Using transactions data to improve consumer returns forecasting

• Published date: 01 April 2020
• Author: Michael R. Galbreth,Guangzhi Shang,Mark E. Ferguson,Erin C. McKie
• DOI: http://doi.org/10.1002/joom.1071
• Date: 01 April 2020

RESEARCH ARTICLE

Using transactions data to improve consumer returns forecasting

Guangzhi Shang

1

| Erin C. McKie

2

| Mark E. Ferguson

3

| Michael R. Galbreth

4

1

Department of Analytics, Information

System, and Supply Chain, College of

Business, Florida State University,

Tallahassee, Florida

2

Department of Management Sciences,

Fisher College of Business, The Ohio State

University, Columbus, Ohio

3

Department of Management Science,

Moore School of Business, University of

South Carolina, Columbia, South Carolina

4

Department of Business Analytics and

Statistics, Haslam College of Business,

University of Tennessee, Knoxville,

Tennessee

Correspondence

Mark E. Ferguson, Department of

Management Science, Moore School of

Business, University of South Carolina,

Columbia, South Carolina.

Email: mark.ferguson@moore.sc.edu

Handling Editor: Manpreet Hora

Abstract

Although generous return policies have been shown to have marketing benefits,

such as a higher willingness to pay and a higher purchase frequency, counter-

balancing these benefits is an increased volume of consumer returns, which pre-

sents significant operational challenges for both retailers and original equipment

manufacturers (OEMs). Since accurate return forecasts are inputs into strategic and

tactic decision support tools for operations managers, advancements in better fore-

cast accuracy can yield significant savings from the returns management practice.

We propose a forecasting approach that incorporates transaction-level data, such

as purchase and return timestamps, and predicts future return quantities using a

two-step “predict-aggregate”process. To enhance the generalizability of our frame-

work, we test it on two distinct datasets provided by a bricks-and-mortar electron-

ics retailer and an online jewelry retailer. We find that our approach demonstrates

significant forecasting error reduction, in the range of 10–20%, over benchmark

models constructed from common industry practices and the existing literature. As

our approach leverages the same data inputs as existing models, it can be easily

adapted by practitioners. We also consider a number of extensions to generalize

our approach into contexts such as restricted return time windows, new product

returns, and inflated same-day returns. Last, we discuss broad implications of

return forecast accuracy improvements in the areas such as inventory management,

staffing level, reverse logistics, and return recovery decisions.

KEYWORDS

closed-loop supply chain, consumer returns, econometrics, forecasting, retail operations

1|INTRODUCTION

Offering a generous return policy has become increasingly

popular among U.S. retailers. From large retail chains such

as Walmart and Target, to more niche stores such as

LuluLemon and Athleta, many stores promise a full money

back guarantee upon return if customers are not satisfied

with their purchases.

1

Behind the trend of lenient return

practices is the belief that consumers highly value the option

to return products penalty-free, which in turn generates

higher demand and better customer satisfaction for the

retailer (Mollenkopf, Frankel, & Russo, 2011). The recent

consumer returns studies offer more specific evidence:

Anderson, Hansen, and Simester (2009) and Heiman, Just,

McWilliams, and Zilberman (2015) estimate that consumers

value a full refund policy for an apparel item purchased

through a catalog or physical channel at 10–25% of the prod-

uct's price. It has also been shown that return-prone cus-

tomers are associated with more frequent, and higher-dollar

future purchases (Griffis, Rao, Goldsby, & Niranjan, 2012),

stronger brand loyalty (Ramanathan, 2011), and even larger

customer lifetime value (Petersen & Kumar, 2015).

Responding to this competitive pressure, Forever 21 recently

updated its exchanges-only, 14-day return policy and now

allows consumers to receive full refunds for their purchases

returned within 30 days (Amatulli, 2017). Although the new

Received: 9 July 2018 Revised: 24 October 2019 Accepted: 28 October 2019

DOI: 10.1002/joom.1071

326 © 2019 Association for Supply Chain Management, Inc. J Oper Manag. 2020;66:326–348.wileyonlinelibrary.com/journal/joom

policy aligns Forever 21 with the majority of the U.S. retail

industry, it may also introduce additional costs associated

with processing returns.

While enjoying the demand-side benefits entailed by a

full refund policy, retailers and OEMs find the management

of returns a challenging task—often ranked among the top

managerial concerns (Brill, 2015). At the macro level, indus-

try research has estimated the total value of U.S. consumer

returns to be around $260 billion annually (Kerr, 2013; Ng,

2015). The National Retail Federation (2008, 2016) has

observed a steady increase in average return rates over the

past decade, from 7 to 12% for brick and mortar retailers,

and the return rate for online retailers is estimated at near

30% (Reagan, 2016). Adding to the sheer volume of returns

are the operational costs that must occur to ensure their flow

along the reverse supply chain and value recovery through

various options. A study by Accenture (Douthit, Flach, &

Agarwal, 2011) suggests that return-related operations such

as inspection, reverse logistics, and refurbishment or dis-

posal account for 5% of a typical OEM's revenue and 4% of

a typical retailer's sales. Overall, just a 1% return rate has

been estimated to cost a large retail chain $16 million

(Douthit et al., 2011, p.9) and the whole U.S. economy $32

billion (National Retail Federation, 2016).

Perhaps the most straightforward “solution”to reducing

returns is to charge a restocking fee that limits consumers'

incentive to return. However, given the prevalent adherence

to full refund policies in practice and the strong belief in

their marketing benefits, much of the return management

burden falls instead on operations, which must minimize the

costs of returns by optimizing decisions along the reverse

supply chain, while treating the return volume as exoge-

nous.

2

Indeed, the operations literature has proposed many

decision support tools for managing returns such as retail

store inventory (Ketzenberg & Zuidwijk, 2009), disposition

decisions (Pince, Ferguson, & Toktay, 2016), and reverse

logistics network design (Guide, Souza, Van Wassenhove, &

Blackburn, 2006). All of these models share the common

trait that an accurate return forecast is critical for their suc-

cessful implementation. The objective of this study, there-

fore, is to develop a forecasting model to help firms better

manage the costs and tactical decisions associated with

processing consumer returns. In the following, we discuss in

detail how accurate return forecasts aid a manager's opera-

tional decisions.

From a retailer's perspective, the influx of returns

requires adjustments to inventory policies, since the current

level of inventory replenishment should consider the volume

of future returns (Ketzenberg & Zuidwijk, 2009). In such

cases, the forecasting of product returns becomes a prerequi-

site for determining the order quantity. Similarly, many other

inventory-related tactical decisions along the reverse supply

chain also involve return forecasting, including determining

the optimal number of parts for refurbishing and repairs, the

number and location of stock points, and the allocation of

inventories across them (Fleischmann et al., 1997).

As the prominence of returns management increases,

how to best staff the return counter using a reliable returns

forecast also becomes part of a retailer's labor planning pro-

cess, complementing the existing traffic-based sales force

staffing plans (Chuang, Oliva, & Perdikaki, 2016). Further-

more, with returns moving upstream in the reverse supply

chain, distribution centers and refurbishment centers face

staffing challenges of similar nature, yet technically more

complex. For example, WalMart sends returns to one of its

six regional return centers across the U.S., where employees

sort the returned goods into four tiers. The return flow is

highly variable: although 45 million pallets of returns are

processed annually, over 40% of this volume occurs in

January and February after the holiday season (Souza,

2013). As a result, WalMart staffs with seasonal employees.

The performance of this approach is highly dependent on the

accuracy of the return forecast. Our conversation with man-

agers at the Bose refurbishment center in South Carolina

3

reveals a similar staffing problem for their part-time labor

force.

Once the merchandise is returned, the value recovery

options include reselling, reusing items for parts,

refurbishing, and recycling. The salvage value of returned

merchandise largely depends on the recovery channel. For

consumer electronics, the salvage value ranges from as high

as 70% value recovery by selling through an online resell

channel to as low as 20% by selling to liquidators who buy

in pallets (Ng, 2015). For example, Optoro, a third-party

reverse logistics provider specializing in value recovery of

electronic products, “recovers”the value of returns through

these various channels based on expected return volume,

consumer preference for open-box items, and discount store

demand (Tabuchi, 2015). An accurate return forecast is a

crucial input parameter. Last, for an OEM who accepts

returns from retailers,

4

the recovery decision often involves

allocating returns between two options: restocking for open-

box sale and earmarking for future warranty demand. Pince

et al. (2016) derive optimal strategies under such a setting

where the future return forecast is again one of the critical

inputs in their disposition decision model.

Despite more profitable options being available, some

returned products accumulate in a warehouse and are even-

tually routed to a low payback salvage channel. Moreover,

an estimated 5 billion pounds of returned items end up in

landfills each year because their value has depreciated past

the point of being profitable to remarket (Phillips, 2018).

One of the main reasons for this wasteful practice is that

return volumes are not built into the logistics network since

SHANG ET AL.327