
- 03 · Run 10,000 simulationsSAMPLE WEEKLY THROUGHPUT
- 04 · Output percentilesP50, P85, P95 COMPLETION DATES
Agile Analytics for Azure DevOps does the math. The interesting decision is which 12 weeks to feed it.
—
SETUP
ADO setup notes.
Use Closed Date on Story / Bug, not Activated Date. The former is the throughput signal you want; the latter measures when work started, not when value shipped.
Filter by team, not by area path. Teams own their throughput; area paths can shift mid-quarter and corrupt the history.
Skip the last partial week — including it inflates probability bands wider than they should be.
—
COMM
How to communicate the forecast.
Show p50 and p85, not p95. P95 sounds rigorous but is rarely useful — by the time it matters you should be replanning, not pointing at the band.
Pair with scope. "P85 ship date is March 14 if scope holds at 47 items." Ship date and scope are coupled — quoting one without the other is meaningless.
Re-forecast weekly. Banded forecasts get tighter as the window shrinks. The narrowing itself is information for stakeholders.
—
RELATED