Received: 7 June 2023 Revised: 16 September 2024 Accepted: 19 September 2024

DOI: 10.1002/joom.1338

RESEARCH ARTICLE

Targeting online sales through last-mile delivery

platform integration

Kevin H. Park 1 | Xiaodan Pan 2 | Martin E. Dresner 3

1
School of Business Administration, University of Dayton, Dayton, Ohio, USA
2
John Molson School of Business, Concordia University, Montreal, Quebec, Canada
3
Robert H. Smith School of Business, University of Maryland, College Park, Maryland, USA

Correspondence
Kevin H. Park, University of Dayton, Abstract
300 College Park, Dayton, OH, USA. We analyze channel integration between a last-mile delivery platform and a
Email: kpark1@udayton.edu
general merchandise retailer in two distinct stages: (1) platform delivery access
Handling Editor: Rui Sousa (PDA), where the retailer continues to offer standard delivery through its own
website but directs customers to the platform's website for new same-day deliv-
ery; and (2) integrated delivery access (IDA), where customers can continue to
use same-day delivery service at the delivery platform website but can purchase
products in a single order with both same-day and standard delivery options at
the retailer's website. We perform a quasi-experiment using consumer spend-
ing data from retailer, target, and delivery platform, Shipt. We find that PDA
provides positive impacts to the delivery platform through increased sales.
IDA, on the other hand, increases the retailer's online channel sales but does
not impact the delivery platform's sales. Moreover, we find that the positive
effects of PDA on the delivery platform's sales are stronger in markets where
online grocery penetration is lower, indicating that the effects were likely
driven by increased purchases for groceries. Finally, the positive effect of IDA
on the retailer's online channel sales is stronger in markets where the retailer
has a greater loyal customer base and online grocery penetration is lower.

KEYWORDS
channel capabilities, channel integration, delivery platform, last-mile delivery

Highlights
• The study analyzes two stages of channel integration between a delivery
platform and a retailer of time sensitive and non-time sensitive products.
• When the online shopper needs to divert from the retailer's website to the
delivery platform's website to purchase time-sensitive products, the delivery
platform's sales increase.

This is an open access article under the terms of the Creative Commons Attribution-NonCommercial-NoDerivs License, which permits use and distribution in any
medium, provided the original work is properly cited, the use is non-commercial and no modifications or adaptations are made.
© 2024 The Author(s). Journal of Operations Management published by Wiley Periodicals LLC on behalf of Association for Supply Chain Management, Inc.

J Oper Manag. 2024;1–25. wileyonlinelibrary.com/journal/joom 1

2 PARK ET AL.

• When the shopper can purchase both time-sensitive and non-time-sensitive

products from the retailer's website, the retailer's sales increase.
• After both levels of integration, sales for both the retailer and delivery plat-
form are higher, signaling a win-win result for the partnership.

1 | INTRODUCTION non-time-sensitive products through its website, Target.

com, while Shipt specialized in the order and same-day
Online shopping and same-day home delivery have delivery of time-sensitive products such as groceries. The
become increasingly popular among consumers, who cite acquisition of Shipt by Target led to Target integrating
factors such as convenience, safety, time savings, and Shipt's same-day delivery capabilities into its online oper-
laziness, for their use of online services (Redman, 2022). ations, thus allowing Target to provide order taking and
From January 2018 to January 2020, for example, online delivery capabilities for both time-sensitive and non-
grocery delivery sales tripled in value in the time-sensitive products.
U.S. (Gessner, 2020), with one in six American con- The integration of Target and Shipt operations was
sumers in 2022 claiming that they ordered online grocer- accomplished in two stages. We term the first stage, plat-
ies for delivery at least once per week (Stern, 2022). The form delivery access (PDA), and the second stage, inte-
rise of digitally-engaged consumers has contributed to grated delivery access (IDA). Prior to PDA, Target offered
this trend (Thomas, 2017), as have changing shopping ordering capabilities through its website, Target.com, for
habits developed during the Covid-19 pandemic. non-time-sensitive products, with standard delivery ser-
Although retailers can develop and manage their own vices (e.g., 2-day or 5-day) undertaken by logistics firms,
same-day delivery service capabilities for time-sensitive such as UPS and FedEx. Shipt, on the other hand, offered
products, offering these services can be costly. Firms need same-day delivery services on behalf of retail partners for
to maintain the labor force required to pick and ship time-sensitive products, using contract labor to pick and
delivered orders, as well as to develop transportation deliver orders.
routes with sufficient densities (Sousa et al., 2020). There- In the first stage of integration, PDA, that occurred
fore, most retailers in the U.S. providing same-day deliv- shortly after Target's acquisition of Shipt, Target was
ery for time-sensitive products, including large national added as a retail client of Shipt. Target continued to offer
chains such as Costco, offer these last-mile services standard delivery services through its website, Target.
through partnerships with technology-based, delivery com, but was then able to direct customers wishing
platforms, such as Instacart and Shipt. The platforms same-day delivery of Target products to the Shipt website
operate by providing a menu of items available for pur- to place orders for delivery.
chase on behalf of retailer clients, and by hiring indepen- In the second stage of integration, IDA, Target inte-
dent contractors to execute shopping and delivery tasks grated Shipt's ordering and same-day delivery capabilities
(Alnaggar et al., 2021; Boyer & Hult, 2005). into its own website. Target was then able to offer both
To obtain the greatest benefits from the order and standard delivery and same-day delivery services through
delivery capabilities offered by delivery platforms, the Target.com website. Shipt continued to offer its capa-
retailers need to integrate their operations with the plat- bilities in online orders and same-day delivery for Target
forms. The focus of this paper is on the impact of retail products.
channel integration on the performance of retailers and Figure 1 provides an illustration of Target's website
their same-day delivery platform partners (Fang after IDA. A new fulfillment option on the website,
et al., 2023; Oh et al., 2012). Using consumer spending “Same Day Delivery with Shipt”, indicates that orders
data on a national retailer and the delivery platform it may be placed on the Target.com website for same-day
acquired, we define two levels of integration and then delivery. As illustrated, after IDA, customers could order
analyze how these two levels of integration impact the Nestle Drumsticks through Target.com with same-day
sales performance of the retailer and the delivery delivery, a service previously available only via Shipt dur-
platform. ing the PDA stage. Women's Birgitte Winter Boots could
U.S. “big box” retailer, Target, acquired the delivery also be ordered through same-day delivery in addition to
platform, Shipt, in December 2017, bringing together standard delivery.
Shipt's and Target's retail and delivery capabilities. Prior Although this study focuses on the integration of Tar-
to the merger, Target offered standard delivery of get and Shipt, it is important to note that other retailers
PARK ET AL. 3

FIGURE 1 Integrated delivery access (Source: Target.com).

have undertaken similar integration efforts. Walmart, for retailer's loyal customer base and, (5) how does online
example, integrated standard delivery services from its grocery penetration in a market moderate the impact of
online site, Walmart.com with same-day delivery ser- channel integration on online retailer/platform sales?
vices, initially delivered through a separate channel, Wal- To address our research questions, we develop a theo-
mart Online Grocery. In 2020, Walmart combined the retical framework rooted in channel capabilities theory
services from the two platforms, creating a unified cus- (Avery et al., 2012). It posits that sales channels
tomer experience catering to both same-day and standard (e.g., physical store channels, online channels, delivery
delivery requirements. Similarly, Amazon.com and its platform channels) have capabilities that enable cus-
grocery subsidiary, Whole Foods Market, transitioned tomers to achieve their shopping goals. If a new channel's
from operating separate delivery channels to an inte- capabilities are duplicative of an existing channel, its
grated online platform. introduction may cannibalize the sales of the existing
We utilize this two-stage integration between the channel; but if the two channels have complementary
retailer, Target, and its platform, Shipt, to address the fol- capabilities, combined channel sales will increase. In our
lowing research questions: (1) How does PDA impact context, the retailer's online channel (Target.com) and
sales of the retailer's online channel and sales of the the platform channel (Shipt) have capabilities to offer
delivery platform's channel? (2) How does IDA impact various delivery services, potentially appealing to diverse
sales of the retailer's online channel and sales of the customer segments. However, the relationship between
delivery platform's channel? (3) What is the impact of these two channels may be complementary or substitu-
integration on total online sales for the retailer and its tive, depending on the degree of integration between the
platform partner? We address these questions to deter- channels (Avery et al., 2012; Wang & Goldfarb, 2017).
mine whether the addition of the delivery platform chan- To test our theories, we analyze propriety customer
nel complements or substitutes for the retailer's online spending data across U.S. markets. We conduct a quasi-
channel.1 experiment on the integration of operations between Tar-
We further assess how characteristics of the retailer get and Shipt using difference-in-difference (DID) models
and of the market may modify the impact of channel with propensity score matching. This approach helps to
integration on retailer online and platform sales (Avery identify the causal impacts on sales performance at two
et al., 2012; Bang et al., 2013; Wang & Goldfarb, 2017). levels of integration, PDA and IDA.
Specifically, we introduce two moderators into our Our study contributes to the channel integration liter-
model: a measure for the retailer's loyal customer base ature in the context of last-mile delivery by providing
and a measure for the penetration of online grocery sales insights into the integration of standard and same-day
in a market prior to integration. Thus, our fourth and delivery services. We respond to the call for theory devel-
fifth research questions are the following: (4) How does a opment in last-mile delivery by formulating a theory on
4 PARK ET AL.

the impact of this channel integration on the perfor- encourages larger customer purchases, retailers should
mance of channel partners (Castillo et al., 2018; Lim work with platforms to ensure that gig workers are ade-
et al., 2018). Furthermore, we respond to the call for quately trained to handle diverse, larger orders and that
middle-range theory development that involves “context- advanced delivery routing algorithms are employed.
specific conceptualization that provides theoretically These algorithms should consider factors such as dis-
grounded insights readily applicable to an empirical con- tance, order size, and unloading times. By focusing on
text” (Carter, 2011). these strategies, retailers can enhance the customer expe-
Previous research has primarily focused on the inte- rience from order to delivery, maximizing the benefits of
gration of offline and online channels (Akturk integration for both the platforms and themselves.
et al., 2018; Fang et al., 2023; Gallino & Moreno, 2014). The rest of this paper proceeds as follows: Section 2
Relative to these papers, our study investigates the inte- proposes a theoretical framework for our analysis.
gration between a retailer's online channel and a delivery Section 3 develops hypotheses. Section 4 details the
platform's online channel in the context of last-mile methods and data used in the analysis. Section 5 presents
delivery. econometric models and estimation results. Section 6 pre-
Our conceptual framework for PDA and IDA departs sents extended analyses. Finally, Section 7 concludes and
from prior conceptualization on retail channel integra- discusses the limitations of the paper.
tion in several ways. First, while studies such as Fang
et al. (2023) and Ren et al. (2023), focus on offline-
to-online integration, our conceptualization of PDA and 2 | THEORETICAL FRAMEWORK
IDA emphasizes online-to-online channel integration
between platforms and retailers. Second, our PDA and This section develops a theoretical framework for our
IDA constructs uniquely draw on channel capabilities research. In Section 2.1, we introduce our overarching
theory. We conceptualize service capabilities in the con- theory, channel capabilities, and discuss how channels
text of last-mile delivery, then use these dimensions to may be complementary or substitutive. In Section 2.2, we
form our PDA and IDA constructs. Third, we distinguish discuss channel capabilities theory in the context of the
between the capabilities of the platform and retailer relationship between a retailer's online channel and a
online channels (Bang et al., 2013). In our context, plat- delivery platform channel. Finally, in Section 2.3 we dis-
forms have strengths in the same day delivery services for cuss the sales benefits from integrating delivery service
time-sensitive products, whereas retailers, through their capabilities into a single channel.
websites, have strengths in offering standard delivery for
non-time-sensitive items. This distinction has not been
addressed in past channel integration studies (Cao & 2.1 | Channel capabilities
Li, 2015; Fang et al., 2023; Oh et al., 2012).
Furthermore, our study provides theoretical contribu- Retailers offer various channels to reach customers,
tions by demonstrating mechanisms leading to the including physical stores, online channels, catalogs,
impact of levels of integration on channel sales (Fisher and mobile channels (Tang et al., 2021; Wang &
et al., 2019; Wang & Goldfarb, 2017). Notably, we find Goldfarb, 2017; Xu et al., 2017). Adding a channel can
that PDA has a positive effect on delivery platform sales increase a retailer's customer base, thus complementing
through increased transaction frequency, while IDA has existing channels, but may also divert customers, thus
positive impacts on both the transaction frequency and acting as a substitute for existing channels (Bang
transaction (basket) size for the retailer's online channel. et al., 2013). Channel capabilities theory, as proposed
Our research reveals critical operational implications, by Avery et al. (2012), offers a theoretical framework to
primarily for retailers, and for delivery platforms as well. determine channel substitutive and complementary
During the PDA stage, as transaction frequencies effects.
increase, it is crucial for retailers to closely collaborate Channel capability is defined as “an enabling char-
with delivery platforms, which manage route planning acteristic of a channel that allows consumers to accom-
and scheduling. Retailers should ensure that their inven- plish their shopping goals” (Avery et al., 2012). For
tory management and product assortment strategies are instance, a physical channel can offer immediate grati-
well-aligned with the operational capabilities of these fication in the purchase of a product, the ability to
platforms. This includes maintaining consistent product touch or smell a product, and access to face-to-face cus-
availability that matches the delivery schedules and pro- tomer support (Alba et al., 1997; Balasubramanian
vides accurate stock data to facilitate efficient order et al., 2005). An online channel, on the other hand,
processing. In the IDA stage, where deeper integration may offer a wider array of products and can save
PARK ET AL. 5

customers travel costs, thus improving shopping conve- Offering a variety of delivery services enhances chan-
nience (Saphores & Xu, 2021). nel capabilities since the appropriate delivery option will
When the characteristics of a new channel subsume vary by product type. Importantly, same-day delivery ser-
those of an existing channel, the new channel may canni- vice may be best suited for refrigerated, frozen, or perish-
balize existing channel sales (Alba et al., 1997; able groceries to shorten the shipping time for cold-chain
Deleersnyder et al., 2002). For instance, an online chan- deliveries and reduce the chance of spoilage (Boyer &
nel may cannibalize a catalog channel, since it provides Hult, 2005). On the other hand, “standard” delivery ser-
similar product information but superior search capabili- vice (two-day, three-day, one week, etc.) may provide a
ties. Conversely, if a new channel has distinct capabilities low-cost option for many non-perishable products when
that appeal to different market segments, then the new short lead times are not important (Sousa et al., 2020).
channel may complement existing channels and have a Delivery platforms in the retail grocery industry spe-
positive impact on a retailer's sales. For instance, an cialize in same-day delivery services for time-sensitive
online channel may complement a physical channel by products, including perishable groceries.2 The platforms
appealing to customers favoring convenience and time generally employ contract personnel to pick, pack, and
savings over instant gratification (Avery et al., 2012). deliver orders (Boyer & Hult, 2005). After receiving an
The channel capabilities theory has been applied to online order for a retailer, basket information is relayed
examine interactions among sales channels. Bang et al. to local contract workers who then pick the order, often
(2013) analyzed the effect of introducing a mobile chan- at the retailer's physical location. Orders are generally
nel on the performance of an online channel accessed by delivered by a second contractor who specializes in the
computer. They found that the impact on sales through delivery function (Lien, 2017). The platform business
the two channels varied by product characteristics. For model is flexible and scalable, with platforms adjusting
instance, the mobile channel may be a poor fit for prod- supply resources (i.e., shoppers and deliverers) to respond
ucts that require in-depth information search, due to the to short-run and long-run changes in demand (Bai
mobile device's small screen and reduced font size. Wang et al., 2019).
and Goldfarb (2017) found complementarities between Although platforms can meet the needs of retailers
physical store channels and online channels, with physi- for same-day delivery service, retailers may reserve
cal stores acting as “living billboards” for retailers, the standard delivery of non-time sensitive products
attracting customers to the retailers' online channels. (e.g., apparel) to their own online operations. Retailers
In the context of our research, a delivery platform can maintain their own distribution centers, and contract
channel may provide distinct advantages to consumers online deliveries of these products to logistics service pro-
compared to a retailer's own online channel. In particu- viders, such as UPS and Fedex. Thus, consumers wishing
lar, the delivery platform may specialize in same-day to purchase both time-sensitive and non-time-sensitive
delivery of time-sensitive products, while the retailer's products may need to access both the website of the
online channel may specialize in the standard delivery of delivery platform and the retailer's website. Conversely,
non-time-sensitive products. Thus, given the distinct to provide a higher level of convenience to consumers,
capabilities of the delivery platform channel from a retailers can integrate a delivery platform's same-day
retailer's own channel, the two channels may be comple- delivery services into their own websites, thus providing
mentary. However, as noted below, this complementarity “one-stop shopping” for consumers. We discuss levels of
may depend on channel integration between the retailer integration, below.
and the delivery platform.

2.3 | Integrating delivery service

2.2 | Delivery service capabilities capabilities

We define delivery service capability as a channel's abil- Retailers can integrate the delivery platform's capabilities
ity to provide a range of delivery options to a retailer's into their own online site to provide channels for the
customers. Delivery service capabilities are increased if ordering and delivery of time-sensitive products, such as
a channel delivers a wider range of products over a groceries, as well as for non-time-sensitive products; for
greater variety of delivery options. Providing a range of example, apparel. This integration can occur in two ways.
delivery services for a variety of products allows a chan- First, customers may be directed from the retailer's
nel to satisfy customer needs by delivering the “right” online site to the delivery platform website to place same-
products at the “right” time (Lim & Srai, 2018; Sousa day delivery orders. Alternatively, the retailers may inte-
et al., 2020). grate the delivery platform's capabilities directly into
6 PARK ET AL.

their own online sites. This allows consumers to place 3.1 | Last-mile delivery platform
same-day and standard delivery orders directly from a integration
retailer's website, providing a one-stop shopping experi-
ence where both time-sensitive and non-time-sensitive The context for our hypotheses development is on
products can be ordered in a single transaction. retailers that sell both time-sensitive and non-
Enabling one-stop shopping through a single website time-sensitive products and delivery platforms that pro-
presents several benefits for customers. First, customers vide same-day delivery on behalf of retail partners.
can conveniently access and review their purchase his- Unlike past integration studies (e.g., Fang et al., 2023;
tory, making repeat purchases and the generation of Ren et al., 2023), we focus on online-to-online channel
shopping lists more efficient (Melis et al., 2016). Second, integration. Our study examines integration between the
customers can benefit from order aggregation, allowing retailer's online channel and the delivery platform chan-
them to more easily meet minimum purchase quantities nel in two stages. Table 1 outlines the delivery service
for free delivery. Without integration, customers may be capabilities of each channel at the different stages of
compelled to make additional purchases to meet free integration.
shipping minimums (Ku, 2019). Third, customers can In our base case (stage 0), the retailer's online channel
enjoy loyalty program benefits and take advantage of can be used for ordering non-time-sensitive products
retailer sales promotions for both same-day and standard with a standard delivery option. The retailer offers no
orders (Oh et al., 2012). For instance, Target's loyal cus- option for same-day delivery.
tomers can apply a 5% discount with their Target Red The first stage of integration, PDA, is conceptualized
card to complete purchase orders. as online-to-online channel integration between the
In summary, delivery channels have distinct delivery platform and the retailer's online channel (see
strengths that appeal to various customer segments. Table 1). Non-time-sensitive products are ordered with
The delivery platform channel can excel in the same- standard delivery through the retailer's online website,
day delivery of time-sensitive products, whereas the while customers can order time-sensitive items with
retailer's own online channel may offer superior capa- same-day delivery through the delivery platform. A cus-
bilities in the standard delivery of non-time-sensitive tomer visiting the retailer's website is directed to the
products. Integrating these capabilities within the delivery platform to place orders requiring same-day
retailer's own online channel provides additional delivery. Thus, at this stage of integration, customers
advantages to the retailer's customer base in terms of wishing to purchase both same-day and standard delivery
convenience and cost savings. orders from the retailer, must visit both the retailer's
online site and the delivery platform's site.
The second stage of integration, IDA, is conceptual-
3 | HYPOTHESES DEVELOPMENT ized as a higher level of online-to-online channel integra-
tion between the delivery platform channel and the
Based on delivery channel capabilities conceptualized in retailer's online channel. IDA offers customers the flexi-
the previous section, we hypothesize two stages of last- bility to order both same-day deliveries and standard
mile delivery integration. Section 3.1 discusses the two deliveries through the retailer's online channel. Same-day
stages of integration. Our hypotheses are then developed orders can also be placed through the delivery platform
and presented in the following subsections. website. This integrated approach allows the customer to

TABLE 1 Delivery platform integration and delivery channel capabilities.

Scope of delivery services Degree of service integration

Standard Same-day Non-integrated Integrated

Integration stage Sales channel delivery delivery delivery delivery
Stage 0 Retailer's online ✓ - - -
Base case Delivery platform - - - -
Stage 1: PDA Retailer's online ✓ - ✓ -
Platform delivery access Delivery platform - ✓ ✓ -
Stage 2: IDA Retailer's online ✓ ✓ - ✓
Integrated delivery access Delivery platform - ✓ ✓ -
PARK ET AL. 7

bundle orders for time-sensitive products, such as perish- customers may access both same-day and standard deliv-
ables, with those for non-time-sensitive items. ery services through the retailer's online channel.
Although same-day deliveries can also be arranged
through the platform's website, this service will no longer
3.2 | Platform delivery access be unique to the platform. Thus, the online retailer's
channel can substitute for the delivery platform channel
Under PDA, the retailer adds same-day delivery service (Wang & Goldfarb, 2017). Since customers have options
through its delivery platform partner to complement its for both same-day and standard deliveries through the
own standard delivery service available through its online retailer's online website, this convenience may be attrac-
channel. The retailer's customers thus benefit from this tive to customers (Ku, 2019; Oh et al., 2012). Thus, we
additional offering (Tucker & Zhang, 2010; Wang propose that IDA will divert customers from the delivery
et al., 2022). Since same-day delivery from the delivery platform channel to the retailer's online channel (Bang
platform is an added service capability, complementary et al., 2013) and pose the following hypotheses:
to the retailer's own standard delivery service, PDA is
likely to attract new customer demand (Bang Hypothesis 2a. Integrated delivery access
et al., 2013). Given that customers need to access same- (IDA) leads to lower sales for the delivery
day delivery through the delivery platform website, this platform channel.
service is projected to increase sales at the delivery plat-
form (Neslin et al., 2006). The delivery platform will be Hypothesis 2b. Integrated delivery access
offering products from a new retail partner that provides (IDA) leads to higher sales for the retailer's
a distinct service to the retailer's customers (i.e., same- online channel.
day delivery). Thus, our first hypothesis is as follows:

Hypothesis 1a. Platform delivery access 3.4 | Moderating effects

(PDA) leads to higher sales for the delivery
platform channel. In this section, we propose potential mechanisms that
may contribute to the impact of PDA and IDA on chan-
At the PDA stage, the delivery platform channel may nel sales: loyal customer base and online grocery market
substitute, at least partially, for the retailer's online chan- penetration.
nel since certain products may be available through both
channels (e.g., household essentials such as paper towels
or cleaning supplies), although with different delivery 3.4.1 | Loyal customer base
options.3 However, other products will be unique to the
platform website (e.g., perishable products) or to Prior studies have emphasized the importance of retailer
the online retailer's website (e.g., small kitchen appli- loyalty programs to attracting and maintaining customers
ances). Thus, customers from the delivery platform chan- (Cao & Li, 2015; Lemon & Verhoef, 2016). In this context,
nel may spill over to the retailer's online channel to make we describe how a retailer's loyal customer base may
purchases only available through that channel (Avery moderate the impact of channel integration on retailer/
et al., 2012; Lemon & Verhoef, 2016; Soysal & platform online sales.
Krishnamurthi, 2016). Although there may be both sub- We expect that the moderating influence of a retailer's
stitute and spillover impacts between the two channels, loyal customer base may vary across the different stages
the spillover effect may dominate, as proposed by Avery of integration. In particular, at the PDA stage, we expect
et al. (2012). Thus, we propose: that a retailer's loyal customer base may have a greater
moderating impact on platform sales, than on retailer
Hypothesis 1b. Platform delivery access online sales. The reason for this distinction is that, while
(PDA) leads to higher sales for the retailer's the retailer maintains standard delivery service for non-
online channel. time-sensitive products through its own website, it adds
new same-day delivery service for time-sensitive products
on the platform website. During the PDA stage, cus-
3.3 | Integrated delivery access tomers are redirected from the retailer's website to the
delivery platform to place same-day delivery orders.
Whereas PDA directs a retailer's online customers to the Thus, the retailer's loyal customers still enjoy the usual
platform's website to use same-day delivery, with IDA, benefits through its online channel, such as product
8 PARK ET AL.

discounts and promotions, when placing non- sales relative to overall online delivery sales (i.e., overall
time-sensitive orders (Kumar & Shah, 2004). Addition- online delivery sales equal non-grocery delivery sales plus
ally, these loyal customers may be more likely to try the grocery delivery sales) in a market, prior to PDA. A high
retailer's new same-day delivery service offered by the OGI value reflects room for market penetration for online
delivery platform (Bijmolt et al., 2010; Rioux, 2020). grocery purchases (Han et al., 2013) while a low OGI
Therefore, a strong loyal customer base could enhance indicates that a sizable percent of consumers may have
the effectiveness of PDA in boosting delivery platform already embraced online grocery purchases (Cao &
sales. Li, 2015). Thus, the potential sales growth in markets
Hence, we propose the following hypothesis: with high OGI for new same-day delivery services may
be the greatest (Eisenhardt, 2013; Pauwels et al., 2013).
Hypothesis 3a. A strong loyalty base for a First, at the PDA stage, OGI may exert a more pro-
retailer in a market amplifies the positive nounced influence on the PDA-platform sales relation-
effects of platform delivery access (PDA) on ship, rather than on the PDA-retailer online sales
the delivery platform channel sales. relationship since, at this stage, the retailer does not sell
perishable groceries through its own website. PDA offers
At the IDA stage, the influence of a retailer's loyalty same-day delivery through the platform channel and this
customers is expected to be more pronounced on the service is likely to be used by customers purchasing gro-
IDA-retailer online sales relationship than on the IDA- cery products (Boyer & Hult, 2006; Sousa et al., 2020).
platform sales relationship. This distinction primarily Hence, in markets where grocery delivery is relatively
stems from the retailer's strategy at this stage, which immature prior to PDA, the growth in delivery platform
involves implementing same-day delivery services channel sales due to PDA may be stronger (Mascarenhas
through its own website. With IDA, customers can con- et al., 2002). Thus, we predict that OGI will positively
veniently purchase both time-sensitive and non-time sen- moderate the impact of PDA on delivery platform chan-
sitive products through the retailer's website, eliminating nel sales and hypothesize the following:
the need for customers to visit the platform website for
same-day deliveries (Cao & Li, 2015; Oh et al., 2012). This Hypothesis 4a. A high level of online gro-
integrated purchasing service through the retailer's cery immaturity in a market amplifies the
online channel may be particularly beneficial to the positive effects of platform delivery access
retailer's loyalty customers. These customers can receive (PDA) on the delivery platform channel sales.
the greatest benefits from this integrated service as their
combined purchases will be eligible for loyalty rewards Next, at the IDA stage, it is anticipated that OGI
offered by the retailer (Stathopoulou & Balabanis, 2016; may have a more substantial impact on the IDA-retailer
Vesel & Zabkar, 2009). online sales relationship, since at this stage, it is the
Thus, we posit the following hypothesis: retailer that expands its grocery offerings with the provi-
sion of same-day delivery services. IDA enhances the
Hypothesis 3b. A strong loyalty base for a retailer's online channel by providing an integrated
retailer in a market amplifies the positive delivery service for both time-sensitive products, such as
effects of integrated delivery access (IDA) on groceries, and non-time-sensitive items (Ku, 2019; Melis
the retailer's online channel sales. et al., 2016; Oh et al., 2012), bolstering the channel's
capability to meet diverse consumer demands. Com-
pared to the pre-IDA period, the retailer's online chan-
3.4.2 | Online grocery penetration nel is better equipped to fulfill the demand for groceries
online, potentially leading to a surge in grocery pur-
Markets that already have significant online grocery chases. This surge may be greatest in markets that had
shopping prior to the integration efforts of the retailer relatively low online grocery sales prior to integration
and the delivery platform may be less impacted by inte- efforts (Pauwels et al., 2013). Thus, we propose that the
gration efforts since online shoppers are already estab- relationship between IDA and retailer online channel
lished in their purchase preferences, especially for same- sales will be greater in markets characterized by high
day delivery services. Conversely, “immature” markets, OGI as follows:
with lower online grocery penetration prior to the inte-
gration efforts may experience the greatest impacts from Hypothesis 4b. A high level of online gro-
the integration efforts. cery immaturity in a market amplifies the
Following Han et al. (2013), we define online grocery positive effects of integrated delivery access
immaturity (OGI) as the intensity of non-grocery delivery (IDA) on the retailer's online channel sales.
PARK ET AL. 9

F I G U R E 2 (a) Research model

(PDA). (b) Research model (IDA). IDA,
integrated delivery access; PDA,
platform delivery access.

Figure 2a,b depict the research model and our stores across the U.S., ranking as the 33rd largest corpo-
hypotheses. ration on the Fortune 500 list (published in 2023) in
terms of revenue. It also operated an online channel,
Target.com. In the second quarter of 2023, Target.com
4 | M E T H O D A N D DA T A delivered 16.9% of revenue for Target (Haleem, 2023).
On December 13, 2017, Target acquired Shipt, a
This section describes our research methods and the technology-based delivery platform, to develop same-day
data used in our estimations. In Section 4.1, we explain delivery capabilities. As of July 21, 2021, Shipt operated
the research context. Section 4.2 describes our data, while in over 5000 cities in the U.S., employing independent
Section 4.3 discusses our estimation method. shoppers to pick products at retail stores and then deliver
these products to customers. In 2023, its website listed
16 retail partners, including the supermarket chains, Kro-
4.1 | Research context ger and Lidl, and the warehouse club retailer, Costco.4
Prior to the first stage of integration, Target only
We analyze our model focusing on integration between offered standard delivery for non-time-sensitive products
the online channel of a national U.S. retailer, Target, and through its online channel. No options were provided for
a delivery platform, Shipt. Integration between these two same-day delivery. Implementation of the first stage inte-
companies occurred in two steps, PDA first and then gration, PDA, for same-day delivery of Target's products
IDA. We perform quasi-experiments on the two levels of through the Shipt website, began 2 months after the
integration to assess their impact on sales at the retailer merger (February 2018). This same-day delivery channel
and the delivery platform. was initially launched in a small number of cities, with a
Target is a Minneapolis-based general merchandise growing number of markets eligible for Shipt's same-day
and grocery retailer. As of 2023, Target operated 1958 delivery of Target products included over time. With the
10 PARK ET AL.

PDA level of integration, users had to visit the Shipt web- the impacts of PDA and IDA from other factors that may
site to order same-day deliveries of Target products. It impact Target.com's and Shipt's sales in a city market.
was not until June 2019 that IDA between the two firms Given that there are two treatments, PDA and IDA,
was implemented. After IDA, same-day delivery service and two dependent variables, Target.com sales and Shipt
was deployed through both Shipt's and Target's online sales in a city market, four estimations are conducted to
channels. measure the following: (1) impact of PDA on Target.com
sales; (2) impact of PDA on Shipt sales; (3) impact of IDA
on Target.com sales; and (4) impact of IDA on Shipt
4.2 | Data description sales.
To isolate the impact of the treatments (PDA and
We use a proprietary consumer spend dataset compiled IDA) on the dependent variables (Target.com sales
by Earnest Analytics (https://www.earnestanalytics.com). and Shipt sales in a market), each of the four DID estima-
Partnering with a U.S. commercial bank, the company tions requires an untreated or control group. Each of the
tracks credit card, debit card, and store-card transactions four control groups must resemble the treated group for
from over 6 million U.S. consumers. From this sample, each of the four estimations. Propensity score matching
Earnest retains the subset of consumers showing consis- procedures, as detailed in Appendix C, are used to iden-
tent behavior in their card spending. For instance, con- tify the four sets of control groups. In addition,
sumers are dropped from the sample if they halt their Appendix B presents the results of parallel trend tests.
credit card purchases. Ultimately, the sample represents
about 0.5% of households for each U.S. state.
Earnest Analytics reports aggregate consumer spend 4.4 | Sample selection
data at the merchant-city-week level. The dataset covers
major retailers such as Target, Walmart, Costco, and Mei- This section outlines the procedures for sample
jer, as well as delivery platforms such as Shipt, Amazon selection of the treatment and control groups. Following
Fresh, Instacart, and Peapod. Additionally, the dataset the approach of Fisher et al. (2019), we create separate
distinguishes between spending made through physical treatment and control groups for each of our four
and online sales channels. In this study, we use consumer estimations.
spending as an approximation for sales performance in a
city market.
We augment the consumer spending data with addi- 4.4.1 | Experimental design: PDA
tional data sources. First, we collect household demo-
graphics by city market from the Consumer Population Figure 3 depicts the timeline for the implementation of
Survey, available from the U.S. Census Bureau. From this PDA. It was rolled out by Target on a market-by-market
dataset, we retain population, population density, median basis beginning in February 2018 and continuing through
age, median household income, and number of house- the remainder of 2018 and 2019. We restrict our sampling
holds who have access to mobile devices (a measure of timeframe to 21 weeks before and after PDA implemen-
the depth of technology use in a market). These demo- tation.5 This restriction allows us to capture the markets
graphic variables are used as controls when we estimate in which PDA was implemented in 2018 as our treatment
weekly Target.com and Shipt sales in city markets. Fur- markets. The markets in which PDA was not implemen-
thermore, we identify the start dates for PDA and IDA ted until 2019 can then be included in the pool of poten-
using Google Search and the Lexis–Nexis database. tial control markets, subject to propensity score matching
Appendix A elaborates on the keyword search process. with the treatment markets. Since PDA was implemented
in both the treatment markets and (eventually) in the
control markets, the two groups of markets may share
4.3 | Estimation method common characteristics, thus facilitating the propensity
score matching process (Hwang & Park, 2016).
Our study investigates the effects of PDA and IDA on As illustrated in Figure 3, PDA was implemented in
online sales in city markets for Target.com and Shipt. To the treatment group at different times (PDA in market
accomplish this, we conduct a quasi-experimental study. 1, PDA in market 2, … PDA in market n) during the sam-
Specifically, we employ DID estimations with propensity ple period of 2018. Thus, the coding for the PDA treat-
score matching to assess the impacts of the treatment ment not only varies by market but also with time. A
effects from PDA and IDA on the sales of Target.com and market is not coded as “treated” until the treatment has
Shipt in city markets. The use of DID allows us to isolate taken place. As noted above, the control group comprises
PARK ET AL. 11

FIGURE 3 Experiment design: PDA. PDA, platform delivery access.

TABLE 2 Sample selection: PDA.

Sample sizes

Dependent Before After

Treatment Design variable Sample units Sample groups(sample year: 2018) matching matching
Platform delivery 1 Platform sales City Treatment Shipt 171 162
access Markets with PDA
(PDA) Control Shipt 447 162
without PDA
2 Retailer online City Treatment Target.com 171 162
sales Markets with PDA
Control Target.com 447 162
without PDA

Abbreviation: PDA, platform delivery access.

the markets where PDA was implemented after the sam- 162 treatment-control paired city markets, with the
ple period of 2018; that is, during 2019. remaining treatment markets dropped because no close
Table 2 presents information on the treatment and match in the control group was found.
control groups utilized for PDA. Our dataset includes To verify that our control group markets (after pro-
information on 8008 city markets. However, we restrict pensity score matching) resemble our corresponding
our observations to city markets in which both Target. treatment group markets, we test for parallel trends prior
com and Shipt operate. Moreover, we retain only those to treatment (Li & Netessine, 2020). This test requires the
markets with at least one non-zero observation control group to demonstrate similar outcome trends to
(i.e., positive sales transaction) for both Target.com and the treatment group in the periods prior to treatment.
Shipt during every five-week rolling period in the sample Results of the parallel trend analysis are shown in
period for the treatment group and in 2019 for the control Tables B2 and B3 of Appendix B and indicate that there
group (Lim et al., 2021). This leaves us with 171 city mar- are no significant differences in sales trends between the
kets in the treatment group and 447 potential city treatment and control groups prior to the treatment. Fur-
markets in the control group. thermore, we introduced a “pseudo” treatment for week
We then match treatment markets to control mar- 11 of each sample year and conducted a falsification test
kets. The details of the propensity score matching tech- for the pre-treatment sample. Results are presented in
nique are outlined in Appendix C. As illustrated in Table B1 in Appendix B, showing no treatment effect for
Table 2, the one-to-one matching (i.e., each treatment the pseudo treatment. Appendix C details the process of
market is matched to one control market) results in propensity score matching.
12 PARK ET AL.

4.4.2 | Experimental design: IDA on June 13, 2019. To estimate the effect of the IDA treat-
ment on Shipt's sales in a city market, we match Shipt's
Figure 4 illustrates the timeline for the IDA treatment. city markets with city markets in which a second online
To estimate the effect of the IDA treatment on Target. platform, Amazon Fresh, operates. Similar to Shipt, Ama-
com and Shipt sales, we require different treatment and zon Fresh specializes in the same-day delivery of perish-
control groups from those used in the PDA analysis. able products, such as groceries.6 Moreover, Amazon
There are two reasons why new groups are required: Fresh had PDA-level integration with a retailer (Whole
First, unlike PDA, IDA was introduced simultaneously Foods Market) but did not implement IDA-level integra-
across all of Target's U.S. city markets. Thus, with IDA, tion with that retailer until after our sample period.
we cannot use a control group from Target and Shipt Finally, as noted below, the sales trends of Shipt and
consisting of untreated, but soon-to-be-treated markets Amazon Fresh were parallel prior to IDA.7
(since all markets are treated at the same time). More- To estimate the impact of IDA on Shipt's sales, we
over, our IDA treatment group is larger than our PDA focus on city markets in which Shipt operated (2564 city
treatment group since it consists of all treated markets markets) and Amazon Fresh operated (2529 city mar-
(unlike PDA where only those markets treated in 2018 kets). We retain city markets with at least one non-zero
are in the treatment group). sales observation for Shipt (Amazon Fresh) during every
Table 3 presents information on the treatment and five-week rolling period in the sample period, following
control groups utilized for IDA. Target implemented IDA Lim et al. (2021). Furthermore, we retain the markets in

FIGURE 4 Experiment design: IDA. IDA, integrated delivery access.

TABLE 3 Sample selection: IDA.

Sample sizes

Dependent Before After

Treatment Design variable Sample units Sample groups (sample year: 2019) matching matching
Integrated 3 Platform sales City markets Treatment Shipt 557 346
delivery access with IDA
(IDA) Control Amazon Fresh 386 346
without IDA
4 Retailer online City Treatment Target.com 557 557
sales Markets with IDA
Control Walmart.com 748 557
without IDA

Abbreviation: IDA, integrated delivery access.

PARK ET AL. 13

which both by Shipt and Amazon Fresh operate enabling lnSALESjt ¼ β0 þ β1  TREATj þ β2  AFTERjt þ β3 ð1Þ
a fairer comparison between the two firms. We then per-
TREATj  AFTERjt þ γ j þ νt þ εjt ,
form one-to-one propensity score matching, resulting in
346 treatment-control city market pairs.8
To estimate the effect of IDA on Target.com's sales in where the dependent variable, SALESjt, denotes the sales
city markets, we use Walmart.com as the control. During of Shipt (Target.com) in market j at time t. TREATj is a
the sample period (similar to Target.com) Walmart.com dummy variable equal to 1 if a market is in the PDA
used standard delivery methods to fulfill orders for non- treatment group and 0 if the market is in the PDA control
perishable products.9 Moreover, Walmart.com did not group. AFTERjt is a dummy variable with value 1 if weekt
provide integrated fulfillment for both standard and is for an observation in the time period after PDA was
same-day deliveries until after the conclusion of the sam- implemented. The coefficient, β3, for the interaction
ple period. effect between TREATj and AFTERjt, is our DID variable
To form our treatment and control groups, we focus assessing the impact of PDA in the treatment markets.10
on city markets in which Target.com operated (8008 city We use a fixed-effects regressions to account for unob-
markets) and in which Walmart.com operated (12,717 served heterogeneity among markets and over time. γ j
city markets). We retain city markets with at least one measures the city market fixed effects, νt represents the
non-zero sales observation for Target.com (Walmart. weekly time fixed effects, and εjt is the error term.
com) during every five-week rolling period in the sample Using similar approaches and notation, we also esti-
timeframe, following Lim et al. (2021). Similar to the mate the impact of IDA on platform channel (retailer
delivery platform sample, we keep the markets in which online channel) sales:
both Walmart.com and Target.com operated. We per-
formed one-to-one propensity score matching, resulting lnSALESijt ¼ β0 þ β1  TREATi þ β2  AFTERt þ β3 ð2Þ
in 557 treatment-control city market pairs.
TREATi  AFTERt þ ηi þ γ j þ νt
A crucial assumption for DID models is the existence
of parallel trends between the treatment and control þ εijt ,
groups prior to the treatment effect (Fisher et al., 2019).
Therefore, to validate our IDA control groups, we test for in Equation (2), subscript i is added to denote a specific
parallel trends in city market sales between the treatment delivery platform or retailer.11
and control groups prior to the treatment effect (Li &
Netessine, 2020). The results of the parallel trend analysis
are provided in Tables B2 and B3 in Appendix B, suggest- 5.1.2 | Moderating-effect models: PDA
ing no significant differences in parallel trends, thus pro- and IDA
viding support for the use of the control groups.
Descriptive statistics for matched samples are provided in To assess the potential moderating impacts of a loyal
Appendix D. retailer customer base and online grocery immaturity on
the relationships between PDA and delivery platform
channel (retailer online channel) sales, we also estimate
5 | ESTIMAT I ON AN D RES U LT S the following model12:

5.1 | Estimating models ln SALESjt ¼ β0 þ β1  TREATj þ β2  AFTERt ð3Þ

þ β3  TREATj  AFTERt þ β4
We conduct four estimations, to analyze the impact of the
two levels of integration (PDA and IDA) on the retailer  TREATj  AFTERt  LOYALj þ β5
(Target.com) and the delivery platform (Shipt). Based on  TREATj  AFTERt  IMMATUREj
the matched sample datasets, we estimate the effects of
þ γ j þ νt þ εjt ,
PDA and IDA on retailer online channel and platform
channel sales using the following DID equations.
where SALESjt represents sales of delivery platform
(retailer) in market j at time t, LOYALj denotes the loy-
5.1.1 | Direct-effect models: PDA And IDA alty customer base of Target in market j, and other nota-
tion follows from Equation (1). To measure Target's loyal
The effect of PDA on platform channel (retailer online customer base in a market, we use the ratio of average
channel) sales is estimated by the following: weekly Target.com sales made using Target loyalty cards
14 PARK ET AL.

to average (total) weekly Target.com sales in the market Figure 6 presents the sales indices for both the deliv-
during the 6-week period prior to the PDA sample ery platform channel (left panel) and the retailer online
period.13 The loyalty moderating effect is assessed channel (right panel) for IDA implementation. The figure
through the coefficient, β4. suggests no discernable divergences between the treat-
IMMATUREj denotes the online grocery immaturity ment and control groups for the delivery platforms after
in market j. We approximate online grocery IDA implementation. However, the right panel indicates
immaturity by the ratio of online non-grocery delivery a potential small, but significant divergence between the
sales in a market to total online delivery sales (grocery treatment and control groups following IDA for retailer
+ non-grocery) during the 6-week period prior to the online sales.
PDA sample period.14 The online grocery immaturity
moderating effect is measured through the coefficient, β5.
Similar to the equation for PDA, we estimate the 5.2.2 | Unique customers
moderating effects of a loyal customer base and online
grocery immaturity on the relationship between IDA and We also present a customer trend analysis from a comple-
platform channel (retailer online channel) sales as mentary dataset obtained from Earnest Analytics. This
follows15: dataset provides customer counts of shoppers that buy
exclusively through a sales channel or cross shop
ln SALESijt ¼ β0 þ β1  TREATi þ β2  AFTERt ð4Þ between two sales channels.17 The data are segmented
into three-month rolling periods spanning 2018 to 2019.
þ β3  TREATi  AFTERt þ β4
For instance, period 1 covers January, February, March
 TREATi  AFTERt  LOYALj þ β5 2018, while period 2 covers February, March, and April
 TREATi  AFTERt  IMMATUREj 2018. The entire sample period, 2018 to 2019, is consis-
tent with the time period from our main analysis.
þ ηi þ γ j þ νt þ εijt :
Figure 7 shows the number of unique customers that
shopped at the delivery platform Shipt and the number of
The variables are computed similar to the computa- unique customers that cross-shopped at Shipt and Target
tions for the variables in Equation (3). and Shipt and Walmart. Increasing numbers of cross-
shoppers between Shipt and Target (compared to Shipt
and Walmart) may provide an indication of synergies
5.2 | Model-free results from the Target-Shipt relationship.
The red vertical line at month = 2 signifies the begin-
5.2.1 | Sales trend ning of PDA (note it was implemented in stages across
the various city markets), while the blue vertical line at
We provide evidence, without employing our models, to Month = 18 signifies the implementation of IDA (across
show how retailer online sales and delivery platform all markets). The figure shows a significant rise in the
sales were impacted by PDA and IDA. We present nor- unique customers using the delivery platform, Shipt, dur-
malized sales, which are obtained by dividing average ing the entire time period, indicating the potential value
sales in a week by the sales at the beginning of the sam- of the Target-Shipt partnership to the delivery platform.
ple period (the average sales at “Event Time—21”) so More significantly, the figure also shows that cross shop-
that sales trends for the treatment and control groups can pers between Target and Shipt increased at a greater rate
be easily compared (Fisher et al., 2019).16 than cross-shoppers between Walmart and Shipt follow-
Figure 5 shows the indices for the delivery platform ing PDA, indicating the potential value of the partnership
channel (left panel) and the retailer's online channel to Target and Shipt. Moreover, there appears to be a sec-
(right panel) for PDA introduction. The vertical line at ond divergence after IDA, potentially indicating the value
week “t” signifies the beginning of the post-treatment to Target and Shipt of the further integration between the
period. The figure illustrates the seasonal adjusted sales two firms.
index by channel (retailer online and delivery platform)
and group (treatment and control). The figure shows a
significant rise in the delivery platform sales index for the 5.3 | Estimation results
treatment group, while the control group index remains
largely unchanged. Significant divergences between the Table 4 presents the estimation results for the PDA
treated and non-treated indices for the retailer online models. Table 5 presents the estimation results for the
channel are not evident from the figure. IDA models.
PARK ET AL. 15

FIGURE 5 PDA: normalized sales by treatment and control group. PDA, platform delivery access.

FIGURE 6 IDA: normalized sales by treatment and control group.

FIGURE 7 Trends in unique

customers.
16 PARK ET AL.

TABLE 4 PDA: Sales performance.

Delivery platform channel Retailer online channel

DV: ln_SALES Model 1.1 Model 1.2 Model 1.3 Model 1.4
AFTER 0.176 (0.133) 0.179 (0.137) 0.155 (0.145) 0.155** (0.072)
TREAT * AFTER 1.108*** (0.215) 7.221*** (2.480) 0.146 (0.162) 0.984 (0.605)
TREAT * AFTER * LOYAL 3.556 (3.329) 0.163 (0.531)
TREAT * AFTER * IMMATURE 9.605*** (2.910) 0.914 (0.681)
Number of observations 14,580 14,580 14,580 14,580
Market fixed effects Yes Yes Yes Yes
Weekly fixed effects Yes Yes Yes Yes
2
R 0.589 0.594 0.592 0.592

Note: Robust standard errors in parentheses. The coefficient for TREAT is omitted because it is absorbed into the market fixed effect, γ j.
Abbreviation: PDA, platform delivery access.
*p < 0.10; **p < 0.05; ***p < 0.01.

TABLE 5 IDA: Sales performance.

Delivery platform channel Retailer online channel

DV: ln_SALES Model 2.1 Model 2.2 Model 2.3 Model 2.4
TREAT * AFTER 0.003 (0.046) 1.290 (0.794) 0.077*** (0.016) 2.507*** (0.139)
TREAT * AFTER * LOYAL 0.626 (0.442) 1.855*** (0.123)
TREAT * AFTER * IMMATURE 1.433 (0.895) 2.725*** (0.161)
Number of observations 29,756 29,756 47,902 47,902
Retailer (Platform) fixed effects Yes Yes Yes Yes
Market fixed effects Yes Yes Yes Yes
Weekly fixed effects Yes Yes Yes Yes
2
R 0.555 0.560 0.706 0.710

Note: Robust standard errors in parentheses. The coefficient for TREAT is omitted because it is absorbed into the retailer (platform) fixed effect, ηi. The
coefficient for AFTER is omitted because it is absorbed into weekly fixed effects, νt, as AFTER does not vary within the control group.
Abbreviation: IDA, integrated delivery access.
*p < 0.10; **p < 0.05; ***p < 0.01.

5.3.1 | Estimation results: PDA we find support for Hypothesis 1a. Hypothesis 1b states
that PDA leads to an increase in retailer online channel
We assess four hypotheses related to PDA, sales. In Model 1.3, the coefficient for TREAT * AFTER
Hypotheses 1a, 1b, 3a, and 4a, in Table 4. The direct is insignificant (β = 0.146, p > 0.1); hence PDA has
effects of PDA on delivery platform channel (Shipt) and no significant impact on Target.com's sales. Thus,
retailer online channel (Target.com) sales are presented Hypothesis 1b is not supported.
in Models 1.1 and 1.3, respectively. The moderating Hypotheses 3a and 4a predict that the positive
effects of loyal customer base and online grocery immatu- effect of PDA on delivery platform sales will be
rity are illustrated in Models 1.2 and 1.4.1819 greater for markets with a stronger loyal customer
Hypothesis 1a predicts that PDA leads to an increase base and greater online grocery immaturity. Model 1.2
in delivery platform channel sales. The results in Model includes the moderating effects of a loyal customer
1.1 indicate that TREAT * AFTER is significant and pos- base and online grocery immaturity on the impact of
itive (β = 1.108, p < 0.01) in the estimation of delivery PDA on Shipt sales. The coefficient for TREAT *
platform sales. This finding implies that PDA increases AFTER * LOYAL is insignificant for Shipt sales
Shipt sales by 202.8% (= exp. (1.108)—1) in a market (β = 3.556, p > 0.1). Thus, Hypothesis 3a is not sup-
relative to expected sales prior to the treatment period ported. The coefficient for TREAT * AFTER * IMMA-
(Halvorsen & Palmquist, 1980; Lim et al., 2021).20 Thus, TURE is positive and significant for Shipt sales
PARK ET AL. 17

(β = 9.605, p < 0.01). Thus, we find support for Table E1 in Appendix E reports the estimation results.
Hypothesis 4a. Results are consistent with the main estimation results.
Second, to test how our estimation results may be
affected by the choice of our propensity score matching
5.3.2 | Estimation results: IDA algorithm, we reconstruct our control groups using a
matching with replacement algorithm. With this algo-
We report the estimation results for the IDA models, rithm, a control retailer (platform) market is allowed to
Hypotheses 2a, 2b, 3b, and 4b, in Table 5. Model 2.1 and be matched to more than one treatment market, thereby,
Model 2.3 test the direct effects of IDA on delivery plat- potentially, providing closer matches to the treatment
form (Shipt) and retailer online (Target.com) sales, markets. Tables E2 and E3 in Appendix E report the esti-
respectively. The moderating effects of loyal customer mation results. Results are consistent with the main
base and online grocery immaturity are illustrated in estimation results.
Models 2.2 and 2.4.2122 Third, we test our models with stricter matching cri-
Hypothesis 2a predicts that IDA leads to a decrease in teria, using a caliper size of 0.10 times standard deviation
delivery platform sales. In Model 2.1, our estimation of propensity scores. The smaller caliper size ensures
results show that the coefficient for TREAT * AFTER is tighter matches between treatment and control markets
insignificant (0.003, p > 0.1). Hence, Hypothesis 2a but may result in fewer matched pairs. Tables E4 and E5
is not supported. Hypothesis 2b states that IDA leads to in Appendix E report the estimation results. Results are
higher sales for the retailer's online channel. In Model consistent with the main estimation results.
2.3, the coefficient for TREAT * AFTER is significantly Fourth, we re-estimate our platform model for IDA
positive (β = 0.077, p < 0.01). This result indicates that with a new control group. Amazon Prime Now and Pea-
IDA leads to increased Target.com sales by 8% (= exp. pod are used for this alternative control group. These
(0.077)—1) relative to expected sales prior to IDA. Hence, platforms showed similar (parallel) trends in sales to our
Hypothesis 2b is supported. treatment group prior to the treatment period (Li &
Hypothesis 3b and 4b predict that the positive effect Netessine, 2020). Table E6 in Appendix E reports the esti-
of IDA on the retailer's online channel sales will be mation results based on this different control group.
greater for markets with a larger proportion of loyal cus- Results are consistent with the main estimation results.
tomers and a greater online grocery immaturity. Model Fifth, it is possible that overall delivery platform activ-
2.4 includes the moderating effects of loyal customer base ity in a market may have driven platform and retail
and online grocery immaturity on the impact of IDA on online sales over time, regardless of the level of delivery
Target.com sales. In Model 2.4, the coefficient for TREAT platform integration. To account for this possibility, we
* AFTER * LOYAL is positive and significant (β = 1.855, add a measure for local platform activity as a control vari-
p < 0.01). Thus, we find support for Hypothesis 3b. The able in our models, PLAT_ACT. The variable is calcu-
coefficient for TREAT * AFTER * IMMATURE is positive lated by first aggregating delivery platform sales across
and significant for Target.com sales (β = 2.725, p < 0.01). major delivery platforms (Amazon Fresh, Amazon Prime
Thus, Hypothesis 4b is supported. Now, Fresh Direct, Instacart, Peapod, Shipt, Thrive Mar-
ket, Walmart Online Grocery) in a market and then
dividing this total by population size of a market. As
5.3.3 | Estimation results: Summary shown in Tables E7 and E8 in Appendix E, the inclusion
of this variable does not affect our main findings.
Figure 8a,b illustrate the summary of our findings. We Lastly, note that our control retailer for IDA,
find support for Hypothesis 1a and 4a in the analysis of Walmart.com, competes with Target.com. Thus, the IDA
PDA and for Hypothesis 2b, Hypothesis 3b, and treatment could impact both the treated firm, Target.
Hypothesis 4b in the analysis of IDA. com, and the control firm, Walmart.com. In particular,
the treatment may have a positive impact on Target.
com's sales and a negative impact on Walmart.com's
5.4 | Robustness checks and alternative sales. Therefore, significant differences between the treat-
explanations ment and control groups following the treatment may be
due not just to a positive enhancement of Target.com's
We illustrate robustness checks in Appendix E. First, to sales (as inferred above) but also to a depression in
account for correlation between retailer online and deliv- Walmart.com's sales. Therefore, the positive effect on
ery platform sales, we re-estimate our model using a Target.com due to the treatment claimed above may be
seemingly unrelated regression for the PDA sample.23 an overestimation to the extent that Walmart.com is
18 PARK ET AL.

F I G U R E 8 (a) Summary of findings

(PDA). (b) Summary of findings (IDA).
Hypotheses with support are bold-faced.
IDA, integrated delivery access; PDA,
platform delivery access.

negatively affected by IDA. However, using year- Decomposing sales into transaction frequency and
over-year growth rates by week level as observations, we transaction size allows us to better determine how PDA
find that IDA between Shipt and Target.com did not sig- and IDA impact retail online and delivery platform sales
nificantly impact Walmart.com's sales. Table E9 reports (Ansari et al., 2008). Specifically, a greater transaction
estimation results. size may signify customers' ability to aggregate orders
across different product categories. Thus, we re-estimate
our econometric models, Equations (1) and (2), with
6 | E X T E N D E D AN A L Y S I S transaction frequency and transaction size as the depen-
dent variables.
In this extended analysis, we identify factors that may Tables F1 and F2 in Appendix F present our estima-
contribute to our understanding of our results. We pre- tion results for PDA. First, in Models F1.1 and F1.2, the
sent the estimation results for the extended analysis in dependent variable is the log of transaction frequency
Appendix F. (ln_TRANS_FREQ) for Shipt and Target.com (respec-
tively). Second, in Models F2.1 and F2.2, the dependent
variable is the log of transaction size (ln_TRANS_SIZE)
6.1 | Transaction frequency and size for Shipt and Target.com (respectively). Similarly, in
Tables F3 and F4, Models F3.1, F3.2, F4.1, and F4.2 pre-
We first divide sales into two components—transaction sent results for IDA.
frequency, measured by the number of weekly sales As shown in Tables F1 and F2, we find that PDA has
transactions by Target.com or Shipt in a market, and a positive effect on delivery platform sales through
transaction size, the average weekly purchase value per increased transaction frequency. Thus, PDA drives addi-
transaction for Target.com or Shipt (Ansari et al., 2008; tional transactions to delivery platforms, but transaction
Lim et al., 2021). size is unaffected by this level of integration. No impact
PARK ET AL. 19

from PDA is found for either retailer online transaction increase in sales attributed to the DID effect varies widely
frequency or retailer online transaction size. depending on prior platform presence: at the 10 percentile
As shown in Tables F3 and F4, IDA has positive level (of prior Shipt's presence in a market), there is a
impacts on both the transaction frequency and transac- 324.6% increase in Shipt's sales, but at 90th percentile
tion size for the retailer's online channel. The increased level, the effect size is only 2.4%.
transaction size is likely driven by the ability of cus-
tomers to aggregate orders of time-sensitive and non-
time-sensitive products. On the other hand, IDA does not 7 | C O N C L U S I O N S AN D
have a significant impact on either transaction size or fre- LIMITATIONS
quency for delivery platforms.
7.1 | Theoretical implications

6.2 | Physical channel sales Our study contributes to the channel integration litera-
ture in the context of last-mile delivery by providing
Our previous discussion focused on the impact of chan- insights into the integration of same-day and standard
nel integration on online sales. However, the impact of last-mile delivery services through a retailer's online
channel integration may spillover onto a retailer's physi- channel and its delivery platform partner's channel. Pre-
cal channel (Fisher et al., 2019). Thus, we explore the vious research on channel integration has primarily
impact of PDA and IDA on physical channel sales.24 focused on the integration of offline and online channels,
Table F5 in Appendix F presents our estimation as exemplified by studies such as Akturk et al. (2018),
results for the impact of PDA and IDA on the retailer's Gallino and Moreno (2014), and Fang et al. (2023). We
physical channel sales. We find that PDA has no signifi- distinguish two levels of retailer-platform integration,
cant impact on physical channel sales (Model F5.1), namely PDA and IDA. We then use channel capabilities
while IDA has a small positive impact (0.9% increase) as theory to hypothesize how these integration levels affect
indicated by Model F5.2. Importantly, these results show the sales performance of the retailer's online channel and
that there does not appear to be online-offline substitu- the delivery platform channel.
tion effects caused by channel integration; that is, we do We find that integration between platform and
not find that online channel integration decreases sales retailer online channels augments sales for both the
in physical channels. delivery platform and the retailer's online sales channel.
Moreover, our research indicates no significant negative
consequences from either the PDA or IDA integration.
6.3 | Prior brand presence Thus, delivery platform integration expands the total
market for the retailer and the delivery platform, indicat-
In our base model, we find that the impact of PDA on ing that integration has complementary benefits for the
platform sales is quite large (202.8% increase). We investi- two channels. Moreover, we find that the delivery plat-
gate whether the size of the impact may be influenced by form and the retailer's online channel are impacted at dif-
the delivery platform, Shipt, entering new markets after ferent stages of the integration process.
integrating with Target. To investigate this possibility, we In the initial integration stage, PDA leads to a signifi-
re-estimate our PDA model, with platform sales as the cant increase in sales (202.8%) for the delivery platform
dependent variable, and include a moderating variable, channel. The effect is driven by the addition of a new retail
PRIOR_PRESENCE, defined as Shipt's sales prior to the offering onto the delivery platform's website, which, in turn,
treatment period. This new variable is interacted with attracts additional customers. The increase in Shipt sales
our post-treatment DID effect, to determine the extent to we calculate corroborates well with Target and Shipt's own
which Shipt's prior presence in a market (or lack thereof) customer records, with Shipt reporting a tripling in mem-
drives the post-treatment impact of PDA integration on bership year-over-year after its PDA integration with Tar-
platform sales. get.25 In addition, online grocery immaturity in a market
Table F6 in Appendix F presents our estimation positively moderates the impact of PDA on platform sales.
results for the impact of PDA on the delivery platform New service offerings can generate a surge in customer
channel sales. We find that the coefficient for TREAT * curiosity, engagement, and sales, contributing to this signifi-
AFTER * PRIOR_PRESENCE is significant and negative, cant effect.26 Thus, it is not surprising that the top-selling
indicating that sales increases due to PDA are greater Target items purchased through Shipt following PDA con-
when the platform presence is lower, prior to the treat- sist of grocery products, such as eggs, milk, bananas, straw-
ment. As reported in Table F7 in Appendix F, the berries, avocados, shredded cheese, butter, limes.27
20 PARK ET AL.

Conversely, during the subsequent integration stage, To illustrate the implications of our estimates,
IDA, there is a significant increase in sales at the Table G1 in Appendix G presents the estimated changes
retailer's online channel and no significant sales change in online sales arising from the two levels of integration,
at the delivery platform channel. Furthermore, online PDA and IDA, in sample markets. We use estimated coef-
grocery immaturity and the presence of a loyal customer ficients from Models 1.1 and 2.3 for these illustrations.28
base in a market positively moderates the impact of IDA For example, in the Chicago, IL, Virginia Beach, VA, and
on a retailer's online sales. Overall, IDA boosts the Washington D.C. markets, the integration efforts have
retailer's online capability with integrated delivery ser- notably increased the combined sales for the retailer's
vices (Ku, 2019; Melis et al., 2016; Oh et al., 2012), meet- online channel and the delivery platform channel by 22%
ing varied consumer demands more effectively. to 43%. Specifically, after implementing PDA, sales in
Finally, we explore the mechanisms through which Chicago surged by 307%, in Virginia Beach by 284%, and
sales are increased following integration. We find that in Washington D.C. by 230%. Following the introduction
PDA increases the frequency of purchases on delivery of IDA, we observe a sales increase of 28% in Chicago, 4%
platforms, while IDA increases both the transaction fre- in Virginia Beach, and 17% in Washington D.C.
quency and transaction size of purchases on the retailer's Moreover, our research reveals that a substantial loyal
online channel. This latter finding suggests that IDA inte- customer base in a market markedly boosts the positive
gration may encourage shoppers to bundle standard and effects of IDA on a retailer's online sales. To fully lever-
same-day delivery purchases, while increasing their shop- age this advantage, retailers should refine their loyalty
ping frequency. programs, ensuring they are finely tuned to meet the spe-
cific needs and preferences of their customers. This could
involve offering tailored promotions, personalized prod-
7.2 | Managerial insights uct recommendations, and exclusive early access to new
products, which all serve to enhance the shopping experi-
As retailers increasingly adopt last-mile delivery plat- ence. Such personalized interactions not only elevate the
forms to fulfill orders, it is crucial to understand how this effectiveness of channel integration but also foster deeper
evolving model impacts customer behavior. Offering customer loyalty and engagement. Consequently, strate-
same-day delivery services, particularly for a diverse gically enhancing loyalty programs becomes an essential
array of perishable products, allows retailers to attract approach for retailers aiming to expand their online cus-
more customers to their online channels. This strategy tomer base and secure a competitive edge in the digital
leverages the distinct advantages of same-day delivery, marketplace.
which can surpass those of standard delivery services. Finally, our research reveals that the benefits of PDA
While standard delivery typically focuses on a broad and IDA are significantly enhanced in markets with low
selection of non-perishable items, same-day delivery online grocery penetration. These markets typically suffer
meets immediate customer needs for freshness and con- from inadequate digital or logistics infrastructure or lim-
venience. This approach enhances customer attraction ited consumer familiarity with online grocery services.
and retention, positioning retailers to be more competi- Strategic investments in channel integration in these con-
tive in the digital marketplace. texts not only address unmet consumer needs but also
Specifically, retailers can choose to only implement accelerate market development and build competitive
PDA or to further integrate with IDA, generating incre- advantage. Consequently, for retailers and delivery plat-
mental sales at each stage. PDA and IDA, while distinct, forms aiming to expand their online presence, focusing
can be perceived as components of an integrated multi- integration efforts in these less mature markets can lead
channel strategy. With PDA, the retailer offers standard to substantial benefits. Potential strategies can include
delivery on its own website and directs customers to the launching targeted marketing campaigns, forging part-
platform's website for same-day delivery. With IDA, cus- nerships with local suppliers to expand product offerings,
tomers can use same-day delivery on the platform's web- and improving logistics for online deliveries, all of which
site or place a single order on the retailer's site that contribute to a stronger market presence and enhanced
includes both same-day and standard delivery options. consumer loyalty in these emerging markets.
Despite enhancing retailer delivery capabilities, the IDA Table 6 summarizes key managerial takeaways of our
stage does not lead to a reduction in delivery platform analyses. Overall, our results offer implications for
channel sales. This result suggests that the two channels retailers exploring the bolstering of their last-mile deliv-
may attract different customer segments, with each seg- ery capabilities through channel integration. Our analysis
ment having a loyal customer base, pointing to specific examines sequential delivery platform integration
promotional and loyalty strategies for each channel. strategies—PDA then IDA—as exemplified by the
PARK ET AL. 21

TABLE 6 Key takeaways.

Integration
level Sales impact Moderators Managerial insights
Platform Significant increase in delivery Online grocery immaturity Managers should leverage the appeal of new online
delivery platform sales enhances the positive impact offerings to boost sales. Addressing immediate
access customer needs for freshness and convenience
(PDA) significantly enhances customer attraction and
retention, particularly in markets with
underdeveloped online grocery sectors
Integrated Significant increase in retailer's Loyal customer base and Retailers should integrate delivery services to more
delivery online sales online grocery immaturity effectively meet diverse consumer demands,
access enhance the positive impact especially in markets with a strong loyal customer
(IDA) base and underdeveloped online grocery sectors.
Promoting bundled delivery options can increase
transaction size and frequency, driving higher
overall sales
Overall Increases total market sales for Managers should consider retailer-platform
impact both channels, with no integration as a strategic market expansion. By
significant negative continuously assessing market maturity and
consequences customer loyalty, they can optimize integration
strategies, enhancing their competitive position in
the digital marketplace

sequence used by Target.com and Shipt, and by other variables for understanding overall business perfor-
retailers, such as Walmart. Additionally, our research mance. Future studies could explore how retailers can
casts light on pivotal considerations regarding platform optimize their profits through the integration of delivery
partnerships. A detailed exploration of retailers that have platform and online retail channels. These results could
not adopted IDA-level integration (such as Costco), may provide valuable insights for retailers seeking to maxi-
further unveil the advantages and disadvantages of mize the benefits of delivery platform integration to
lower-level channel integration. enhance their business performance.
Fourth, we analyze integration between channels
within a corporate environment. However, some
7.3 | Limitations and extensions retailers, such as Aldi and Publix, simply partner with
delivery platforms without a merger. Future studies could
While our study provides important findings, several lim- examine whether the results of this study generalize to
itations should be noted that may impact the generaliz- these settings.
ability and scope of the results. First, our analysis focuses Fifth, our analysis hinges on sequential platform inte-
on the integration between a general merchandise and gration, PDA followed by IDA. This implies that the con-
grocery retailer (Target) and a delivery platform (Shipt), sumer perceptions and behaviors, and the operational
that may limit the generalizability of the findings to other dynamics of firms during the IDA phase, may be
types of retailers. Future research could investigate impacted by prior PDA experiences. Moreover, the
whether the results can be applied to different types of sequential analysis inherently omits the exploration of
retailers and delivery platforms to provide a more com- alternate pathways towards integration, notably, the
prehensive understanding of the impact of delivery plat- direct implementation of IDA without a PDA stage.
form integration on sales performance. Direct IDA adoption could present a distinct set of strate-
Second, our study does not assess potential differ- gic, operational, and consumer dynamics, potentially
ences across various types of products. This may limit the unveiling different outcomes. Future research can exam-
understanding of the impact of PDA and IDA on sales ine these divergent pathways, exploring both sequential
performance. Future studies could conduct product-level and direct adoption of IDA.
analyses to provide more nuanced insights into the Sixth, we did not control for the quality of delivery
impact of delivery platform integration on the types of services in our model. Service quality may influence the
products to optimize retailers' sales performance. positive impact of integration on sales. Future studies
Third, our study did not assess the impact of delivery could examine how the quality of delivery services
platform integration on costs or profits, which are critical offered by retailers and delivery platforms evolve after
22 PARK ET AL.

channel integration. This analysis could provide valuable robustness check does allow for the use of two other delivery
insights into the long-term sustainability of delivery plat- platforms (Amazon Prime Now and Peapod) as controls. See
form integration. Appendix E for the results with these alternate platforms.
7
Finally, our study relies on aggregate spending data As noted by Li and Netessine (2020), treatment and control
in markets to assess buyer behavior, including purchases groups need not be similar in every aspect, as long as the parallel
trend assumption holds.
on the Target.com and Shipt websites. A dataset that pro-
8
vides information on the characteristics of individuals For both the retailers and the delivery platforms, the control
observations do not necessarily come from the same market as
who choose to shop for time-sensitive products online
the treatment observations. For example, Walmart.com in Mar-
and the behavior of these individuals might reveal greater ket Y can serve as the control for Target.com in Market X,
insights into the behavioral responses to online channel according to the selection criteria used.
integration. 9
Walmart operated two separate online channels, Walmart.com
and Walmart Online Grocery, prior to 2020. Walmart.com is sim-
A C K N O WL E D G M E N T S ilar to Target.com, and Walmart Online Grocery to Shipt. Cus-
We thank the Earnest Analytics Company for their provi- tomers could visit Walmart.com to order non-perishable
sion of consumer spending data. products for delivery and visit Walmart Online Grocery to order
same-day delivery of groceries. Thus, over the sample period,
Walmart.com was similar to the pre-IDA Target.com.
ORCID 10
PDA was implemented in different markets at various times.
Kevin H. Park https://orcid.org/0000-0003-0734-698X When PDA was implemented in market A at time t, the subse-
Xiaodan Pan https://orcid.org/0000-0001-5836-2133 quent period is designated as “AFTER.” When market A is
paired with a control, market B, the same time definitions apply
E N D N O T ES for consistency; hence “AFTER” is specific to each matched pair
1 and can be identified using weekly fixed effects in the PDA
Although Target acquired Shipt and may be indifferent to the
sample.
source of revenue (i.e., either through Target.com or Shipt), other
11
major retailers partner with delivery platforms that are indepen- For Equations (1) and (3), the subscript i is not used, since, for
dently owned (e.g., Costco and Instacart). Therefore, the comple- PDA, we compare a Target (Shipt) treated market to a Target
mentarity or substitutability of the channels at the two stages of (Shipt) untreated market. Conversely, in Equations (2) and (4),
integration is a relevant question, since it helps determine the the subscript i is needed since, for IDA, we contrast Target.com
balance of revenues between the two channels (Pozzi, 2013). with Walmart.com, and Shipt with Amazon Fresh; thus, we need
Moreover, the impact of integration on total revenues is a con- the subscript to denote the different retailers and platforms.
12
cern for partners regardless of whether they share common Although not all moderating relationships are hypothesized, we
ownership. include variables for these moderations in our model for the sake
2
Delivery platforms do not exclusively deliver perishable items. of completeness.
Customers can order household essentials, personal care prod- 13
We use the ratio ‘prior to’ the PDA integration because of endo-
ucts, and sporting goods for same-day delivery. However, this geneity concerns (Wang & Goldfarb, 2017). Specifically, last-mile
service may not cover many items, such as certain clothing, elec- delivery platform integration could affect the loyal customer base
tronics, alcohol, and items weighing over 40 pounds (Source: in a market. We also tested our model with different period
Target.com). lengths (5-week, 7-week, and 8-week). The main results remain
3
Source: Target (https://corporate.target.com/news-features/ unchanged.
article/2017/12/target-acquires-shipt); accessed on December 14
We use the ratio ‘prior to’ PDA integration, since both types of
7, 2023). integration may affect online grocery immaturity, resulting in
4
Source: Shipt (https://www.shipt.com/partner). Accessed April endogeneity bias. Total online non-grocery delivery sales include
18, 2023. sales from online retailers and marketplaces such as Amazon,
5 Best Buy, eBay, Walmart.com, and Target.com. The total sales
We selected a 21-week time frame to ensure a balanced period
from these sites are divided by the larger total that also includes
before and after the treatment implementation while minimizing
sales from the sites that specialize in the distribution of groceries
end-of-year seasonal effects. Adjusting this time window did not
(e.g., Shipt, Instacart) to compute IMMATUREj.
alter our findings. Alternatively, we analyzed the 2018 data
15
exclusively, without balancing the pre- and post-treatment The subscript i is added to variables Equation (4) to denote a spe-
periods for the PDA sample, and the 2019 data for the IDA sam- cific delivery platforms or retailers; Target.com, Walmart.com,
ple. The results remained qualitatively consistent across these Shipt and Amazon Fresh. See details in footnote 11.
16
different approaches. For the PDA sample, the pre-treatment period during the year of
6
Other delivery platforms, including Instacart, and Walmart 2018 encompasses weeks 1 through 22 of 2018. The pre-
Online Grocery, had significantly higher sales growth than Shipt treatment period during the year of 2019 for the IDA sample
prior to the treatment effect. Therefore, these firms could not be spans from week 1 to week 21 of 2019.
17
used in the control group because the parallel trend assumption, Data used for this exploratory analysis are at the national level
prior to treatment, was violated (Li & Netessine, 2020). Our since city-market level data are not available.
PARK ET AL. 23

18
Moderating variables are included even if hypotheses are not Ansari, A., Mela, C. F., & Neslin, S. A. (2008). Customer channel
proposed for completeness and transparency. migration. Journal of Marketing Research, 45(1), 60–76.
19
Two-way interaction terms are excluded due to their collinearity Avery, J., Steenburgh, T. J., Deighton, J., & Caravella, M. (2012).
with various fixed effects, as recognized and employed in prior Adding bricks to clicks: Predicting the patterns of cross-channel
literature (Lim et al., 2021; Son et al., 2020; Wang & elasticities over time. Journal of Marketing, 76(3), 96–111.
Goldfarb, 2017). Bai, J., So, K. C., Tang, C. S., Chen, X., & Wang, H. (2019). Coordi-
20 nating supply and demand on an on-demand service platform
In Section 6.3, we provide evidence that this large impact is
with impatient customers. Manufacturing & Service Operations
driven by sales in city markets where Shipt had minimal or no
Management, 21(3), 556–570.
sales prior to PDA.
Balasubramanian, S., Raghunathan, R., & Mahajan, V. (2005). Con-
21
Two-way interaction terms are excluded due to their collinearity sumers in a multichannel environment: Product utility, process
with various fixed effects, as recognized and employed in prior utility, and channel choice. Journal of Interactive Marketing,
literature (Lim et al., 2021; Son et al., 2020; Wang & 19(2), 12–30.
Goldfarb, 2017). Bang, Y., Lee, D. J., Han, K., Hwang, M., & Ahn, J. H. (2013). Chan-
22
Moderating variables are included even if hypotheses are not nel capabilities, product characteristics, and the impacts of
proposed for completeness and transparency. mobile channel introduction. Journal of Management Informa-
23
We thank an anonymous reviewer for this point. For the IDA tion Systems, 30(2), 101–126.
estimations, we use different samples for the retailer and the Bijmolt, T. H., Leeflang, P. S., Block, F., Eisenbeiss, M.,
platform estimations. Therefore, a seemingly unrelated regres- Hardie, B. G., Lemmens, A., & Saffert, P. (2010). Analytics for
sion is not possible. customer engagement. Journal of Service Research, 13(3),
24 341–356.
We thank an anonymous reviewer for this point.
Boyer, K. K., & Hult, G. T. M. (2005). Extending the supply chain:
25
https://corporate.target.com/news-features/article/2018/12/ Integrating operations and marketing in the online grocery
shipt-anniversary, accessed 9/14/2024. industry. Journal of Operations Management, 23(6), 642–661.
26 Boyer, K. K., & Hult, G. T. M. (2006). Customer behavioral inten-
During the period of our study, it is plausible that dynamic
growth rates and other developments might have been at play, tions for online purchases: An examination of fulfillment
influencing consumer purchasing behaviors and sales trajectories method and customer experience level. Journal of Operations
across channels, confounding our treatment effects. One poten- Management, 24(2), 124–147.
tial confounder might be new partnerships with other retailers. Cao, L., & Li, L. (2015). The impact of cross-channel integration on
However, this likely would not have a major impact on our retailers' sales growth. Journal of Retailing, 91(2), 198–216.
results because Shipt, our delivery platform, partnered with Carter, C. R. (2011). A call for theory: The maturation of the supply
smaller group of retailers during our sample period in contrast to chain management discipline. Journal of Supply Chain Man-
other platforms, such as Instacart, which aggressively partnered agement, 47(2), 3–7.
with numerous retailers. Furthermore, in our model-free results, Castillo, V. E., Bell, J. E., Rose, W. J., & Rodrigues, A. M. (2018).
the initial surge in sales after treatment is maintained consis- Crowdsourcing last mile delivery: Strategic implications and
tently during the treatment period without significant ups and future research directions. Journal of Business Logistics, 39(1),
downs in sales trends. We thank an anonymous reviewer for the 7–25.
insightful comment. Deleersnyder, B., Geyskens, I., Gielens, K., & Dekimpe, M. G.
27
https://corporate.target.com/news-features/article/2018/12/ (2002). How cannibalistic is the internet channel? A study of
shipt-anniversary, accessed 9/14/2024. the newspaper industry in the United Kingdom and The
28
Netherlands. International Journal of Research in Marketing,
For instance, in Chicago, PDA is estimated to increase Shipt's
19(4), 337–348.
sales by 307% if online grocery immaturity is 0.898 (Model 1.1).
Eisenhardt, K. M. (2013). Top management teams and the perfor-
We multiply this number by the average Shipt sales prior to PDA
mance of entrepreneurial firms. Small Business Economics, 40,
to calculate the dollar impact of PDA on Shipt sales in Chicago.
805–816.
Fang, J., Liu, H., Cai, Z., & Tan, C. W. (2023). Omni-channel retail-
ing on platforms: Disentangling the effects of channel integra-
R EF E RE N C E S tion and inter-platform function usage difference. Journal of
Akturk, M. S., Ketzenberg, M., & Heim, G. R. (2018). Assessing Operations Management, 69(2), 197–216.
impacts of introducing ship-to-store service on sales and Fisher, M. L., Gallino, S., & Xu, J. J. (2019). The value of rapid deliv-
returns in omnichannel retailing: A data analytics study. Jour- ery in omnichannel retailing. Journal of Marketing Research,
nal of Operations Management, 61, 15–45. 56(5), 732–748.
Alba, J., Lynch, J., Weitz, B., Janiszewski, C., Lutz, R., Gallino, S., & Moreno, A. (2014). Integration of online and offline
Sawyer, A., & Wood, S. (1997). Interactive home shopping: channels in retail: The impact of sharing reliable inventory avail-
Consumer, retailer, and manufacturer incentives to participate ability information. Management Science, 60(6), 1434–1451.
in electronic marketplaces. Journal of Marketing, 61(3), 38–53. Gessner, K. (2020). Instacart dominates delivery, while top gro-
Alnaggar, A., Gzara, F., & Bookbinder, J. H. (2021). Crowdsourced cers see little shift in market share. Second Measure https://
delivery: A review of platforms and academic literature. Omega, secondmeasure.com/datapoints/grocery-spending-delivery-
98, 102139. trends
24 PARK ET AL.

Haleem, A. (2023). Target digital sales, total sales both dip in Q2. purchase process in emerging and mature markets. Interna-
DigitalCommerce360. https://www.digitalcommerce360.com/ tional Journal of Research in Marketing, 30(1), 57–68.
article/target-online-sales/#::text=Target%20digital%20sales% Pozzi, A. (2013). The effect of internet distribution on brick-
2C%20or%20those,for%2017.2%25%20of%20total%20sales and-mortar sales. The Rand Journal of Economics, 44(3),
Halvorsen, R., & Palmquist, R. (1980). The interpretation of dummy 569–583.
variables in semilogarithmic equations. American Economic Redman, R. (2022). Study: Online grocery shifts from need to con-
Review, 70(3), 474–475. venience. https://www.supermarketnews.com/online-retail/
Han, C., Dong, Y., & Dresner, M. (2013). Emerging market penetra- study-online-grocery-shifts-need-convenience
tion, inventory supply, and financial performance. Production Ren, X., Windle, R. J., & Evers, P. T. (2023). Channel transparency
and Operations Management, 22(2), 335–347. and omnichannel retailing: The impact of sharing retail store
Hwang, M., & Park, S. (2016). The impact of Walmart supercenter product availability information. Journal of Operations Manage-
conversion on consumer shopping behavior. Management Sci- ment, 69(2), 217–245.
ence, 62(3), 817–828. Rioux, P. (2020). The value of investing in loyal customers. Forbes.
Ku, H. H. (2019). Consumer affects when making undesirable https://www.forbes.com/sites/forbesagencycouncil/2020/01/29/
purchases to meet the minimum purchase requirement: the-value-of-investing-in-loyal-customers/?sh=56be989121f6
Decision-related variables as moderators. Journal of Consumer Saphores, J. D., & Xu, L. (2021). E-shopping changes and the state
Behaviour, 18(1), 53–62. of E-grocery shopping in the US-evidence from national travel
Kumar, V., & Shah, D. (2004). Building and sustaining profitable and time use surveys. Research in Transportation Economics,
customer loyalty for the 21st century. Journal of Retailing, 87, 100864.
80(4), 317–329. Son, Y., Kwon, H. E., Tayi, G. K., & Oh, W. (2020). Impact of cus-
Lemon, K. N., & Verhoef, P. C. (2016). Understanding customer tomers' digital banking adoption on hidden defection: A com-
experience throughout the customer journey. Journal of Mar- bined analytical–empirical approach. Journal of Operations
keting, 80(6), 69–96. Management, 66(4), 418–440.
Li, J., & Netessine, S. (2020). Higher market thickness reduces Sousa, R., Horta, C., Ribeiro, R., & Rabinovich, E. (2020). How to
matching rate in online platforms: Evidence from a quasi- serve online consumers in rural markets: Evidence-based rec-
experiment. Management Science, 66(1), 271–289. ommendations. Business Horizons, 63(3), 351–362.
Lien, T. (2017). Apoorva Mehta had 20 failed start-ups before Insta- Soysal, G., & Krishnamurthi, L. (2016). How does adoption of the
cart. Los Angeles Times https://www.latimes.com/business/ outlet channel impact customers' spending in the retail
technology/la-fi-himi-apoorva-mehta-20170105-story.html stores: Conflict or synergy? Management Science, 62(9),
Lim, S. F. W., Jin, X., & Srai, J. S. (2018). Consumer-driven 2692–2704.
e-commerce: A literature review, design framework, and research Stathopoulou, A., & Balabanis, G. (2016). The effects of loyalty pro-
agenda on last-mile logistics models. International Journal of grams on customer satisfaction, trust, and loyalty toward high-
Physical Distribution & Logistics Management, 48(3), 308–332. and low-end fashion retailers. Journal of Business Research,
Lim, S. F. W., Rabinovich, E., Park, S., & Hwang, M. (2021). Shop- 69(12), 5801–5808.
ping activity at warehouse club stores and its competitive and Stern, M. (2022). Will more Americans make e-grocery delivery a
network density implications. Production and Operations Man- weekly habit? https://retailwire.com/discussion/will-more-
agement, 30(1), 28–46. americans-make-e-grocery-delivery-a-weekly-habit/
Lim, S. F. W., & Srai, J. S. (2018). Examining the anatomy of last- Tang, Q., Lin, M., & Kim, Y. (2021). Inter-retailer channel competi-
mile distribution in e-commerce omnichannel retailing: A sup- tion: Empirical analyses of store entry effects on online pur-
ply network configuration approach. International Journal of chases. Production and Operations Management, 30(8), 2547–
Operations & Production Management, 38(9), 1735–1764. 2563.
Mascarenhas, B., Kumaraswamy, A., Day, D., & Baveja, A. (2002). Thomas, L. (2017). Amazon officially owns whole foods; here are the
Five strategies for rapid firm growth and how to implement products that are getting marked down. CNBC. https://www.
them. Managerial and Decision Economics, 23(4–5), 317–330. cnbc.com/2017/08/28/amazon-officially-owns-whole-foods-
Melis, K., Campo, K., Lamey, L., & Breugelmans, E. (2016). A big- here-are-the-products-that-are-getting-marked-down.html
ger slice of the multichannel grocery pie: When does con- Tucker, C., & Zhang, J. (2010). Growing two-sided networks by
sumers' online channel use expand retailers' share of wallet? advertising the user base: A field experiment. Marketing Sci-
Journal of Retailing, 92(3), 268–286. ence, 29(5), 805–814.
Neslin, S. A., Grewal, D., Leghorn, R., Shankar, V., Teerling, M. L., Vesel, P., & Zabkar, V. (2009). Managing customer loyalty through
Thomas, J. S., & Verhoef, P. C. (2006). Challenges and opportu- the mediating role of satisfaction in the DIY retail loyalty pro-
nities in multichannel customer management. Journal of Ser- gram. Journal of Retailing and Consumer Services, 16(5),
vice Research, 9(2), 95–112. 396–406.
Oh, L. B., Teo, H. H., & Sambamurthy, V. (2012). The effects of Wang, K., & Goldfarb, A. (2017). Can offline stores drive online
retail channel integration through the use of information tech- sales? Journal of Marketing Research, 54(5), 706–719.
nologies on firm performance. Journal of Operations Manage- Wang, L., Rabinovich, E., & Richards, T. J. (2022). Scalability in
ment, 30(5), 368–381. platforms for local groceries: An examination of indirect net-
Pauwels, K., Erguncu, S., & Yildirim, G. (2013). Winning hearts, work economies. Production and Operations Management,
minds and sales: How marketing communication enters the 31(1), 318–340.
PARK ET AL. 25

Xu, K., Chan, J., Ghose, A., & Han, S. P. (2017). Battle of the chan-
nels: The impact of tablets on digital commerce. Management How to cite this article: Park, K. H., Pan, X., &
Science, 63(5), 1469–1492.
Dresner, M. E. (2024). Targeting online sales
through last-mile delivery platform integration.
S UP PO RT ING IN FOR MAT ION
Journal of Operations Management, 1–25. https://
Additional supporting information can be found online
doi.org/10.1002/joom.1338
in the Supporting Information section at the end of this
article.