Why we ask for four weeks of roster history before forecasting

Forecasting windows need enough signal to separate noise from structure. Four weeks is our pragmatic minimum for mixed desk and field teams; shorter spans trigger an explicit limitation memo instead of false precision.

When history is thin, we lean on scenario branches rather than point estimates. Each branch carries assumptions about arrival variance, training holds, and vendor delays. That sounds heavy, but the alternative is a single line chart that leadership treats as promise.

We publish the assumption log beside the chart so planners can challenge inputs without dismantling the entire model in a single meeting.