The Value of Inventory Accuracy in Supply Chain Management: Correlation Between Error Sources and Proactive Error Correction

Abstract

One of the key elements in supplay cahin management is accurate information. Decision makers are aware of inaccuracies in inventory levels and, therefore, routinely conduct inventory reviews to correct the discrepancies between IT records and actual inventory. Several studies have investigated error sources and the cumulative effect of errors on holding costs, shortage costs, order-up-to levels and time between inventory counts. In most works, the errors were independent of the demand, which is neither realistic nor accurate. Here we use familiar inventory errors and information scenarios already proposed in several previous papers. We offer a model that considers the correlation between inventory errors and demand. The effect of the relationship between the random variables is tested in the context of several different scenarios. Each scenario contains a different level of information about the underlying demand and inventory errors. We then analyze the effect of changes of the covariance on the cost and time between inventory counts in each scenario. Using these results we formulate the value of information and its dependence on the covariance. We use analytical methods to draw conclusions regarding single parameter set cases and a numerical full factorial study for average multiparameter cases. In both settings, we show that the value of information decreases as the covariance increases. Moreover, the reduction is more significant when the information scenario makes less assumptions. The same behavior is observed in stock review frequency. As covariance increases, the optimal number of periods between inventory reviews drops sharply. Finally, we propose several simple methods for proactive error correction. We show that without prior knowledge, these methods perform better than the basic information scenario. Using these results we are able to formulate recommendations for businesses with different profiles of correlation between demand, and demand and errors, e.g., automated warehouses with weak correlation compared to grocery stores.