Managing cross-channel fulfillment impact within shared inventory demand systems

Patent metadata

Field	Value
Publication number	US-10423922-B2
Application number	US-201615199205-A
Country	US
Kind code	B2
Filing date	Jun 30, 2016
Priority date	Jun 30, 2016
Publication date	Sep 24, 2019
Grant date	Sep 24, 2019

Abstract

System and method for omni-channel retailer operations that integrate a network of brick-and-mortar stores with their online channel. The system and method includes calibrating a demand model for both brick-and-mortar sales and on-line channels over which a product is sold, the calibrating based upon a cross-channel fulfillment-aware inventory effect. An omni-channel sales prediction and fulfillment model is then constructed based on the calibrated demand model. Using constructed linear demand and revenue models, a plan is generated to optimize one or more: allocating of the product across physical stores, partitioning of the product for sales, and pricing of the product. Customer choices are jointly modeled across channels to allow switching, and a ship-from-store (SFS) inventory effect feature only for brick choice is applied to capture asymmetry. Inventory decision variables are introduced into an omni-channel pricing formulation to manage SFS-effect induced non-convexity and specific reformulations applied to recover a linear MIP.

First claim

The invention claimed is:

1. A computer-implemented method for a network comprising multiple physical stores or warehouses configured for storing an inventory of a product, said inventory being shared to fulfill demand by at least an on-line sales channel and a physical store sales channel at distributed locations, said method comprising:
receiving, at a processor unit, historical sales transaction data related to sales and shipment of a product at one or multiple physical stores or warehouses;
determining, using the processor unit, a decay value representing an inventory effect based on a channel switching behavior wherein on-line sales channel orders are fulfilled using product inventory at physical store sales channel locations;
modifying an inventory level at a location using said decay value;
applying a cross-channel non-linear thresholding function to the modified inventory level for said product at said physical stores;
calibrating, using the processor unit, a demand model to forecast demand of the product for each of one or more zones and each of one or more on-line sales or physical store sales channels over which the product is sold, said calibrating based upon the modified inventory level;
solving an omni-channel sales prediction and fulfillment optimization model constructed to jointly predict a demand and determine inventory effect values across all locations for a future period of time based on said calibrated demand model, said optimization model comprising a non-convex formulation as a result of the applied non-linear thresholding function, said solving comprising:
using inventory decision variables as constraints to transform the non-convex formulation into a piecewise linear function;
incorporating said piecewise linear function into said optimization model; and
solving said optimization model using a mixed integer programming (MIP) parallel processing solver unit; and
generating, from an output of said MIP solver unit, an optimized plan specifying, using the processor unit, at least one of allocation of the product across physical stores to satisfy future demand predictions for said product across said on-line sales and retail store sales channels subject to said channel switching behavior, a partitioning of the product for virtual sales, and a pricing of the product.

2. The computer-implemented method of claim 1 , wherein said calibrating comprises:
calculating from said historical sales transaction data a historical ship-from-store (SFS) flow out of every physical store or warehouse location as a function of time (t), and computing a modified inventory level amount I′(t) as a function of time at a store according to:
I ′( t )= I ( t )−SFS( t )

where I(t) is an initial inventory level amount at a point in time, and SFS(t) is the calculated SFS product out flow from said store.

3. The computer-implemented method of claim 2 , further comprising:
applying a cross-channel inventory effect function f( ) to the modified inventory level I′(t) for said product at said store, said f( ) being said non-linear thresholding function; and
modifying said demand model based on said applied a cross-channel inventory effect function f(I′(t)), said modified demand model used in said predicting demand for said product at said store for a period of time.

4. The computer-implemented method of claim 3 , further comprising:
computing a partition of store inventory using the applied cross-channel inventory effect thresholding function based on said demand predictions at said sales location.

5. The computer-implemented method of claim 3 , wherein said demand model is an attraction-based model, said modifying said demand model based on said applied cross-channel inventory effect function f(I′(t)) comprising:
computing A′ brick ( t )= A brick ( t )+β f ( I ′( t )),

where A′ brick (t) is a multivariable function representing a retailer's attractiveness for a physical retail sales channel A brick (t) as reduced by said applied inventory effect function and β is a coefficient representing a sensitivity of channel attractiveness to inventory level.

6. The computer-implemented method of claim 5 , further comprising modeling a demand level for said product through said physical retail sales channel from a starting time period as:
Predicted D B ( t )= M ( t )exp( A′ brick ( t ))/1+exp( A′ brick ( t ))+exp( A′ online ( t )); and

modeling a demand level for said product through said on-line sales channel from a starting time period as:
Predicted D O ( t )= M ( t )exp( A′ online ( t ))/1+exp( A′ brick ( t ))+exp( A′ online ( t ))

where M(t) is a market demand size model, and A′ online (t) represents an attractiveness level relating to a customer's preference to or order the product via on-line sales channel.

7. The computer-implemented method of claim 6 , wherein to construct said omni-channel sales prediction and fulfillment optimization model comprises:
creating a function that comprises a piecewise linear function of available inventory;
incorporating said piecewise linear function in an optimization model program to generate a discrete optimization problem; and
solving said discrete optimization problem using the MIP solver unit.

8. The computer-implemented method of claim 7 , wherein said creating a piecewise linear function comprises:
specifying a piecewise linear store inventory decision variable z(I, t), where z it =1 if effective inventory level I′(t) is Ī it , where i represents a store of said physical retail sales channel network and z represents a Special Ordered Set type-2 (SOS-2) variable;
computing I′(t)=sum(i) z it Ī it , where i represents a store of said sales channel network, wherein for each Ī it , specifying demand values D B (t) for said physical retail sales and D O (t) for said on-line channel as piecewise linear function of z(I, t).

9. A system comprising:
a network comprising multiple physical stores or warehouses configured for storing an inventory of a product, said inventory being shared to fulfill demand by at least an on-line sales channel and a physical store sales channel at distributed locations;
a memory storage device; and
a hardware processor in communication with said memory storage device, the hardware processor configured to:
receive historical sales transaction data related to sales and shipment of a product at one or multiple physical stores or warehouses;
determine a decay value representing an inventory effect based on a channel switching behavior wherein on-line sales channel orders are fulfilled using product inventory at physical store sales channel locations;
modify an inventory level at a location using said decay value;
apply a cross-channel non-linear thresholding function to the modified inventory level for said product at said physical stores;
calibrate a demand model to forecast demand of the product for each of one or more zones and each of one or more on-line sales or physical store sales channels over which the product is sold, said calibrating based upon the modified inventory level;
solve an omni-channel sales prediction and fulfillment optimization model constructed to jointly predict a demand and determine inventory effect values across all locations for a future period of time based on said calibrated demand model said optimization model comprising a non-convex formulation as a result of the applied non-linear thresholding function, wherein to solve said optimization model, said hardware processor is further configured to:
use inventory decision variables as constraints to transform the non-convex formulation into a piecewise linear function;
incorporate said piecewise linear function into s

Assignees

• IBM

Inventors

• Harsha Pavithra
• Subramanian Shivaram

Classifications

• G06Q10/087 Primary

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

• G06Q30/0201 Primary

  Market modelling; Market analysis; Collecting market data · CPC title

• G06Q30/0202

  Market predictions or forecasting for commercial activities · CPC title

• G06Q10/08744

  by distributed inventory management · CPC title

• G06Q10/08726

  for replenishment processing, procedures, or recommendations using forecasting or optimisation · CPC title