Superceded by documentation at Phlogiston/Forecasting Model.
Assumptions and Methods
Each forecast is fundamentally the size of the backlog (in story points) divided by velocity (in story points per week). We produce an Optimistic, Pessimistic, and Nominal forecast. Each is a combination of three factors:
- Best = current backlog;
- Nominal = current backlog * 1.5
- Worst = current backlog * 2
We will not produce a forecast for Tranche until the backlog of the Tranche appears stable, which is a judgment call of the Product Owner, depending on their knowledge for what belongs in the Tranche backlog, the thoroughness of deconstructions of that work, and the subjective stability of the curve in the Project Burnup .
The FY2015 Q3 and Q4 backlogs both grew over 100% during the course of the quarter. However, this may not apply to the Tranches, or at least Tranches 1 and 2, which based on the historical curve may have already stopped growing.
- Best = Average of highest 3 velocity weeks in last 3 months.
- Worst = Average of lowest 3 velocity weeks in last 3 months
- Nominal = Average velocity for last 3 months
Assumption: We should limit history to the last 3 months for purposes of calculating velocity and maintenance fraction, because this window will reduce the number of anomalies in the data (like holidays) and because the composition of the team may change over time.
The average maintenance fraction over the last 3 months is subtracted from all Velocities.
Assumption: There is no correlation between velocity and maintenance fraction, or at least not a strong enough one to try and deal with