Forecasting Returns for Retail Demand Planning

Abstract

A system and method are disclosed for returns forecasting for a retail store inventory of one or more supply chain entities. Embodiments include receiving a current state of items in a supply chain network, receiving a sales time series and a returns time series, computing a returns forecast comprising an expected quantity of a particular product to be returned for a future time period using a sales forecast a and a transfer function estimated from the sales time series and the returns time series.

What is claimed is: 
     
         1 . A computer-implemented method for returns forecasting for a retail store inventory of one or more supply chain entities, comprising:
 receiving, by a computer, a sales time series comprising a sequence of sales quantities of a particular product and a returns time series comprising a sequence of returns quantities of the particular product; and   computing, by the computer, a returns forecast comprising an expected quantity of the particular product to be returned for a future time period, the returns forecast computed using a sales forecast and a transfer function, the transfer function estimated from the sales time series and the returns time series and comprising one or more weights, each of the one or more weights associated with a time period and comprising a probability that the particular product will be returned during the associated time period.   
     
     
         2 . The computer-implemented method of  claim 1 , further comprising:
 recording, by a point of sale system, one or more transactions of one or more items of the particular product by scanning an identifier associated with each of the one or more items and identifying the particular product based, at least in part, on the scan;   tabulating, by the computer, the one or more transactions of the one or more items of the particular product; and   generating, by the computer, the sales time series and the returns time series based, at least in part, on the tabulated one or more transactions.   
     
     
         3 . The computer-implemented method of  claim 1 , wherein the one or more transactions comprise one or more of a sales transaction and a returns transaction. 
     
     
         4 . The computer-implemented method of  claim 1 , wherein the computer computes the weights of the transfer function by minimizing the error between an estimated returns quantity and an actual returns quantity using a normalized least mean square method. 
     
     
         5 . The computer-implemented method of  claim 4 , further comprising:
 updating, by the computer, the weights of the transfer function by computing an error between a quantity of items of the particular product predicted by the transfer function for a time step and the actual quantity of items returned.   
     
     
         6 . The computer-implemented method of  claim 5 , wherein the returns forecast comprises an expected quantity of returned items from a customer segment, the customer segment identified based, at least in part, on a customer behavior the influences the likelihood of a customer in the customer segment to return an item at a particular location. 
     
     
         7 . The computer-implemented method of  claim 6 , wherein the returns time series comprises an aggregation of a quantity of returned items from a second product. 
     
     
         8 . A system of returns forecasting for a retail store inventory of one or more supply chain entities, comprising:
 a computer, comprising a processor and a memory, and configured to:
 receive a sales time series comprising a sequence of sales quantities of a particular product and a returns time series comprising a sequence of returns quantities of the particular product; and 
 compute a returns forecast comprising an expected quantity of the particular product to be returned for a future time period, the returns forecast computed using a sales forecast and a transfer function, the transfer function estimated from the sales time series and the returns time series and comprising one or more weights, each of the one or more weights associated with a time period and comprising a probability that the particular product will be returned during the associated time period. 
   
     
     
         9 . The system of  claim 8 , further comprising:
 a point of sale system, wherein the point of sale system records one or more transactions of one or more items of the particular product by scanning an identifier associated with each of the one or more items and identifying the particular product based, at least in part, on the scan.   
     
     
         10 . The system of  claim 9 , wherein the computer is further configured to:
 tabulate the one or more transactions of the one or more items of the particular product; and   generate the sales time series and the returns time series based, at least in part, on the tabulated one or more transactions.   
     
     
         11 . The system of  claim 8 , wherein the one or more transactions comprise one or more of a sales transaction and a returns transaction. 
     
     
         12 . The system of  claim 8 , wherein the computer computes the weights of the transfer function by minimizing the error between an estimated returns quantity and an actual returns quantity using a normalized least mean square method. 
     
     
         13 . The system of  claim 12 , wherein the computer is further configured to:
 update the weights of the transfer function by computing an error between a quantity of items of the particular product predicted by the transfer function for a time step and the actual quantity of items returned.   
     
     
         14 . The system of  claim 13 , wherein the returns forecast comprises an expected quantity of returned items from a customer segment, the customer segment identified based, at least in part, on a customer behavior the influences the likelihood of a customer in the customer segment to return an item at a particular location. 
     
     
         15 . A non-transitory computer-readable medium embodied with software, the software when executed configured to forecast returns for a retail store inventory of one or more supply chain entities by:
 receiving a sales time series comprising a sequence of sales quantities of a particular product and a returns time series comprising a sequence of returns quantities of the particular product;   computing a returns forecast comprising an expected quantity of the particular product to be returned for a future time period, the returns forecast computed using a sales forecast and a transfer function, the transfer function estimated from the sales time series and the returns time series and comprising one or more weights, each of the one or more weights associated with a time period and comprising a probability that the particular product will be returned during the associated time period.   
     
     
         16 . The non-transitory computer-readable medium of  claim 15 , wherein the software when executed is further configured to:
 record one or more transactions of one or more items of the particular product by identifying the particular product based, at least in part, on a scan of an identifier associated with each of the one or more items.   
     
     
         17 . The non-transitory computer-readable medium of  claim 16 , wherein the software when executed is further configured to:
 tabulate the one or more transactions of the one or more items of the particular product; and   generate the sales time series and the returns time series based, at least in part, on the tabulated one or more transactions.   
     
     
         18 . The non-transitory computer-readable medium of  claim 15 , wherein the one or more transactions comprise one or more of a sales transaction and a returns transaction. 
     
     
         19 . The non-transitory computer-readable medium of  claim 15 , wherein computing the weights of the transfer function o minimizing the error between an estimated returns quantity and an actual returns quantity using a normalized least mean square method. 
     
     
         20 . The non-transitory computer-readable medium of  claim 19 , wherein the software is further configured to:
 update the weights of the transfer function by computing an error between a quantity of items of the particular product predicted by the transfer function for a time step and the actual quantity of items returned.