Background

Noting service level agreements (SLAs) with key customers, the client requested a review of a segment of inventory to characterise the costs of customer forecast inaccuracy. The extra inventory investment required to maintain a service level goal was estimated for different levels of forecast error, and the outcomes used to inform SLA negotiations.

Challenge

• Client Back Orders not available – the forecast error derived from historical Sales may be understated if Demand > on hand stock (inc SS)
• Average Inventory = ½ OQ + SS, used to estimate the necessary stock on hand to satisfy the forecast demand
• Actual Inventory = Average of 4 weekly stock on hand snapshots in, made up of SS and Cycle Stock
• Client’s forecasting system rarely re-run to select appropriate statistical models

Approach

A project was commissioned to quantify the value of improved forecasting based on a segment of high value inventory:

• Historical data collection of monthly forecasts and actual achieved sales by product - demand assumed to be at the factory gate. Lead times, Service Level targets and On hand by item
• Calculation of Forecast Error; 12 Values of Forecast Deviation (Actual Sales – Forecast) determined. Average deviation in $ used to characterise forecast error performance by item = $ Standard Deviation (STDDEV)
• Use of an Advanced Planning System (APS) to calculate Optimal Reorder Quantity (ORQ) and Safety Stock (SS). The APS raw forecast (choice of 34 models) based on 24 month history to fix ORQ and client adjusted forecast error (overwrite in APS database)
• Varied forecast error in APS simulations (50% to 200% of current value) to characterise the relationship between Average Inventory and STDDEV

Outcome

• Generally, there was a correlation between increasing forecast error and increased Safety Stock at the Group level
• However, at the Item Level
  - The Safety Stock Policy (days cover) was only loosely connected to Service Level
  - The Safety Stock Policy was not connected to Lead-time or Forecast Error
• Generally carried stock in excess of SS requirements ($1.3M, 14% of Current Average Stock on hand), but not achieving SL target for some items. In some cases service levels could be achieved without Safety Stock
• Average Stock On Hand was well in excess of calculated Average Cycle Stock and Safety Stock ($3.5M or 37% of Average Stock on Hand). No policy was in place to decide when to scrap excessive stock of “non-stock” items
• Better forecasts would reduce the amount of Safety Stock necessary to achieve the service level promise