top of page

Using algorithms for spares management. Easy, right?

Academic references exist for inventory optimization to reflect current reliability and current market conditions. Within manufacturing, managing spares with algorithms is a strategic step to manage cash while ensuring you have the right amount of spares.

The two classic calculations for Reorder Point (ROP) and Economic Order Quantity (EOQ) bring in common inventory inputs like issues and lead time. Introducing more control constants like purchasing cost, the cost to process a purchase order, inventory risk, and inventory holding cost allows you to fine-tune the equations to give you the values you are looking for. The challenge is that inventory patterns have atypical behaviors that can affect these arithmetically derived inventory values, catching you sometimes without the required inventory in the most critical of times.

Consider how the algorithms would behave when the amount issued had an extra zero issued by mistakes (e.g. 100 parts issued versus 10)? Or what happens when the historical data has a unit of measure for the same part number that changed from a “box” to an individual “count”? Or what happens when we need 100 bags for a baghouse job but the ROP calls for one and the EOQ calls for 20 because lead times are three days? We sure can’t do much with one to twenty bags for a baghouse preventive maintenance job that requires you to change out all of them on an annualized basis.

Running this inventory optimization process many times, I have learned that you have to write logic within the algorithms to look for common asset management and inventory management scenarios. Then intense scrutinization needs to take place to vet out scenarios such as these. Based upon the results, write “smart” logic within the algorithms to adjust. For example, instead of looking at issued parts, consider a net calculation to capture the 90 that get returned in the example above. In the baghouse example, the application of some standard deviation and monitorization of the mean can highlight the optimum amount needed for a job.

The algorithms are only as smart as the inputs and the maturity of the logic. Accepting that the algorithms that manage your inventories should be organically changing with effective vetting of scenarios is critical for the organization to have the most optimized amount.

You want to evaluate future borrowers, but in order to train an algorithm that will help you identify future defaults, you have to train it and evaluate it on past data. - Anthony Goldbloom, Founder and CEO Kaggle


bottom of page