Use pandas for product-level error, rolling forecasts and cumulative stock.
You work on Pine Harbor Distribution's fictional planning file. Before calculating, check the time period, the unit and which business event each input records. Keep those definitions in the workbook so another analyst can reproduce the result. The first task gives you the steps. Later tasks use new quantities, a boundary case and a handoff to another planner. Your answer is recalculated on a second copy of the inputs, so enter a formula rather than a number copied from the example.
Return a Series indexed by sku: total absolute baseline error divided by total actual demand per SKU. Use all periods. The notebook starts with pandas and the distributor's demand extract loaded as demand. Each row is a SKU-period observation. Keep the time order explicit, preserve the requested output shape, and calculate from columns instead of typing a result. Demand can be zero; an initial rolling window can be missing. Those states have different meanings and must not be filled without a reason.