The requested items are retrieved and delivered to the single entrance of the warehouse simultaneously by a group of human order pickers. For the sake of clear responsibilities, disjoint areas of the warehouse are assigned to individual pickers. At the entrance of the warehouse, the picked items are loaded onto little trains, which circulate according to a pre-defined periodic schedule in a milkrun system. The schedule contains four train departures—so-called milkrun cycles—in each hourly period. Each cycle serves the material requirements of different departments of the factory. About the half of the items are consumed in several milkrun cycles. It is possible to estimate the probability pik that a pick list in cycle k contains item i based on historic data, with occasional adjustments based on forecasts when the product mix changes. However, more precise data, e.g. the covariance of the random variables cannot be reliably extracted from this data. Stock level forecasts are, however, available.