Establishing dynamic expiration dates for perishables: An application of RFID and sensor technology

Our research addresses the value of information (vox) for the use of a produces time and temperature history (Trx). Using Trx information, the retailer can set expiration dates dynamically, based on known environmental conditions. This dynamically set expiration date corresponds to the maximum number of periods that inventory may remain available for sale before it must be removed from inventory and discarded (outdated). In current static practice, however, without the availability of nil, environmental conditions are not known and all units of inventory receive the same expiration date, generally predicated on worst case conditions. Our research demonstrates that information on the Tin as a product flows through the supply chain can be very valuable. Using the example of a supply chain for fresh packaged tomatoes, we quantify the value of Trx information when used for dynamic expiration date setting. We find that the vox is quite sensitive to environmental and parametric settings, ranging upwards to 90.5% with a mean of 41.2%. Our studies demonstrate that the cost savings that leads to the vox from Trx and expiration dating stems from two major sources: eliminating the chance of selling perished product, and greatly decreasing the rate at which lost sales occur. In addition, we show that when dynamic expiration dating is used, average product freshness at the time of sale increases significantly. This indicates a win-win situation where costs to the retailer are reduced, and also additional value for the consumer is created. We also extend our analysis into the impact of imperfect information and find that the vim is fairly robust, up to error levels corresponding to a mean absolute percentage error (NAPE) of approximately 12%. Median vox at those error levels is 16.5%. The impact of errors, however, differs depending on the model parameterization and we find that under certain settings, the vox can remain significant for much larger values of MAPE.