Optimizing stock‐keeping unit selection for promotional display space at grocery retailers

• DOI: http://doi.org/10.1002/joom.1075
• Author: Su‐Ming Wu,Olga Pak,Mark Ferguson,Olga Perdikaki
• Published date: 01 July 2020
• Date: 01 July 2020

RESEARCH ARTICLE

Optimizing stock-keeping unit selection for promotional

display space at grocery retailers

Olga Pak

1

| Mark Ferguson

2

| Olga Perdikaki

2

| Su-Ming Wu

3

1

Supply Chain and Information Systems, Pennsylvania State University, State College, Pennsylvania

2

Management Science, University of South Carolina, Columbia, South Carolina

3

Oracle Retail Science, Burlington, Massachusetts

Correspondence

Mark Ferguson, University of South

Carolina, Columbia, SC.

Email: mark.ferguson@moore.sc.edu

Funding information

Oracle External Research Office, Grant/

Award Number: 10003885

Handling editor: Alan Mackelprang

Abstract

Promotional displays, such as end-of-aisle displays, provide a stimulus for dis-

cretionary and incremental sales at grocery stores, offering a powerful yet

affordable tool to boost profit. In this study, we examine a store manager's

choice of which stock-keeping units (SKUs) from a given category to assign to

a promotional display space. While the academic literature and retail software

solution providers offer a variety of optimization solutions for the assortment

optimization problem, there is very little guidance for retailers on how to opti-

mally determine when and what products to place on their promotional dis-

play space. Consequently, retailers often default to simple heuristics, which are

typically suboptimal from a profit maximization standpoint, when making this

decision. Thus, there is a need for a decision support tool to facilitate the prod-

uct selection for promotional display space. Using a grocery store sales transac-

tion data set, we demonstrate how to measure the incremental lift in sales of

placing a particular SKU on promotional display space. Our optimization

model includes the incremental lifts (from the estimation method) combined

with the estimated base-sales rates and profit margins of each SKU so that the

profit-maximizing SKU can be chosen for a promotional display space for each

week of the year. Placing a SKU on promotional display can result in a signifi-

cant lift in sales. For example, the estimated average display effect (i.e., sales

lift) for the beer category across all SKUs and all weeks is 27%, which makes

promotional display a very effective tool for stimulating incremental product

sales. Overall, our methodology results in at least 1.6 times improvement in

incremental profit when compared to a common industry benchmark. We also

test our methodology on two additional product categories and demonstrate

that it performs equally well across product categories. Our work provides an

easy-to-implement, promotional display SKU-selection methodology that

includes both an estimation and an optimization model. Our estimation model

can handle an extensive and complex product assortment and accounts for

Received: 15 July 2018 Revised: 26 October 2019 Accepted: 5 November 2019

DOI: 10.1002/joom.1075

J Oper Manag. 2020;66:501–533. wileyonlinelibrary.com/journal/joom © 2019 Association for Supply Chain Management, Inc. 501

important aspects of promotional activities such as the cannibalization of the

inner aisle sales and halo effects. Our optimization model is flexible enough to

consider practical aspects such as common business rules that restrict the selec-

tion of the same SKU over a consecutive set of weeks, display-related change-

over costs, and slotting fees offered by the manufacturers. Overall, our study

underscores the importance of making effective promotional display decisions.

KEYWORDS

grocery retailers, optimization, promotion

1|INTRODUCTION

The grocery store sector is one of the major players in the

US retail and foodservice industry. In 2017 alone, its total

seasonally adjusted annual sales approached $639 million

(US Census Bureau, 2018). Grocery store managers have

several operational and marketing levers at their disposal

to increase the profitability of their stores. Operational

levers include store layout, the product assortment, and

the stocking/replenishment policies while marketing

levels include the pricing and promotion of the products

that were previously selected in the assortment decision.

While all of these levers have received considerable atten-

tion in the academic literature, there is an additional

lever at the operations/marketing interface that has been

mostly overlooked—optimizing the product selection for

which stock-keeping units (SKUs) to place on a limited

amount of promotional display space. This lack of atten-

tion is surprising because an increase in product exposure

is known to result in one of the most coveted shopper

outcomes, persuading the shopper to make impulse pur-

chases (purchases of products that the shopper did not

originally plan on purchasing) of full-priced products

(Inman & Winer, 1998; Kollat & Willett, 1967).

It is well understood within the grocery industry that

impulse purchases are driven by product exposure. Com-

mon impulse buy items such as candy and magazines are

almost universally placed near the checkout lanes of

most national grocery store chains. A recent industry

study of 2 million grocery store shoppers shows that

products placed on a promotional display are seen by

nearly twice as many store visitors as products that are

located in the inner aisles of the store (Wade, 2014). Pro-

motional display space increases product visibility

(Garrido-Morgado & Gonzalez-Benito, 2015) and

increases the number of impulse buys (Kacen, Hess, &

Walker, 2012). Thus, placing a product on promotional

display is a very valuable tool for stimulating incremental

product sales (via impulse buys) for new or existing prod-

ucts by exposing products to potential customers that

may not have originally planned to purchase an item from

that particular product category. An example of an

unplanned impulse buy is when a father stops by a store

to pick up some baby formulae but, upon passing bya pro-

motional display space of beer, decides to include a six-

pack of beer in his purchase. Promotional display space

includes endcap displays, front walls, wings, display racks,

countertops, gondola checkouts, showcases, specialty

shelves, dump bins, and dump tables. They are located in

high-traffic and high-visibility areas around and on the

outside edges of a store's shelf space perimeter.

A typical grocery store perimeter offers fresh, high-

margin foods like produce, bakery items, meats, dairy

products, and deli (Johnson, 2018). The center-store

aisles are located inside the perimeter and include 12 or

more aisles with shelf-stable consumer-packaged goods

(CPG) ranging from toiletries to frozen pizzas (Rupp,

2015). Promotional endcap displays face the perimeter of

the store, while other displays can be placed in-between

the aisles, the lobby area, or in other high traffic areas of

the store (Frazier, 2014). Recent technological advances

like radio frequency identification tags attached to retail

shopping carts and shopper movement video-reading

software show that an overwhelming majority of grocery

shoppers shop the perimeter of a grocery store and avoid

entering the center aisles (Shop Association, 2016;

Sorensen, 2008). As a result, in contrast to inner-aisle

merchandise, products located on the endcaps enjoy sub-

stantially more face time (Larson, Bradlow, & Fader,

2005). A large-scale industry study of millions of grocery

shoppers found that products placed on promotional dis-

plays receive 93% more exposure than those on the inner

aisles of the same store (Shop Association, 2016). The

implication of this behavior is that products placed on

endcaps have a much higher probability of driving

impulse purchases (Bezawada, Balachander, Kannan, &

Shankar, 2009; Nakamura et al., 2014; Phillips &

Bradshaw, 1993), where the lift in sales comes from pur-

chases that are often not combined with a price reduc-

tion. Thus, promotional display space does not face the

502 PAK ET AL.

trade-off that other common promotion levers such as

price discounting do, since the product margins are often

maintained and combined with a lift in unit sales.

1.1 |Problem and motivation

The choice of which SKUs to place on promotional display

space is also very different from the more commonly dis-

cussed assortment planning/optimization problem. Assort-

ment optimization is concerned with selecting a subset of

products/items to include in a store's assortment among a

set of potential candidates, factoring in important issues

such as space limitations and potential substitution effects

as the demand of an item depends on the presence of

other items in the assortment. In contrast, promotional

display optimization is concerned with selecting a subset

of products/items to place on promotional display from a

given assortment that the store already carries. Since the

effectiveness of promotional displays relies on generating

impulse purchases, frequently changing the products on

the promotional displays is necessary for impulse pur-

chases to continue to occur. Hence, while the nature of

promotional display optimization is dynamic and may

change from week to week, assortment optimization gives

recommendations meant to stand for a considerable

length of time, as it would be confusing to shoppers to

have the store's assortment and shelf contents change as

frequently as promotional displays do.

Due to these import ant differences in the overa ll objec-

tive of the problem, conventional methods for the assort-

ment optimization problem are not appropriate for the

promotional display allocation problem for the following

reasons: First,promotional display decisions,unlike assort-

ment planning decisions, provide retailers with an impor-

tant degree of freedom that is, the ability to adjust a

product's visibility, generating impulse buys without

changing the overall product assortment. Thus, an impor-

tant characteristic of a promotional display decision sup-

port tool is the ability to measure the effectiveness of

promotional display space in terms of generating incre-

mental sales rather than total sales. Second, the nature of

substitution that shapes consumer demand is different in

assortment planning decisions as compared to promotional

display allocation decisions. In assortment planning, an

important consideration is consumers' willingness to sub-

stitute their original preferred product if it is not offered in

the assortment, referred to as assortment-based subs titu-

tion (Kok & Fisher, 2007). In the context of promotional

display, however, substitution arises from the increased

visibility that a product achieves when placed in a promo-

tional display relative to a product located in an inner-

aisle. Given these distinct differences in the types of

substitution, there is a need for distinct methods to address

these two types of decisions. Finally, a methodology that

can assist with promotional display allocation decisions at

a given store location should be able to estimate sales lifts

of SKUs that have never been chosen for promotional dis-

play at that particular store. Thus, sales data from numer-

ous stores is needed to estimate these sales lifts.

Given the low-cost

1

but high-return nature of promo-

tional display space, it is surprising that, while the aca-

demic literature and retail software solution providers

offer a variety of optimization solutions for the assortment

optimization problem (Chong, Ho, & Tang, 2001; Kok &

Fisher, 2007; Rooderkerk, van Heerde, & Bijmolt, 2013),

there is very little guidance for retailers on how to opti-

mally determine whenand what products to place on their

promotional display space.

2

Consequently, retailers often

default to simple heuristics when making this decision,

such as selecting the best-selling SKUs or SKUs where the

OEM offers the most generous trade allowances for plac-

ing their products on promotional displays. In some cases,

large grocery chains make the decision at the firm's head-

quarters, thus imposing the same selection of SKUs across

different geographical store locations and time periods.

Such heuristics are often far from optimal because some

SKUs do not need any additional exposure to continue to

be a best-selling SKU and what may be a profitable SKU

at one geographic location may not be so at a different

location. In the absence of any decision support system to

help with this decision, it is difficult to know which SKUs

are the most profitable to allocate to a limited number of

promotional display locations at each store.

1.2 |An overview of our methodology

Next, we discuss our methodology for providing such a

decision support system. While the selection of which

product category to put on promotional display is rather

straight forward, given that there are only a few product

categories that generate significant impulse buying

demand (i.e., candy versus toilet paper), choosing a spe-

cific SKU from an existing category is significantly more

challenging. Thus, we start by assuming that a store man-

ager has already decided which category to place on a

promotional display space such as an endcap at his store,

and offer an initial effort in filling this research gap by

providing a methodology that helps select the profit-

maximizing SKU from within this category to place on

promotional display for each week.

3

We first discuss how the store manager can short-list

the potential SKU candidates (from that store assortment

within the preselectedcategory) to put on promotional dis-

play and offer an estimation methodology that allows him

PAK ET AL.503