Putting the Demand Forecasting Fundaments to Work
A demand forecast always contains errors, but the errors diminish with an increasing aggregation level (in time, product hierarchy, or location hierarchy), and when forecasting less far into the future. It’s a good idea to look for ways to use the aggregation and time horizon dynamics of forecasting to increase total forecast accuracy and inventory management efficiency from that perspective. A good example is work we have done with food retailers in managing seasons and promotions.
In high seasons, and during major retailers’ big promotions, many suppliers require advance information to be able to ensure production capacity and delivery capability so that the products needed can be delivered.
We were working with a customer in a market where the major food producers required pre-orders of products before the start of major seasons to ensure availability. Thus the retail stores had to calculate orders for each delivery day within that season for 6-8 weeks in advance. However the supplier naturally does not need to know the final allocation of the goods that early, just the total volume.
We Implemented a Process Where We Used Our SCM Software:
- A day level demand forecast was calculated for each product and each store well (months) before the season
- The forecast was nevertheless re-calculated daily with the most recent information available to increase accuracy
- The demand forecast and customer’s presentation plans were used to build a day-level delivery plan for each product and each store
- The delivery plan was recalculated with the updated demand forecast daily
- Just before the time items were needed the day-level delivery plan was consolidated into aggregated store chain level data and shared with the suppliers
- Ordering was performed at a normal lead-time
- The aggregate forecast was so well in line with the actual orders, that the suppliers could easily cope with the variation, and the retailer had no need to allocate scarce or extra stock
The accuracy of the forecasts was increased by working at a higher aggregation level when dealing with a longer time horizon and moving to store level only when dealing with the shortest possible time horizon. The change was a big success. For example where fresh products were concerned the retailer was able to increase availability while simultaneously reducing spoilage, both by several percentage points – so a worthwhile development!