Inventory Optimization

Inventory Optimization: Challenges and Opportunities

While most companies can set up their safety stock using either desired customer service level or days of supply approach, they experience the common problems:

  • Difficult to Predict Service Level: It’s difficult to predict what service level we will get until a few months later. This is because inventory policy will not impact inventory till the purchasing lead time passes (that is, when the purchase order is received)
  • Unforeseeable Inventory Problem: It’s hard to foresee how the demand and supply planning will fare until a few months later. By that time, any damage (over stock or out of stock) will be done. And user will need to re-adjust the inventory policy and then wait again to see the results
  • Hard to Suit Demand Planning System: Every demand planning tools are different. How can we adjust the safety stock policy so that it fits best for the demand planning tool in use
  • Tough to Obtain Problematic and Opportunity Item List: It’s tough to go over tens or hundreds of thousands of items to list the problematic items ( and opportunity items) based on current inventory policy

Solution

Analytics United Inventory Optimizer is developed using unique machine learning technology and utilize historical service level and inventory to optimize the safety stock settings. Specifically

  1. Conducts machine learning using historical data. This is no normal distributed noise assumption used as is by other software vendors
  2. Optimization is tailored to demand planning tool
  3. Simulate what will happen using the current safety stock policy, identify issues and problems before they can happen
  4. Identify problematic items and opportunity items so that we can focus on them during demand forecasting and supply planning

Results

  • Customer one achieved 33% on inventory reduction while still meeting the current customer service level
  • Customer two achieved 16% on inventory reduction while meeting high customer service level by focusing demand forecasting on the top 10% problematic items