Predicting inventory record-keeping errors with discriminant analysis: A field experiment

As more organizations adopt perpetual inventory monitoring systems, concerns about discrepancies between physical counts and perpetual record balances are likely to increase. Depending on the direction of error, these discrepancies can lead to unnecessary inventory holding costs or to delays in production. Ultimately, they can affect relationships with customers and the financial viability of an organization. This study involved a stockroom experiment that was designed to assess the causes of inventory record errors in a manufacturing setting. Data from the experiment were analyzed using discriminant analysis, and the results indicated that the following variables influenced inventory record accuracy: unit value, weigh-counting, quantity on-hand, dollar value of the stock on-hand, number of places the part was used, and the stockroom staff’s rating of the error likelihood for a part. Contrary to expectations, neither the quantity per transaction for a part nor the number of transactions for a part were significant predictors of error. Experimental controls, as well as the nature of the organizations’s inventory system, suggest that the results may be generalized to similar settings. Moreover, the research methods described here may be adopted or modified by firms wishing to develop diagnostic models of their own.