Inventory is one of the working capital components that can be influenced by poor demand information and can be a big drain in the cash-to-cash cycle. Optimal inventory is defined as the inventory that is just sufficient to meet your customer orders. 

Since this is a balancing act, demand forecast accuracy determines both your inventory levels and service levels. You can make a trade-off between the two for a given demand forecast error 

  • Higher the required service levels, higher is the needed inventory. 

  • Lower the forecast accuracy, higher the needed inventory to attain a specific service level. 

99% or more customer service level can be achieved by keeping extremely high levels of inventory. This could be possible if you are not aware of your demand forecasts have no idea how customer demand is generated. 

You may hear a number of supply chain experts say that demand forecast is tough and hence a waste of time. They may have fancy solutions that advocate an expensive re-engineering of the supply chain to accommodate any type of demand. Although that is technically possible, it is not a useful way to spend organizational resources. 

A small effort on the demand side will result in tremendous benefits and improved information to help the supply chain. Even if the demand is highly uncertain, you can install a demand management process that helps determine the order forecast for the lead time. 


How demand metrics affect your inventory levels? 

Historical demand forecast metrics affect the inventory cushion you need to cover forecast volatility. This is done through the safety stock measure that in statistical terms, is an estimation of confidence intervals using a one-tailed test.          

        Safety Stock = Customer Service Level * Standard Error of the Demand during the Lead time 

        Safety Stock = Customer Service Level * RMSE * Square-root of (Lead time) 

It is based on the customer service level and the deviation of the demand during the lead time. In practice, people use the daily average of the historical data as an approximation for the mean and in essence, they ignore the forecast. This correction requires breaking out the monthly forecast into a weekly or daily forecast. 

This will result in two components to the deviation namely:

  1. Total Forecast Error 
  2. Daily Distribution Error 


Safety stock calculations also need to be adjusted depending on how wide the Lead Time is compared to how accurate the daily distribution is. For example, if the monthly forecast is very accurate but the daily distribution is incorrect but the Lead Time is three weeks then, you may end up carrying incorrect safety stock during the Lead time. Generally, if the Lead Time is the same as the forecast bucket, most problems are solved. 

If you have a question or comment