Systems and methods of utilizing multiple forecast models in forecasting customer demands for products at retail facilities

Patent metadata

Field	Value
Publication number	US-10740773-B2
Application number	US-201615367530-A
Country	US
Kind code	B2
Filing date	Dec 2, 2016
Priority date	Dec 9, 2015
Publication date	Aug 11, 2020
Grant date	Aug 11, 2020

Abstract

In some embodiments, apparatuses and methods are provided herein useful to forecasting product demand. In some embodiments, a system comprise a forecasting control circuit to: apply each of a plurality of different models to forecast demand of a first product over a first historic period generating historic forecasted demands of the first product, wherein at least a first model uses selected one or more variables that are predicted to have an uncharacteristic effect on predicted demand; select one of the models and apply the model in generating a forecasted future demand, wherein the selection of the model is based on a difference between each of the generated historic forecasted demands and actual sales; and identify actions to modify inventory of the first product at the first shopping facility based on the forecasted future demand.

First claim

What is claimed is:

1. A system to forecast product demand at retail facilities, comprising:
a product forecasting system comprising a control circuit and memory storing computer instructions that when executed cause the control circuit to:
apply, for each of hundreds of different products, each of a plurality of different models to forecast demand of a first product over a first historic period of time to generate a plurality of different historic forecasted demands of the first product at a first shopping facility, wherein: at least a first model uses selected one or more variables, of tens of different variables maintained in a variable database of status data corresponding to each of the tens of different variables, that are predicted to have an uncharacteristic effect on predicted demand of the first product at the first shopping facility in generating a corresponding first historic forecasted demand of the different historic forecasted demands, and at least a second model does not use the variables maintained in the variable database in generating a corresponding second historic forecasted demand of the different historic forecasted demands;
select one of the plurality of different models and apply the selected one of the models in generating a forecasted future demand of the first product at the first shopping facility over a fixed future period of time, wherein the selection of the one of the plurality of different models is based on a difference between each of the generated historic forecasted demands and actual sales of the first product over the first historic period of time;
identify at least a first action to modify inventory of the first product at the first shopping facility based on the forecasted future demand, comprising identifying a quantity of the product to be transferred to the first shopping facility based on the forecasted future demand and relative to an on-hand quantity of the product at the first shopping facility, and identify one or more other shopping facilities from which the quantity of the first product is to be transferred; and
carry out at least the first action initiating a transfer, from the one or more other shopping facilities, of the quantity of the first product to the first shopping facility and reallocating and redistributing the quantity of the first product to the first shopping facility consistent with and satisfying the forecasted future demand of the first product at the first shopping facility.

2. The system of claim 1 , wherein the control circuit is further configured to determine an error factor for each of the different historic forecasted demands relative to the actual sales of the first product, where a first error factor corresponding to the selected model has a lowest error factor.

3. The system of claim 1 , wherein the control circuit is further configured to:
determine an error factor for each of the different historic forecasted demands relative to the actual sales of the first product; and
confirm, prior to selecting the one of the plurality of different models, that the first error factor corresponding to the selected one of the plurality of different models is less than an error factor of an additional historic forecasted demand generated by an alternative inventory replenishment application.

4. The system of claim 1 , wherein the control circuit is further configured to:
determine an error factor for each of the different historic forecasted demands relative to the actual sales of the first product over the first historic period of time;
apply each of the plurality of different models to forecast a secondary demand of the first product over a second historic period of time that corresponds in duration to the fixed future period of time to generate a plurality of additional different historic forecasted demands of the first product at the first shopping facility;
determine an additional error factor for each of the additional different historic forecasted demands relative to additional actual sales of the first product over the second historic period of time; and
determine, for at least the selected one of the plurality of different models, a confidence factor based on the corresponding error factor and additional error factor wherein the selection of the one of the plurality of different models comprises confirming the confidence factor corresponding to the selected one of the plurality models has a predefined relationship with a confidence factor threshold.

5. The system of claim 4 , wherein the control circuit is further configured to adjust the actual sales as a function of on-hand inventory of the first product at the first shopping facility over at least a portion of the first historic period, wherein the determining the error factor for each of the different historic forecasted demands determines the error factor relative to the adjusted actual sales of the first product.

6. The system of claim 5 , wherein the control circuit in selecting the one or more variables selects the one or more variables as a function of a residual between historical sales data of the first product relative to a previously forecasted demand forecasted without consideration of the change in status of the one or more variables.

7. The system of claim 1 , wherein the control circuit further applies a first set of two or more models comprising the first model, wherein each of the two or more models of the first set of models uses the selected one or more variables; applies a second set of two or more models comprising the second model, wherein each of the two or more models of the second set of models do not use the one or more variables; and compares forecasted future demand from each model of the first set of models to forecasted future demand determined from each model of the second set of models in confirming an uncharacteristic change in demand.

8. The system of claim 1 , wherein the control circuit is further configured to, in parallel and independent of predicting whether there is an uncharacteristic demand of the first product at the first shopping facility:
receive, for each of the hundreds of products at the first shopping facility and from the variable database, a change in status data corresponding to selected one or more variables, of the tens of different variables, that are predicted to have effects on predicted demand of corresponding ones of the hundreds of products at the first shopping facility;
forecast, independent of the other of the hundreds of products, a forecasted future demand for each of the hundreds of product at the first shopping facility by: applying one or more of a set of models using the selected one or more variables to historic data relative to the product being forecasted, applying one or more of a set of models that do not use the one or more variables to historic data relative to the product being forecasted, and confirming there is a change in demand for multiple of the hundreds of product relative to the first shopping facility; and
identify one or more additional actions to modify inventory at the first shopping facility relative to each of the multiple of the hundreds of products in response to the forecasted future demand resulting in part from changes in conditions corresponding to the first shopping facility as reflected in the change of status of the selected one or more variables corresponding to each of the one or more of the hundreds of products.

9. The system of claim 1 , wherein the control circuit is further configured to evaluate inventory at multiple other shopping facilities, comprising the one or more other shopping facilities, relative to the on-hand quantity of the first product at the first shopping facility, and in-stock quantities of

Assignees

• Walmart Apollo Llc

Inventors

• Li Ming
• Tripathi Rahul
• Dey Anindya S
• Mitra Mainak
• Barnwal Manish K
• Pop Mariana
• Srivasatava Nikesh K
• Steel Stephen
• Bharadwaj Yashas
• Ryan Aaron
• Schaefer Paul
• Ramireddy Lakshmi Bhaswanth

Classifications

• G06Q30/0202 Primary

  Market predictions or forecasting for commercial activities · CPC title

• G06Q10/087

  Inventory or stock management, e.g. order filling, procurement or balancing against orders · CPC title