Commentary: Using order data to optimize inventory

MacKay & Company

By Richard Ilseman, MacKay & Company

Inventory optimization is a tricky business. Most inventory is made up of two major segments, expected demand over a replenishment lead time stock and safety stock to handle any demand variability.

So if the lead time from your supplier on a part is typically five days and your customers on average order 10 pieces per day, you would need 50 pieces in inventory to handle that average demand. However, if demand is really 10 pieces per day plus or minus four pieces, you would need to add enough safety stock to handle those times when customer demand exceeds 10 pieces per day.

Optimizing inventory is largely about optimizing safety stock. Variability of demand and the fill rate you wish to provide to your customers drives safety stock. Optimizing safety stock begins with making the most accurate forecast you can of demand and demand variability.

Using order data to forecast demand: Using order data from your customers is a good starting point to develop an accurate demand and variance forecast. Here are a few helpful hints to get you started:

  • Smooth the data: When customers order they are normally using some type of inventory replenishment model themselves. This means what they order from you not only includes true end-customer demand, but may also include safety stock for their system, extra units to take advantage of any volume discounts or perhaps a sales program they may be offering that can temporarily spike demand. Smoothing order data helps reduce forecast error.
  • Determine seasonality: Keeping at least 24 to 36 months of customer order data will allow you to see any seasonality in the parts you sell. Identifying seasonality reduces forecast error.
  • Keep track of part supersessions: When keeping several years’ worth of data in your forecasting model, it is very important to track supersessions so you can more accurately forecast trends in like families of parts.

Beyond order data: Using true point-of-sale (POS) data from your customers is a big step up from raw order data. POS data represents true end-customer demand and strips out most of the noise that you see in order data.

In addition, you can see how many pieces a customer purchases per transaction. Using this information should provide more accurate estimates of variability that will allow you to better forecast your optimal safety stock.

Big Data – The Next Frontier: Big-data models are generally more appropriate for large suppliers/OEMs that need to adapt quickly to new model/parts introductions and make more accurate all-time buys on older components. While order or POS data works well for parts with stable demand, they are not the best for parts at the beginning or end of their lifecycle. The increasing speed with which OEMs are introducing new models and/or new technologies make this a common occurrence.

With the increasing availability of sensor data, warranty data, engineering mean-time-to-failure data and VMRS data, you can begin to utilize the software tools now available to analyze this “big data.” For example, demand for a particular truck component can be forecasted based upon expected failure rates and the population of trucks having that component now and in the future.

For most of you, using order or POS data as the basis for forecasting is all you need to effectively optimize your inventories and better serve your customers.

Richard Ilseman joined MacKay & Company in September 2014 after 40 years with Navistar International. While at Navistar, he piloted, designed, implemented and supported a vendor-managed inventory system. As an accomplished SAS programmer, Ilseman conducts statistical analysis and survey tabulation at MacKay & Company.

Learn how to move your used trucks faster
With unsold used inventory depreciating at a rate of more than 2% monthly, efficient inventory turnover is a must for dealers. Download this eBook to access proven strategies for selling used trucks faster.
Download
Used Truck Guide Cover