How Manifst forecasts when a work package will finish

Forecasting by simulation from the team's real pace: what "likely finish" and "safe finish" mean, and when the forecast is reliable.

Plan

Portfolio and above

#The question

"When will this work package be finished?" Rather than a single date, Manifst answers with a probability, based on the team's real pace. Forecast blocks are in the dashboard, Forecast category: end of the backlog, and end of each epic or Super Epic. On a plan that does not include them, they are shown blurred, with a fictitious example.

#The method

1. Manifst records how many tickets the team completed each week over the recent period (up to 90 days). With fewer than 8 weeks of history, it adds the latest completed sprints.
2. It replays the future 1,000 times: for each simulated week, it randomly draws one of the weeks actually observed, until the remaining tickets run out.
3. The 1,000 resulting dates give two markers:
   • Likely finish: one chance in two of finishing before it. It is a coin toss: do not announce it.
   • Safe finish: 85 chances in 100 of finishing before it. This is the date to announce.

Scope is counted in number of tickets, not points: many teams do not estimate all their tickets, and the number of tickets completed per week turns out to be a good predictor.

#Several work packages in parallel

The team does not work on one epic at a time. Each work package moves at the share of activity it received over the period: if 30% of completed tickets belonged to an epic, the simulation gives it 30% of the pace. A package that received no activity has no pace: Manifst only shows the date it would reach if the whole team worked on it.

When a work package has an end date, the forecast classifies it:

• On track: the safe finish is before the deadline;
• At risk: the likely finish holds, but not the safe finish;
• Late: even the likely finish passes the deadline, or the deadline has passed.

#The assumptions

The forecast assumes that the past pace continues, that each package's share stays the same, and that no ticket is added. If you add scope, the date moves back: the forecast updates itself.