Reducing food waste through digital platforms: A quantification of cross-side network effects

Abstract

To fight food waste, retail stores have begun selling perishable food close to the expiration date at discounted prices. To render this form of last-minute discounting effective, digital platforms have been developed with the major aim to connect local retail stores and their consumers by sharing information about these discounts. To sustain digital platforms, platform leaders need to ensure both consumers and retail stores remain active on it. To provide platform leaders with advice on how to create a sustainable digital platform, we examine how retail store activity on the digital platform affects consumer activity, and vice versa (also known as cross-side network effects). By combining a PVAR model and an impulse response function, along with data from a digital platform aimed at food waste reduction, we find that the effect of consumer activity on retail store activity is stronger and more long-lasting than the effect of retail store activity on consumer activity. We discuss the implications of our findings for both retail stores and digital platform leaders.

Full text

1
Reducing food waste through digital platforms:
A quantification of cross-side network effects
Shantanu Mullick1,5
Néomie Raassens2,5
Hans Haans3
Edwin J. Nijssen4
1 Corresponding author, Assistant Professor in Marketing, Eindhoven University of
Technology, P.O. Box 513, 5600 MB Eindhoven, The Netherlands, s.mullick@tue.nl.
2 Assistant Professor in Marketing, Eindhoven University of Technology, P.O. Box 513, 5600
MB Eindhoven, The Netherlands, n.raassens@tue.nl.
3 Assistant Professor in Marketing, Tilburg University, P.O. Box 90153, 5000 LE Tilburg,
The Netherlands, haans@uvt.nl.
4 Professor in Entrepreneurial Marketing, Eindhoven University of Technology, P.O. Box
513, 5600 MB Eindhoven, The Netherlands, e.j.nijssen@tue.nl.
5 These authors contributed equally to this work.

2
Reducing food waste through digital platforms:
A quantification of cross-side network effects
To fight food waste, retail stores have begun selling perishable food close to the expiration
date at discounted prices. To render this form of last-minute discounting effective, digital
platforms have been developed with the major aim to connect local retail stores and their
consumers by sharing information about these discounts. To sustain digital platforms,
platform leaders need to ensure both consumers and retail stores remain active on it. To
provide platform leaders with advice on how to create a sustainable digital platform, we
examine how retail store activity on the digital platform affects consumer activity, and vice
versa (also known as cross-side network effects). By combining a PVAR model and an
impulse response function, along with data from a digital platform aimed at food waste
reduction, we find that the effect of consumer activity on retail store activity is stronger and
more long-lasting than the effect of retail store activity on consumer activity. We discuss the
implications of our findings for both retail stores and digital platform leaders.
Keywords: food waste reduction, digital platform, two-sided market, cross-side network
effects, retailing

3
1. Introduction
Almost one third of all food produced globally is wasted or lost every year (FAO, 2017).
Food waste has environmental, social, and economic costs (Stöckli, Niklaus, & Dorn, 2018).
As retail store Tesco stated: “Food waste resulted in significant costs to our business, as well
as our suppliers and our customers” (Little & Castella, 2017). Because these costs are
increasing along the supply chain (Capgemini, 2017; Schanes, Dobernig, & Gözet, 2018),
movements against food waste are emerging (Cicatiello et al., 2017). Especially grocery
retailers are beginning to recognize the financial and reputational potential of reducing food
waste in their operations (Winsight Grocery Business, 2018) and are, therefore, at the heart of
initiatives to reduce food waste (Capgemini, 2017).
Industry studies find that food waste represents an $18 billion profit opportunity for
grocery retailers (ReFED, 2018), with retailers able to obtain a median of five dollars in return
for every dollar invested in food waste reduction (Capgemini, 2017). To tackle the food waste
problem, an increasing number of different food sharing models are being developed
(Michelini, Principato, & Iasevoli, 2018). The opportunity for these models mainly arises in
digital technology and the emerging phenomenon of the sharing economy (Ciulli, Kolk, Boe-
Lillegraven, 2019; Michelini, Principato, & Iasevoli, 2018). Table 1 shows some illustrative
examples of initiatives around food waste reduction that have recently been taken, by grocery
retailers as well as food retailers in general.
[Insert Table 1 about here]
An interesting food sharing initiative that has seen a rapid increase in adoption by grocery
retailers in several Western countries is last-minute discounting of perishables (Aschemann-

4
Witzel, 2018; Robertson & Parfitt, 2018). In last-minute discounting, grocery retail stores
lower the price of perishable products close to their date of expiry. When perishables
approach their expiration date, consumers perceive them as sub-optimal due to food safety
and risk concerns (Stangherlin, Duarte Ribeiro, & Barcellos, 2019). However, the use of last-
minute discounting motivates consumers to buy these sub-optimal products (Buisman,
Haijema, & Bloemhof-Ruwaard, 2019; de Hooge et al., 2017), thereby generating revenue
from perishables that would otherwise have been wasted (Adenso-Díaz, Lozano, & Palacio,
2017). Thus, last-minute discounting creates a win-win situation where (i) retail stores reduce
food waste thereby mitigating the financial loss associated with it (Tsiros & Heilman, 2005;
Wang & Li, 2012), and (ii) consumers are able to save money while contributing to solve the
food waste problem (Aschemann-Witzel, 2018; Aschemann-Witzel et al., 2017).
However, information on these last-minute discounts do not always reach consumers in
time (i.e., before the recommended expiry date), thereby rendering these discounts ineffective.
Typically, last-minute discounts are given one or two days before the date of expiry of
perishables (cf. Aschemann-Witzel, 2018; Theotokis, Pramatari, & Tsiros, 2012), leaving
consumers with little time to take advantage of them. As a consequence, IT solutions in the
form of digital platforms have been launched to create ties between grocery retailers and
consumers, which help facilitate timely information exchange (cf. Capgemini, 2017). The
importance of creating ties is recognized by the extant sustainability literature, which even
argues that the absence of suitable ties between retail stores and consumers (i.e., supply chain
partners) hinders a more sustainable and appreciative handling of food (Schanes, Dobernig, &
Gözet, 2018). Digital platforms can foster food waste reduction by creating the necessary ties
(Ciulli, Kolk, & Boe-Lillegraven, 2019). Specifically, by connecting retail stores and
consumers, digital platforms play a crucial role in enhancing the effectiveness of last-minute
discounts, thereby also helping to reduce food waste.

5
Since digital platforms are a relatively recent phenomenon, research is yet to explore how
ties between retail stores and consumers are created (Ciulli, Kolk, & Boe-Lillegraven, 2019).
Therefore, the objective of our study is to examine how digital platforms tie together retail
stores and consumers to enhance the effectiveness of last-minute discounts. Specifically, we
explore whether the effect of consumer activity on retail store activity is different from the
effect of retail store activity on consumer activity. To provide a better understanding of these,
possibly asymmetric, interactions, we rely on the literature of two-sided markets and dynamic
cross-side network effects (CNEs). We model the retail store-consumer interactions using
Panel Vector Auto Regression (PVAR) and apply it to panel data from a digital platform
designed for reducing food waste of grocery retail stores. The PVAR model allows us to
characterize the dynamics of CNEs in terms of both short-term and long-term effects. The
results of our study will help digital platform leaders effectively market their service, thereby
increasing their chances of sustaining the platform. Currently, digital platform leaders market
their solutions to retail stores and consumers alike, since little is known about which party is
more pivotal in sustaining digital platforms.
We contribute to the literature in two ways. First, we contribute to the literature on food
waste. Previous work investigates the general impact of temporary price discounts on the
sales of perishables and finds that price promotions on perishables can be effective (e.g.,
Donselaar et al., 2016; Nijs et al., 2001). Also in the context of last-minute discounting to
reduce food waste, research suggests that consumers react favorably to last-minute discounts
(e.g., Aschemann-Witzel et al., 2019; Theotokis, Pramatari, & Tsiros, 2012). However, this
research focuses on consumers who have viewed these discounts in the store and are thus in a
position to take advantage of them. As a consequence, they do not address the question of
how these discounts can be disseminated to a wider audience in a short time span (one to two
days), in order to increase their effectiveness. Given the transient nature of last-minute

6
discounts, our study investigates how digital platforms can help enhance the effectiveness of
last-minute discounts by creating ties between retail store and consumers. In addition, we
make advances to the empirical literature on last-minute discounting to reduce food waste by
using secondary data, since prior research in this area is primarily based on experimental
studies (e.g., Aschemann-Witzel et al., 2019 and Theotokis, Pramatari, & Tsiros, 2012).
Second, our study contributes to the literature of two-sided markets and cross-side
network effects. Most work in this domain has emerged in the areas of strategic management
and economics (e.g., Kouris & Kleer, 2012; Rysman, 2009; Stummer, Kundisch, & Decker,
2018) and is of a theoretical nature. Thus, studies primarily use stylized analytic models
(Sriram et al., 2015), leaving ample scope for empirical work. In particular, we build upon
sparse empirical work on two-sided markets that quantifies cross-side network effects, i.e., the
relative importance of one side over the other (e.g., Chu & Manchanda, 2016; Cong et al.,
2019; Song et al., 2018; Thies, Wessel, & Benlian, 2018; Voigt & Hinz, 2015). We add to the
extant literature by studying how CNEs operate for digital platforms that focus on food waste
reduction by connecting consumers with retail stores that sell perishable consumer products,
such as fruits and vegetables, close to their expiration date. In particular, digital platforms that
focus on food waste reduction need to disseminate information from retailers to consumers
and generate consumer interest very quickly (within one or two days) to develop and sustain
CNEs. Although our study focuses on food waste and relates closely to the grocery retail
context, our findings can be generalized to other food retailers such as bakeries and
restaurants and other retailers that sell consumer perishable products, such as florists and even
pharmacies, since they sell products such as over-the counter food supplements and dietary
aids.
This study is organized as follows. First, we provide the theoretical background for our
study. Particularly, we discuss how digital platforms that focus on food waste reduction can

7
push food waste reduction. Then, we argue that these digital platforms can be conceptualized
as two-sided markets and explain the asymmetric cross-side network effects that occur within
these markets. Subsequently, we describe the method and present the results. In the final
section, we discuss the implications of our study and provide directions for future research.
2. Theoretical background
2.1. Tackling food waste at retail stores by last-minute discounting
Last-minute discounting in the context of food waste has been studied in the past
(Buisman, Haijema, & Bloemhof-Ruwaard, 2019). From the perspective of the retailer, the
effect of discounting on retailer performance (e.g., Adenso-Díaz, Lozano, & Palacio, 2017;
Gauri et al., 2017) and determining the optimal price for perishables has been the focal point
of interest (e.g., Adenso-Díaz, Lozano, & Palacio, 2017; Chung, 2019). From the consumer
perspective, researchers have focused on the willingness to buy and the attitude (e.g.,
satisfaction) towards products that are discounted but suboptimal (e.g., de Hooge et al., 2017;
Le Borgne, Sirieix, & Costa, 2018) and on the effects of last-minute discounting on consumer
perceptions of brand quality (e.g., Theotokis, Pramatari, & Tsiros, 2012). Although these
studies provide valuable insights, they are primarily analytical or experimental in nature. We
are among the first to empirically assess last-minute discounting in the context of food waste
using secondary data.
Last-minute discounts are typically offered by a retail store one or two days before the
product expiration date by placing discount stickers on the physical product. Consumers
visiting the retail store may see them and take advantage. Research suggests that consumers
react favorably to last-minute discounts (Aschemann-Witzel et al., 2019), thereby providing

8
reassurance to retail stores that their initiative is well received. However, while valuable
insights are gained from this research, we note that the study only speaks to situations where
the consumer is already in the store, and thus notices the retail store’s initiative to reduce food
waste. But most consumers are not in the store when the discount is available and as a result
remain unaware of the offer, implying an information asymmetry problem. To exploit this
untapped potential, the effectiveness of retail stores’ current initiatives to reduce food waste
largely depend on digital platforms.
2.2. Digital platforms aimed at food waste reduction: Two-sided markets
Digital platforms play a central role in generating the interplay between its users by
providing efficient and effective ways to match users who have offerings with those who want
those offerings (Rangaswamy et al., 2020). Specifically, by posting last-minute discounts
from grocery retail stores to the digital platform where they can be viewed by consumers,
digital platforms enable these stores to alert existing traffic or even create additional traffic by
connecting retail stores to local consumers and, thus, resolve the information asymmetry
problem (Ciulli, Kolk, & Boe-Lillegraven, 2019). Hence, digital platforms, which focus on
food waste reduction, are conceptualized as two-sided markets with the primary function of
connecting retail stores and consumers (cf. Ciulli, Kolk, & Boe-Lillegraven, 2019;
Frishammar et al., 2018).
Two-sided platforms are specific multi-sided platforms composed of a platform leader and
two distinct user networks that provide each other with network benefits (Eisenmann, Parker,
& Van Alstyne, 2006; Landsman & Stremersch, 2011; Muzellec, Ronteau, & Lambkin, 2015;
Rysman, 2009). The platform leader facilitates the user networks and tries to build and
maintain the digital platform’s business model (Helfat & Raubitschek, 2018). The business

9
model should ensure that value is created and enough money is earned to keep the digital
platform alive. For last-minute discounting of perishables close to expiry date, value is created
(i.e., value proposition) by connecting retail stores that upload these discounts and consumers
who search for these discounted perishables by using the digital platform. This allows better
utilization of perishable food, a category which normally witnesses high food waste (Parfitt,
Barthel, & Macnaughton, 2010). To remain sustainable, the digital platform should aim for
positive cross-side network effects (Cong et al., 2019): if the digital platform leader can
attract enough consumers, retail stores will be eager to join the digital platform, and vice
versa. This so-called chicken and egg problem is widely prevalent in digital platforms, which
operate as two-sided markets (Evans & Schmalensee, 2010; Stummer, Kundisch, & Decker,
2018). Specifically, uploading a large number of perishable products to the digital platform by
retail stores have a positive effect on the utility of consumers, thereby attracting more
consumers to subscribe to and view discounted perishables on the digital platform and hence
increasing the probability of food waste reduction. At the same time, consumers on the digital
platform positively affect retail stores’ decision to upload perishable products to the digital
platform. The more active consumers, the larger is the market potential for retail stores to
achieve food waste reduction goals (which also provide economic benefits); more retail stores
in turn attract more consumers to the digital platform.
2.3. Cross-side network effects
Research on two-sided markets point out that the benefit of joining one side of a digital
platform depends on the total number of users on the other side of the platform, and this is
called the cross-side network effect (Chu & Manchanda, 2016). Empirical work on cross-side
network effects has remained limited, but has shown rapid growth of late (Chu & Manchanda,

10
2016; Song et al., 2018; Sridhar et al., 2011; Voigt & Hinz, 2015). While past research has
studied symmetric cross-side network effects in two-sided markets (see, e.g., Sriram et al.,
2015 and Voigt & Hinz, 2015 for an overview), little attention has been paid to the
asymmetric interactions between the two sides of a digital platform (Chu & Manchanda,
2016; Song et al., 2018; Thies, Wessel, & Benlian, 2018; Voigt & Hinz, 2015). To sustain a
digital platform (including the focal initiative to reduce food waste), however, both CNEs are
important (Cong et al., 2019), for example, retail stores on consumers and consumers on retail
stores.
We identify a few studies that empirically quantify asymmetric two-way cross-side
network effects.1,2 Chu and Manchanda (2016) quantify two-way CNEs (i.e., buyers on sellers
and sellers on buyers) and their evolution over the platform’s life cycle for a customer-to-
customer (C2C) online platform. The findings show a large and positive cross-network effect
on both sides of the platform, but also that this effect is asymmetric in that the installed base
of sellers have a much larger effect on the growth of buyers than vice versa. Also, Voigt and
Hinz (2015) focus on a C2C platform to investigate network effects. In particular, they
investigate users’ spending behavior on an online dating platform and find asymmetric CNEs,
i.e., men are more likely to pay for online dating services when there is a sufficient installed
user base of women than vice versa. In a business-to-business (B2B) setting, Song et al.
(2018) examine how CNEs on different platform sides (application software side and user
side) of Mozilla Firefox are temporally asymmetric and find a long-term CNE from the user
side to the application side and a short-term CNE from the application side to the user side.
Finally, Thies, Wessel, and Benlian (2018) empirically assess CNEs for a crowdfunding
1 Another identified study that empirically assesses CNEs is the study of Cong et al. (2019). Unlike the other
identified studies and our own research context in which we focus on one specific digital platform and its users,
this study examines CNEs across multiple platforms with the focus on platform failure.
2 Please note that the identified studies may document same-side network effects and non-network factors, but
because of our focus on CNEs, we only report findings on cross-side network effects.

11
platform (which is classified as a business-to-consumer (B2C) context). By analyzing data on
the evolution of Kickstarter, a large crowdfunding platform, they find evidence for
asymmetric network effects. In particular, while an increasing number of entrepreneurial
projects has a strong effect on the installed base of funders, an increased installed base of
funders does not have an effect on the growth of the number of entrepreneurial projects.
Our paper is related to these studies, but with one notable difference. The two-sided
market we study is unique because discounting perishables to reduce food waste is very time
sensitive. Specifically, the quality of perishables deteriorates continuously until it is no longer
suitable for sale or consumption (Wang & Li, 2012). Thus, when retail stores upload last-
minute discounts on the digital platform, consumers have a small time window to view these
discounts and act (about one or two days). If a ‘mismatch’ between retail stores and
consumers arises, it renders last-minute discounts ineffective, and, in turn, it becomes more
difficult to sustain the digital platform and to accomplish the goal of reducing food waste.
We follow previous research (e.g., Song et al., 2018; Thies, Wessel, & Benlian, 2018;
Voigt & Hinz, 2015) by adopting user activity as a measure of installed base.3 The main
reason is that user networks contribute to each other in dimensions other than membership
(i.e., the size of the two user networks) (Landsman & Stremersch, 2011). In particular, retail
stores and consumers who simply join the digital platform without taking any action do not
create any value (i.e., they do not contribute to the overall goal of food waste reduction).
Thus, the interdependency between user networks on both sides extends beyond membership
and involves usage too (Rysman, 2009). In our setting, when focusing on digital platforms
aimed at food waste reduction, it is not just the number of retail stores and consumers that join
3 In essence, while some papers refer to installed bases, they actually operationalize installed bases as an activity-
based measure (e.g., Song et al., 2018; Thies, Wessel, & Benlian, 2018; Voigt & Hinz, 2015). For example, Song
et al. (2018), who study a platform with applications on one side and users on the other side, use application
quantity (i.e., number of applications) as a proxy for application usage on one side and a measure of user activity
(i.e., the average daily times of actual usage) on the other side. Thus, like other studies, our paper actually
captures the installed base, which we operationalize as retail store and consumer activity.

12
the platform that matters, but more important is the activity of these retail stores and
consumers. The current study, therefore, focuses on the relation between the activities of retail
stores and consumers and explores whether asymmetric CNEs exist between these two sides:
how does retail store activity, such as the number of perishables on last-minute discount that
are uploaded, impact consumer activity, such as the number of views of discounted
perishables, and vice-versa?
2.4. Asymmetric interactions between retail stores and consumers
For digital platforms to survive, platform leaders must maintain positive cross-side
network effects (Helfat & Raubitschek, 2018; Thies, Wessel, & Benlian, 2018), which is a
major challenge (Voigt & Hinz, 2015). Each side of the market has its own set of distinct
features and agenda that may not align. These distinct features of each of the two sides of the
market may result in asymmetric influences of CNEs between the two sides which can have
important implications for sustaining the digital platform (Cong et al., 2019; Song et al.,
2018). In particular, platform leaders are typically concerned with understanding the primacy
of one side versus the other. By exploring whether the two cross-side network effects are
asymmetric, it is possible to pinpoint the side of the digital platform that is more important in
sustaining the platform (Chu & Manchanda, 2016).
Retail store-to-consumer effect. The main goal of retail stores participating in digital
platforms aiming to reduce food waste is to sell more perishables before they expire and need
to be thrown away. Thus, retail stores seek to increase their revenues by diminishing costs
inherent in food waste (Tsiros & Heilman, 2005; Wang & Li, 2012). Since uploading of last-
minute discounts by retail stores stimulates consumers to use the digital platform to search for
discounted perishables, we refer to it as the retail store-to-consumer effect. Empirical

13
evidence of this effect is provided by Nijs et al. (2001), who find that the short- and long-term
effectivity of price promotions is greater for perishable goods than for other product
categories. Moreover, since consumers are less willing to pay for perishable products close to
expiry (Tsiros & Heilman, 2005), discounting them may be an effective strategy to waste less
food and increase profits (Aschemann-Witzel et al., 2019; Tsiros & Heilman, 2005; Wang &
Li, 2012). In addition, last-minute discounting of perishables can build trust between retail
stores and consumers because it minimizes food waste and will benefit society as a whole
(Tsiros & Heilman, 2005). This reputational effect further encourages retail stores to
participate in a digital platform and upload discounted perishables.
On the other hand, in the long term, retail stores bear the risk of eroding their pricing
power by using last-minute discounts on perishables. Research shows that when retail stores
use price discounts on a regular basis, consumers’ reference prices are lowered and they may
wait for prices to be reduced in the future, and, as a result, postpone their purchases (van
Heerde, Leeflang, & Wittink, 2000). However, Chung (2019) states that it is not clear whether
consumers strategize in their purchase of perishable products as they do when purchasing
seasonal goods such as fashion products. It is possible that consumers make a tradeoff
between price and quality (i.e., freshness) without behaving strategically.
Consumer-to-retail store effect. Consumers participate in the digital platform for two
reasons. First, consumers have the strong feeling that participating in digital platforms
dedicated to reducing food waste is morally correct. It can make consumers feel guilty for
wasting food (van Geffen et al., 2020). By using the digital platform to identify and
subsequently buy perishables on last-minute discount, consumers feel that they are helping in
the fight against food waste. Second, by buying perishables on last-minute discount,
consumers save money. Consumers’ use of the digital platform will increase sales of
perishables in retail stores which, in turn, will prompt these stores to be more active on the

14
platform by uploading more perishables on last-minute discount. We call this the consumer-
to-retail store effect. Although consumers foresee immediate moral and economic benefits of
using these digital platforms that focus on food waste, their usage may dwindle in the long
run. Literature finds that the reason for the decline in usage is due to the level of consumers’
cooking skills and the accuracy of planning skills. Consumers with developed cooking skills
who are able to prepare a wide variety of food can improvise based on the perishables
available for sale (Stangherlin, Duarte Ribeiro, & Barcellos, 2019), while consumers who lack
these skills are more rigid about the perishables they can use, thereby making them less
proactive in using digital platforms to reduce food waste. In a similar vein, consumers with
accurate planning skills are better in making decisions on the quantity to purchase, cook, and
consume, but experience less freedom in making spontaneous decisions concerning what to
eat (van Geffen et al., 2020), which makes these consumers less prone to last-minute
discounts.
Quantifying the retail store-to-consumer and consumer-to-retail store effects. We expect
that the retail store-to-consumer and consumer-to-retail store effects differ in strength, i.e.,
they are asymmetric because of the distinct features and objectives outlined above (cf. Cong
et al., 2019). Note that there are different mechanisms at work in the short and long term, for
both retail stores and consumers. To summarize, while, in the short term, grocery retail stores
use the digital platform as an avenue to reach consumers and thus recoup some of their costs
of food that otherwise would go to waste (or be sent to local food banks such as food pantries
and meal programs), in the long-term, they may hold back to some extent, to prevent
stimulating opportunistic and strategic buying behavior by consumers (e.g., delaying their
purchases when expecting a last-minute bargain to emerge) that may make them more price
sensitive. Consumers, on the other hand, want to save money and help the environment by
making use of the digital platform in the short term, but in the long term may be hesitant to

15
change their buying patterns (e.g., consumers need flexibility and creativity in their cooking
repertoire to fully benefit from the products offered on the digital platform).
Quantifying the relative importance of one side over the other of a two-sided network
provides digital platform leaders with the necessary information to allocate their available
resources more efficiently (Chu & Manchanda, 2016; Sridhar et al., 2011).
3. Methodology
3.1. Research context
To study the dynamics and (a)symmetry of CNEs we use data from a company in Europe
that built a digital platform in 2016 for reducing food waste. The firm managed to attract a set
of grocery retail stores and their consumers that interact on the digital platform to jointly
prevent waste of perishable food. The data pertain to retail store behavior (i.e., uploading of
last-minute discounts on the digital platform) and consumer behavior (i.e., viewing the
available last-minute discounts).
To participate in the digital platform, grocery retail stores subscribe for a year and pay a
fixed yearly fee. In addition, retail stores incur some transaction cost, because perishables
placed on last-minute discount are scanned by personnel at the retail store to post the items on
the digital platform. Consumers can download the mobile phone app and use the digital
platform for free. The app uses the consumers’ geolocation to indicate nearby retail stores and
the discounted perishables available. Consumers can (pre)select the retail stores they would
like to receive information from. The mobile app does not send push notifications (i.e., no
proactive messages are sent when new perishable products are added to the digital platform).

16
3.2. Data
Retail stores gradually joined the digital platform in the year following its launch in 2016,
as illustrated by Figure 1. We track the activity of retail stores for the first year since they
joined the digital platform. On the grocery retail store side, we observe the perishable
products that are uploaded by a retail store on the digital platform on a daily basis. Primarily
perishables from categories such as vegetables, fruits, meats, and dairy were placed under
last-minute discount. On the consumer side, we observe the day as well as the time at which a
consumer views the perishables offered on last-minute discount by his or her preferred retail
store(s). We have data on 48 grocery retail stores who uploaded 159,040 perishable products
in their first year (i.e., 52 weeks) of joining the digital platform, and which are viewed by
9,985 consumers. Each consumer’s list of preferred retail stores is also known. For the
analysis, the data were aggregated from day to week level in accordance with the fact that
grocery trips in the European country where the study is based, are not made every day.
To measure retail store activity on the digital platform, we calculate the number of distinct
products offered on last-minute discount by a retail store. To measure consumer activity on
the digital platform, we calculate the number of distinct consumers who viewed these last-
minute discounts in the same store. Table 2 provides some descriptive statistics.
[Insert Figure 1 and Table 2 about here]
Although the activity at the consumer side is measured by the number of consumers who
viewed the last-minute discounts on perishables and not the actual purchases that these
consumers made, it still serves as a good proxy for purchase behavior of the perishables
posted on the digital platform. As the mobile app did not send push notifications, a reasonable

17
assumption is that consumers only open the app just before or during their grocery shopping
journey. Previous research shows that the level of interest in mobile apps of retailers is indeed
directly associated with purchase intent (Taylor & Levin, 2014). Empirical support for this
assumption is offered by our data too. Figure 2 shows a plot confirming that consumers
generally opened and thus used the app during popular times for grocery shopping, i.e., from
9 am to 9 pm.
[Insert Figure 2 about here]
3.3. Econometric modeling
To analyze the data and study the dynamic interactions and feedback effects over time
between (i) the number of last-minute discounts uploaded on the digital platform by the retail
stores, and (ii) the number of distinct consumers – linked to these stores on the digital
platform – who viewed these discounts, we used a vector autoregressive (VAR) model. VAR
models address the modeling needs of our study as it controls for endogeneity, reverse
causality and feedback loops between variables (Granger & Newbold, 1986; Pauwels, 2004),
thereby better illustrating the cross-side network effects between the two sides of the digital
platform. VAR has been used to study cross-side network effects in two-sided markets in
different contexts such as digital platforms (Song et al., 2018) and crowdfunding campaigns
(Thies, Wessel, & Benlian, 2018), and is also an established modeling technique in marketing
research (e.g., Colicev et al., 2018; DeKimpe & Hanssens, 1995; Luo, Raithel, & Wiles, 2013;
Yang, See-To, & Papagiannidis, 2019).
As we have panel data which involve time-series information related to 48 grocery retail
stores, we use the panel vector autoregressive model (PVAR) to estimate the relationship

18
between the number of last-minute discounts uploaded by the retail stores and the number of
distinct consumers who view these discounts (Holtz-Eakin, Newey, & Rosen, 1988). PVAR
pools data across retail stores, while controlling for cross-sectional heterogeneity by using
store fixed effects (for similar practice, see Borah & Tellis, 2016; Colicev, Kumar, &
O’Connor, 2019; Hewett et al., 2016).
Before specifying our PVAR model, we need to conduct two important tests. First, we use
Granger causality tests (Granger, 1969) to establish the relationship between the number of
last-minute discounts uploaded and the number of distinct consumer views of these discounts.
Granger causality of variable Y by variable X implies that we can better predict Y by also
including the lagged value of X (along with the lagged values of Y) than by only including the
lagged value of Y. This is one of the closest causality tests that can be carried out with non-
experimental data (Srinivasan, Rutz, & Pauwels, 2016) and is also known as temporal
causality. Specifically, we adopt the Granger causality test for panel data proposed by
Dumitrescu and Hurlin (2012). This approach offers a test that takes into account
heterogeneity across retail stores when estimating the causal relationship between last-minute
discounts and distinct consumer views and provides an overall Granger causality statistic for
the entire data set (by averaging across retail stores). The null hypothesis for the test assumes
that the variables do not (Granger) cause each other. Our results show that the p-values are
significant for last-minute discounts causing distinct consumer views (p < 0.000) and distinct
consumer views causing last-minute discounts (p < 0.002),4 and thus we can reject the null
hypothesis. As the results indicate that last-minute discounts (Granger) cause distinct
consumer views, both these variables are endogenous.
Second, we use unit root tests to check whether our two key variables, i.e., number of last-
minute discounts and number of distinct consumer views, are stationary. This is a requirement
4 Reporting the minimum p-value across 15 lags.

19
for using PVAR. We use panel unit root tests which are unit root tests for multiple-series and
can be applied to panel data (see Baltagi 2013 for more details). These panel unit root tests
build upon classical tests performed on single time series, such as the Augmented Dickey-
Fuller (ADF) and Phillips-Perron tests. We use a range of panel root tests, specifically, Fisher-
type panel unit root tests as indicated by both Choi (2001) and Maddala and Wu (1999). We
also use the Levin-Lin-Chu (LLC) test (Levin, Lin, & Chu, 2002). These tests have been
frequently used by previous research in marketing (e.g., Colicev, Kumar, & O’Connor, 2019;
Kim & Hanssens, 2017; Kübler et al., 2018; Rego, Morgan, & Fornell, 2013). Table 3
includes the results of these tests. These results permit us to reject the unit root hypothesis for
last-minute discounts and distinct consumer views, and suggest that these variables can enter
the PVAR model in levels.
[Insert Table 3 about here]
Based on the results of the panel Granger causality tests and the unit root tests, the PVAR
model is specified as follows:
(1) 
= .
.

.
.


 
+,
,+ 1,
2,,
where LD = last-minute discounts, CV = distinct consumer views, i = 1,…S (= 48) retail
stores and t = 1,…T (= 52) weekly observations since a retail store i joined the platform.
The off-diagonal terms in the matrix ,
 (kl) represent the lagged effects of distinct
consumer views on last-minute discounts, and the lagged effects of last-minute discounts on
distinct consumer views. The diagonal elements in the matrix ,
 (k=l) represent the
autoregressive effects. Store fixed effects () are included to control for any time-invariant

20
store factors.5,6 Finally, the error terms () capture the contemporaneous effects between the
variables.
Next, we estimate equation (1) using Generalized Method of Moments (GMM). However,
since previous research has shown that fixed effects () are likely to be correlated with
regressors due to the lags of the dependent variable (Arellano & Bond, 1991; Arellano &
Bover, 1995; Blundell & Bond, 1998), we use forward orthogonal deviations (also known as
Helmert transformation) to eliminate the fixed effects. As the Helmert transformation only
removes the forward mean, that is the mean of all future observations for each store-month in
our data, the orthogonality between the transformed variables and the lagged regressors is
preserved, and thus we may use lagged instruments in our estimation (Arellano & Bover,
1995). Further, Helmert transformations do not induce autocorrelation in the error terms. To
estimate our PVAR model we use the panelvar package in R (Sigmund & Ferstl, 2019), which
has been used by a number of studies (e.g., Koengkan, Losekann, & Fuinhas, 2018; Liu &
Kim, 2018), and permits us to calculate robust standard errors.
To choose the optimal lag order for our PVAR model estimation using GMM, the
consistent moment and model selection criteria (MMSC) proposed by Andrews and Lu (2001)
are used. These criteria resemble common Maximum Likelihood based model selection
criteria such as Akaike Information Criterion (AIC) (Akaike, 1969), Bayesian Information
Criterion (BIC) (Schwarz, 1978), and the Hannan-Quinn Information Criterion (HQIC)
(Hannan & Quinn, 1979). Based on the optimal lags, we then check for multicollinearity
among the lagged variables by calculating the Variance Inflation Factor scores (VIF values).
If the VIF values are below 10, multicollinearity is not a concern (Yoder & Pettigrew-Crosby,
5 Week fixed effects are not included as the weeks in our model correspond to the weeks retail stores joined the
digital platform rather than calendar weeks.
6 By using the Hausman test on a simplified version of our PVAR model, we test whether random effects are
preferred to fixed effects in our context. The results indicate that the preferred model is fixed effects (rather than
random effects).

21
1995). Finally, we assess model stability by checking whether the eigenvalues of the
companion matrix are within the unit circle (Heij et al., 2004).
Our PVAR comprises a system of equations as the endogenous variables, i.e., number of
last-minute discounts uploaded by retail stores and number of distinct consumer views,
depend on each other. Because of this interdependency, the coefficient estimates only provide
limited information about the reaction of the system to a shock (i.e., a unit increase in one of
these variables). Hence, impulse response functions are calculated to provide a better
understanding of the model’s dynamic behavior. To exemplify, if a shock is given to distinct
consumer views this week and one would like to see its response on last-minute discounts, not
just for the same week but also for a specified number of weeks in the future, impulse
response functions are required. More generally, impulse response functions describe the
evolution of our focal variable along a specified time horizon, after a shock is given to the
system at a particular point in time. To calculate the impulse response functions, we recast our
PVAR into its vector moving-average form and then derive the impact of a one-unit shock of
one endogenous variable on the other. For an in-depth exposition, we direct the interested
reader to Hamilton (1994).
As the impulse response function permits us to trace the impact over time of one-unit
shock of an endogenous variable on other endogenous variables, we are able to examine the
short-term as well as the long-term effects of a unit shock of the number of last-minute
discounts uploaded by retail stores on the number of distinct consumers who view these
discounts and vice versa. We use the generalized impulse response functions (GIRFs) which
are robust to the causal ordering of variables (Pesaran & Shin, 1998). The standard errors are
calculated by bootstrapping, using cross-sectional resampling as prior research recommends
the use of this method for panel models such as PVAR (Kapetanios, 2008).

22
4. Results
4.1. PVAR model
The optimal lag length for the PVAR model according to all three MMSC criteria, i.e.,
MMSC-AIC, MMSC-BIC, and MMSC-HQIC, is one lag (see Table 4). The calculated VIF
values show that multicollinearity is not a concern because all the VIF values are around 2
(VIF for lagged last-minute discounts is 1.84 and for lagged distinct consumer views it is
2.12). Our stability check finds that the PVAR model is stable as the eigenvalues of the
companion matrix are .90 and .57, and are thus within the unit circle. As the parameter
estimates of the PVAR model are not interpretable (Sims, 1980; Song et al., 2018), the GIRFs
are used to estimate the effect of the variables.
[Insert Table 4 about here]
4.2. Generalized impulse response functions (GIRFs)
The GIRFs are used to examine how a one-unit shock to distinct consumer views affects
last-minute discounts and vice versa, and are depicted in Figure 3. Panel A displays the
consumer-to-retail store effect and, thus, shows how a one-unit shock to distinct consumer
views affects last-minute discounts over a 15-week period. Panel B displays the retail store-
to-consumer effect and, thus, shows how a one-unit shock to last-minute discounts affects
distinct consumer views over a 15-week period. The dotted line represents the 95%

23
confidence interval. Our results show that the consumer-to-retail store effect is stronger than
the retail store-to-consumer effect, but both these effects dwindle over time.
[Insert Figure 3 about here]
To better illustrate how the effects depicted by the GIRFs dwindle over time, the short-
term and long-term effects are reported in Table 5, where short-term effects refer to the
immediate effects (i.e., effects at week 1), while long-term effects are effects that persist over
a relatively long period of time (i.e., effects at week 15). In line with previous research using
VAR type models (e.g., DeKimpe & Hanssens, 1999; Pauwels & Weiss, 2008), we report
values of significant effects while assigning a value of 0 for non-significant effects. We find
that although the consumer-to-retail store effect dwindles over time, it continues to have a
relatively long-term impact. The retail store-to-consumer effect also decreases over time, but
dwindles quickly to zero. A possible explanation for a strong short-term consumer-to-retail
store effect may be that extra consumer interest leads to an important increase in sales of
discounted perishables on offer which stimulates retail stores to upload more eligible products
on the platform and to do this more promptly and systematically. Prior findings support this
conjecture (Taylor & Levin, 2014). The effect lasts longer since an increase in sales of a
product near its expiration date implies that other, alternative products - perishables with
longer shelf life - will remain unsold (substitution effect). These, in turn, may become
candidates for the next days’ last-minute discounts. Theoretically, it could precipitate a
gradual overall increase in the number of perishables, discounted last-minute, which may lead
to a relatively long-term consumer-to-retail store effect that we find.
[Insert Table 5 about here]

24
In comparison, the retail store-to-consumer effect is small in the short-term. To motivate
consumers to purchase a perishable at last-minute discount, retail stores need to increase the
variety of perishables. Consumers considering the extra discounted perishables are more
likely to buy different than similar items. Within a particular category, the quantity they can
absorb and use in a short period of time will be limited. This is further boosted by the fact that
the perishable products in a certain category act as substitutes or need to be integrated in a
meal, which means planning for or already having the additional meal components at home.
More creative and flexible (in cooking repertoire) consumers may be able to absorb more
products than more rigid counterparts. This may explain why we find the long-term retail
store-to-consumer effect being zero.
4.3. Robustness checks
We perform three different robustness checks. For these robustness checks, we report the
cumulative effect in Table 6 by taking the sum of significant GIRFs (Colicev, Kumar, &
O’Connor, 2019). First, we increase the number of lags to two to check whether it affects our
results. Second, we control for any variations arising due to the time since a retail store joined
the platform by incorporating time fixed effects (where time represents the week since joining
the digital platform). Third, we take the log transformation of both our variables (i.e.,
consumer views and last-minute discounts) before estimating the PVAR model. We find that,
for all the robustness checks, the consumer-to-retail store effect remains stronger than the
retail store-to-consumer effect.
[Insert Table 6 about here]

25
5. Discussion
Food waste is a huge problem. By 2030, the total cost of food waste could be as high as
$600 billion (Waste Wise Products, 2017), without factoring in environmental and social
consequences. The problem is expected to increase unless efforts are made to tackle it.
Grocery retailers can help reduce food waste by adopting strategies such as last-minute
discounting of perishables that are close to their expiration date. Digital platforms and
integrated mobile apps can help retail stores swiftly disseminate information on these
discounts to consumers.
This study used data from such a digital platform to study cross-side network effects using
a dynamic perspective. Drawing on two-sided market theory, we have advanced the literature
on digital platforms by empirically examining the asymmetry between the cross-side network
effects of the two sides of the market in the context of business-to-consumer industries (cf.
Chu & Manchanda, 2016; Song et al., 2018; Thies, Wessel, & Benlian, 2018; Voigt & Hinz,
2015). In our research setting, we have retail stores on one side and consumers on the other.
In addition, in the context of research on food sharing models (Michelini, Principato, &
Iasevoli, 2018), this is the first study – to the best of our knowledge - to quantify the relative
importance of one side over the other side of the network. This quantification is especially
relevant in our research context of food waste in which time sensitivity is high. Indeed, to
increase the effectiveness of last-minute discounts, retail stores and consumers need to be
connected within a short timespan (i.e., one or two days). Our study shows interesting results.
Particularly, we find evidence of asymmetric cross-side network effects in that the magnitude
of the effects, between both sides of the network, is different. Specifically, the consumer-to-
retail store effect is stronger and more long lasting than the retail store-to-consumer effect.

26
5.1. Theoretical implications
Our findings contribute to recent endeavors, primarily in the area of cross-side network
effects and food waste, to understand retail store-consumer interactions in the fight against
food waste. First, regarding the literature on two-sided markets and cross-side network
effects, our findings provide further evidence for the existence of asymmetric cross-side
network effects (e.g., Chu & Manchanda, 2016; Song et al., 2018; Thies, Wessel, & Benlian,
2018; Voigt & Hinz, 2015). Unlike existent studies, we focus on a digital platform that needs
to connect both sides of the market (i.e., retail stores and consumers) very quickly, i.e., within
one or two days, to develop and sustain CNEs. Our results provide evidence that dynamic
environments, in which time sensitivity is high, ask for knowledge on the relative importance
of one side over the other side of the market. Hence, our findings may apply to other retailers
as well, especially those that rely on (almost) simultaneous interactions of users on both sides
of the digital platform, e.g., retailers that focus on perishable products other than groceries
such as bakeries, restaurants, florists, and pharmacies, but also services that rely on sharing-
economy businesses such as Uber.
Second, while literature on food waste is extensive, food waste at the retail level is under-
researched (Cicatiello et al., 2017; Filimonau & Gherbin, 2017; Michelini, Principato, &
Iasevoli, 2018) and studies on the use of digital platforms to mitigate the food waste problem
are still largely lacking (for an exception see, e.g., Corbo & Fraticelli, 2015). By investigating
how digital platforms, focused on reducing food waste, can be sustained, we respond to
multiple calls related to food waste such as (i) better understanding managerial approaches to
reduce food waste (Filimonau & Gherbin, 2017; Stöckli, Niklaus, & Dorn, 2018) and (ii)
assessing the effectiveness of strategies against food waste (Cicatiello et al., 2017; Stöckli,

27
Niklaus, & Dorn, 2018). Specifically, our study shows that digital platforms are able to
effectively tie together retail stores and consumers within a short timespan (i.e., one or two
days) by levering the different CNEs, which, in turn, enhances the effectiveness of last-minute
discounts and reduces food waste. Additionally, we extend prior, often descriptive and
qualitative work in this domain (e.g., Corbo & Fraticelli, 2015; Halloran et al., 2014; Pirani &
Arafat, 2016) by empirically assessing digital solutions to the food waste problem using
secondary data.
Our findings suggest that digital platforms are able to effectively create ties between retail
stores and consumers in that consumer-to-retail store and retail store-to-consumer effects are
present but asymmetric. Specifically, the discounted offerings uploaded by retail stores trigger
consumers to use the digital platform to identify which offers to take advantage of.
Consumers thus value the option to save money and, possibly, also the thought of actively
helping to prevent food waste. However, this effect is shortly lived, which implies that, while
academic research and popular press urge retailers to take initiatives to tackle food waste
(e.g., Capgemini, 2017; Cicatiello et al., 2017; Winsight Grocery Business, 2018), firms
should find ways to keep consumers alert and active. Previous research already indicates that
the reasons for household food waste are complex and not yet fully understood (Lee, 2018),
which make it hard to develop successful food waste reduction strategies (Aschemann-Witzel
et al., 2018). Advancing the literature on food waste in grocery retailing, our results show that
the effectiveness of retail stores’ food waste reduction strategies are dependent on consumer
actions and behavior. Both digital platform leaders and retail stores should take this into
account. They may explore the use of push messages to trigger consumers to respond and
remain active, but also suggesting recipes in which the perishables on discount can be used
can trigger consumer activation. In line with the product transformation salience theory (e.g.,
Winterich, Nenkov, & Gonzales, 2019), we expect that providing recipes will lead consumers

28
to think how the discounted perishables could be used, thereby encouraging them to behave
sustainably by increasing their purchase of perishables on discount.
5.2. Managerial implications
Our findings provide evidence to retail stores that their initiative to provide last-minute
discounts on close to expiry products persuades consumers to use the digital platform, via its
mobile app, to view these products. However, since the consumer-to-retail store effect is
stronger than the retail store-to-consumer effect, it is advisable for the digital platform leader
to intervene on the consumer side of the market (cf. Chu & Manchanda, 2016). Based on
existing literature on mobile apps, we speculate that the digital platform leader should
improve the perceived compatibility and interactivity of their app. These characteristics have
a positive influence on affective involvement, which increases consumers’ intention to use
mobile apps (Kang, Mun, & Johnson, 2015). Perceived compatibility is the degree to which
consumers believe the digital platform fits with their needs and preferences (Kang, Mun, &
Johnson, 2015). Digital platform leaders will benefit from a detailed understanding of
consumers’ service needs. By providing recipes based on the perishables on discount, the
digital platform leader and retail store can foster compatibility and stimulate willingness to
use. It also adds to interactivity, because the provision of recipes can make using the digital
platform more appealing and interesting (Kang, Mun, & Johnson, 2015). Interactivity could
also be enhanced by offering consumers the option to create a digital grocery shopping list at
the beginning of their journey that includes the close to expiry date products on sale. Push
notifications may, as was mentioned, enhance perceived interactivity too. We encourage
scholars to empirically validate these theoretical presumptions in future research.

29
Finally, the results may hold important information for policy makers. They can help
subsidize parts of the two-sided market to help the digital platform leader to promote the
platform and enhance its interconnectivity. By stimulating the overall installed base and use,
the chances of such digital platforms to become sustainable will increase. Governments may
also consider changing regulations regarding the quantification and transparency of food
waste in order to provide retail stores more incentives to reduce the food waste problem.
Indeed, in order to reduce food waste, one needs to quantify the food waste problem but also
the outcome of food waste prevention strategies (such as last-minute discounting). For
instance, if retail stores are encouraged to quantify their food wastage and analyze its causes,
information is more readily available that can be used to take further measures to reduce food
waste. Moreover, transparency about the success rates of certain food waste reduction
strategies enables best practices to be shared more easily.
5.3. Limitations and future research directions
This study has several limitations, some of which provide worthwhile avenues for further
research. First, while this research focuses on the asymmetric cross-side network effects
between retail stores and consumers on a digital platform aimed at food waste reduction, we
do not empirically investigate which strategies the digital platform owner should adopt to
increase consumers’ and retail stores’ usage of the digital platform and its associated app. We
encourage researchers to investigate the effectiveness of different intervention strategies (e.g.,
providing recipes, enabling grocery shopping lists, push notifications), especially on the
consumer side of the market, in order to elicit a virtuous cycle of the mobile app’s use and to
reduce the food waste problem. In this regard, it may also worthwhile to investigate why the
consumer-to-retail store effect is stronger and lasts longer than the retail store-to-consumer

30
effect. Due to a lack of data, we only present theoretical reasons for this finding (e.g.,
substitution effect), but we encourage scholars to empirically test our presumptions.
Second, while literature shows that the level of interest in mobile apps of retailers is
directly associated with purchase intent (Taylor & Levin, 2014), providing us with confidence
that consumer views serve as a good proxy for actual purchase behavior of the perishables
uploaded on the digital platform, it might be that there is a gap between what consumers aim
to do (e.g., buying discounted perishables) and what they actually do (e.g., viewing the app
without making purchases) (cf. van Geffen et al., 2020). Due to the unavailability of data, we
were not able to measure actual purchase behavior and compare intentional and actual
consumer behavior. Future research might take up this issue and investigate whether
consumer views outweigh actual purchases or vice versa, and the implications for retail stores
and the objective of reducing food waste.
Third, an interesting avenue for further research is to examine the performance
implications for retail stores that actively participate on a digital platform to reduce food
waste. One can argue that active participation can have both a negative and positive impact,
which makes the performance implications ambivalent. On the negative side, using the digital
platform can make consumers more price sensitive as they may defer their purchase until
these last-minute discounts appear. On the positive side, apart from an uptake in sales of the
perishables on last-minute discount, retailers can expect a boost in store visits as well as
basket size.
Fourth, our sample only consists of retail stores and consumers who choose to participate
in the digital platform. While this sample suits the purposes of this study, it can imply that we
have a sample selection bias. Retail stores and consumers who are more concerned about food
waste are more likely to be active on the digital platform but are not necessarily representative
of, respectively, the retailing industry and consumer population. Future work can examine

31
what motivates retail stores and consumers to use the digital platform to reduce food waste. In
addition, it is also important to examine the motivation of existing users of the digital
platform. For example, does the self-interested economic reason of saving money or the
altruistic and moral reason of contributing to food waste reduction dominate in the decision to
use the digital platform (see e.g., Aschemann-Witzel, 2018)?
Finally, to tackle the increasing problem of food waste, different actors should initiate
multiple initiatives. While our focus was on one of these initiatives, i.e., providing last-minute
discounts on perishables close to their recommended expiry date, retail stores could take
multiple actions at the same time. Therefore, future research can explore the synergies
between different food waste reduction initiatives to see whether they complement or compete
with each other. Additionally, while we focus on initiatives taken by grocery retail stores,
other parties within the supply chain are also responsible for taking action. Future research
could take a supply chain perspective and investigate the interrelationships between food
waste reduction initiatives taken by different parties within the supply chain.

32
References
Adenso-Díaz, B., Lozano, S., & Palacio, A. (2017). Effects of dynamic pricing of perishable
products on revenue and waste. Applied Mathematical Modelling, 45, 148-164.
Akaike, H. (1969). Fitting autoregressive models for prediction. Annals of the institute of
Statistical Mathematics, 21(1), 243-247.
Andrews, D. W., & Lu, B. (2001). Consistent model and moment selection procedures for
GMM estimation with application to dynamic panel data models. Journal of
Econometrics, 101(1), 123-164.
Arellano, M., & Bond, S. (1991). Some tests of specification for panel data: Monte Carlo
evidence and an application to employment equations. The review of economic studies,
58(2), 277-297.
Arellano, M., & Bover, O. (1995). Another look at the instrumental variable estimation of
error-components models. Journal of econometrics, 68(1), 29-51.
Aschemann-Witzel, J., Otterbring, T., de Hooge, I. E., Normann, A., Rohm, H., Almli, V. L.,
& Oostindjer, M. (2019). The who, where and why of choosing suboptimal foods:
Consequences for tackling food waste in store. Journal of Cleaner Production, 236, 1-10.
Aschemann-Witzel, J. (2018). Consumer perception and preference for suboptimal food under
the emerging practice of expiration date based pricing in supermarkets. Food Quality and
Preference, 63, 119-128.
Aschemann-Witzel, J., de Hooge, I. E., Almli, V. L., & Oostindjer, M. (2018). Fine-tuning the
fight against food waste. Journal of Macromarketing, 38(2), 168-184.
Aschemann-Witzel, J., Jensen, J. H., Jensen, M. H., & Kulikovskaja, V. (2017). Consumer
behavior towards price-reduced suboptimal foods in the supermarket and the relation to
food waste in households. Appetite, 116, 246-258.

33
Baltagi, B. H. (2013). Econometric analysis of panel data. 5th ed. Chichester: John Wiley &
Sons Ltd.
Blundell, R., & Bond, S. (1998). Initial conditions and moment restrictions in dynamic panel
data models. Journal of econometrics, 87(1), 115-143.
Borah, A., & Tellis, G. J. (2016). Halo (spillover) effects in social media: do product recalls
of one brand hurt or help rival brands? Journal of Marketing Research, 53(2), 143-160.
Buisman, M. E., Haijema, R., & Bloemhof-Ruwaard, J. M. (2019). Discounting and dynamic
shelf life to reduce fresh food waste at retailers. International Journal of Production
Economics, 209, 274-284.
Capgemini (2017). The role of food retailers and technology in reducing consumer food
waste. Available at https://www.capgemini.com/wp-content/uploads/2017/10/food-waste-
management_web.pdf.
Choi, I. (2001). Unit root tests for panel data. Journal of International Money and Finance,
20(2), 249-272.
Chu, J., & Manchanda, P. (2016). Quantifying cross and direct network effects in online
consumer-to-consumer platforms. Marketing Science, 35(6), 870-893.
Chung, J. (2019). Effective pricing of perishables for a more sustainable retail food market.
Sustainability, 11(17), 4762-4777.
Cicatiello, C., Franco, S., Pancino, B., Blasi, E., & Falasconi, L. (2017). The dark side of
retail food waste: Evidences from in-store data. Resources, Conservation & Recycling,
125, 273-281.
Ciulli, F., Kolk, A., & Boe-Lillegraven, S. (2019). Circularity brokers: Digital platform
organizations and waste recovery in food supply chains. Journal of Business Ethics, 1-33.
Colicev, A., Malshe, A., Pauwels, K., & O’Connor, P. (2018). Improving consumer mindset
metrics and shareholder value through social media: The different roles of owned and

34
earned media. Journal of Marketing, 82(1), 37-56.
Colicev, A., Kumar, A., & O’Connor, P. (2019). Modeling the relationship between firm and
user generated content and the stages of the marketing funnel. International Journal of
Research in Marketing, 36(1), 100-116.
Cong, L. W., Miao, Q., Tang, K., & Xie, D. (2019). Survival scale: Marketplace lending and
asymmetric network effects. Available at SSRN: https://ssrn.com/abstract=3461893 or
http://dx.doi.org/10.2139/ssrn.3461893.
Corbo, C., & Fraticelli, F. (2015). The use of web-based technology as an emerging option for
food waste reduction. In Envisioning a future without food waste and food poverty:
Societal challenges. pp. 1-2. Wageningen Academic Publishers.
De Hooge, I. E., Oostondjer, M., Aschemann-Witzel, J., Normann, A., Mueller Loose, S., &
Lengard Almli, V. (2017). This apple is too ugly for me! Consumer preferences for
suboptimal food products in the supermarket and at home. Food Quality and Preference,
56, 80-92.
Dekimpe, M. G., & Hanssens, D. M. (1995). The persistence of marketing effects on sales.
Marketing science, 14(1), 1-21.
Dekimpe, M. G., & Hanssens, D. M. (1999). Sustained spending and persistent response: A
new look at long-term marketing profitability. Journal of Marketing Research, 36(4), 397-
412.
Donselaar, K. H., Peters, J., de Jong, A., & Broekmeulen, R. A. C. M. (2019). Analysis and
forecasting of demand during promotions for perishable items. International Journal of
Production Economics, 172, 65-75.
Dumitrescu, E. I., & Hurlin, C. (2012). Testing for Granger non-causality in heterogeneous
panels. Economic modelling, 29(4), 1450-1460.
Eisenmann, T. R., Parker, G. G., & Van Alstyne, M. W. (2006). Strategies for two-sided

35
markets. Harvard Business Review, 84(10), 92-101.
Evans, D. S., & Schmalensee, R. (2010). Failure to launch: Critical mass in platform
businesses. Available at https://ssrn.com/abstract=1353502 or http://dx.doi.org/10.2139/
ssrn.1353502.
FAO (2017). Key facts on food loss and waste you should know! Available at
http://www.fao.org/save-food/resources/keyfindings/en/.
Filimonau, V., & Gherbin, A. (2017). An exploratory study of food waste management
practices in the UK grocery retail sector. Journal of Cleaner Production, 167, 1184-1194.
Frishammar, J., Cenamor J., Cavalli-Björkman, H., Hernell, E., & Carlsson, J. (2018). Digital
strategies for two-sided markets: A case study of shopping malls. Decision Support
Systems, 108, 34-44.
Gauri, D. K., Ratchford, B., Pancras, J., & Talukdar, D. (2017). An empirical analysis of the
impact of promotional discounts on store performance. Journal of Retailing, 93(3), 283-
303.
Granger, C. W. J. (1969). Investigating causal relations by econometric models and cross-
spectral methods. Econometrica, 37(3), 424-438.
Granger, C. W. J., & Newbold P. (1986). Forecasting economic time series. San Diego:
Academic Press.
Halloran, A., Clement, J., Kornum, N., Bucatariu, C., & Magid, J. (2014). Addressing food
waste reduction in Denmark. Food Policy, 49, 294-301.
Hamilton, J. D. (1994). Time series analysis. New Jersey: Princeton University Press.
Hannan, E. J., & Quinn, B. G. (1979). The determination of the order of an autoregression.
Journal of the Royal Statistical Society: Series B (Methodological), 41(2), 190-195.
Heij, C., de Boer, P., Franses, P. H., Kloek, T., & van Dijk, H. K. (2004). Econometric
methods with applications in business and economics. Oxford: Oxford University Press.

36
Helfat, C. E., & Raubitschek, R. S. (2018). Dynamic and integrative capabilities for profiting
from innovation in digital platform-based ecosystems. Research Policy, 47(8), 1391-1399.
Hewett, K., Rand, W., Rust, R. T., & Van Heerde, H. J. (2016). Brand buzz in the echoverse.
Journal of Marketing, 80(3), 1-24.
Holtz-Eakin, D., Newey, W., & Rosen, H. S. (1988). Estimating vector autoregressions with
panel data. Econometrica: Journal of the Econometric Society, 1371-1395.
Kang, J. Y. M., Mun, J. M., & Johnson, K. K. (2015). In-store mobile usage: Downloading
and usage intention toward mobile location-based retail apps. Computers in Human
Behavior, 46, 210-217.
Kapetanios, G. (2008). A bootstrap procedure for panel data sets with many cross-sectional
units. The Econometrics Journal, 11(2), 377–395.
Kim, H., & Hanssens, D. M. (2017). Advertising and word-of-mouth effects on pre-launch
consumer interest and initial sales of experience products. Journal of Interactive
Marketing, 37, 57-74.
Koengkan, M., Losekann, L. D., & Fuinhas, J. A. (2018). The relationship between economic
growth, consumption of energy, and environmental degradation: renewed evidence from
Andean community nations. Environment Systems and Decisions, 1-13.
Kouris, I., & Kleer, R. (2012). Business models in two-sided markets: An assessment of
strategies for app plarforms. Available at https://aisel.aisnet.org/cgi/viewcontent.cgi?
article=1002&context=icmb2012.
Kübler, R., Pauwels, K., Yildirim, G., & Fandrich, T. (2018). App popularity: Where in the
world are consumers most sensitive to price and user ratings? Journal of Marketing, 82(5),
20-44.
Landsman, V., & Stremersch, S. (2011). Multihoming in two-sided markets: An empirical
inquiry in the video game console industry. Journal of Marketing, 75, 39-54.

37
Le Borgne, G., Sirieix, L., & Costa, S. (2018). Perceived probability of food waste: Influence
on consumer attitudes towards and choice of sales promotions. Journal of Retailing and
Consumer Services, 42, 11-21.
Lee, K. C. L. (2018). Grocery shopping, food waste, and the retail landscape of cities: The
case of Seoul. Journal of Cleaner Production, 172, 325-334.
Levin, A., Lin, C.-F., Chu, C.-S. J. (2002). Unit root tests in panel data: asymptotic and finite-
sample properties. Journal of Econometrics, 108(1), 1-24.
Little, M., & Castella, T. (2017). Tesco’s operations in the United Kingdom: Food waste in
stores and depots. Available at http://flwprotocol.org/case-studies/tescos-operations-
united-kingdom-food-waste-stores-depots/.
Liu, H., & Kim, H. (2018). Ecological Footprint, Foreign Direct Investment, and Gross
Domestic Production: Evidence of Belt & Road Initiative Countries. Sustainability,
10(10), 3527.
Luo, X., Raithel, S., & Wiles, M. A. (2013). The impact of brand rating dispersion on firm
value. Journal of Marketing Research, 50(3), 399-415.
Maddala, G. S., & Wu, S. (1999). A comparative study of unit root tests with panel data and a
new simple test. Oxford Bulletin of Economics and statistics, 61(S1), 631-652.
Michelini, L., Principato, L., & Iasevoli, G. (2018). Understanding food sharing models to
tackle sustainability challenges. Ecological Economics, 145, 205-217.
Muzellec, L., Ronteau, S., & Lambkin, M. (2015). Two-sided Internet platforms: A business
model lifecycle perspective. Industrial Marketing Management, 45, 139-150.
Nijs, V. R., Dekimpe, M. G., Steenkamp, J. B. E. M., & Hanssens, D. M. (2001). The
category-demand effects of price promotions. Marketing Science, 20(1), 1-22.
Parfitt, J., Barthel, M., & Macnaughton, S. (2010). Food waste within food supply chains:
Quantification and potential for change to 2050. Philosophical Transactions of the Royal

38
Society B: Biological Sciences, 365(1554), 3065-3081.
Pauwels, K. (2004). How dynamic consumer response, competitor response, company
support, and company inertia shape long-term marketing effectiveness. Marketing
Science, 23(4), 596-610.
Pauwels, K., & Weiss, A. (2008). Moving from free to fee: How online firms market to
change their business model successfully. Journal of Marketing, 72(3), 14-31.
Pesaran, H. H., & Shin, Y. (1998). Generalized impulse response analysis in linear
multivariate models. Economics Letters, 58(1), 17-29.
Pirani, S. I., & Arafat, H. A. (2016). Reduction of food waste generation in the hospitality
industry. Journal of Cleaner Production, 132, 129-145.
Rangaswamy, A., Moch, N., Felten, C., van Bruggen, G., Wieringa, J. E., & Wirtz, J. (2020).
The role of marketing in digital business platforms. Journal of Interactive Marketing,
Forthcoming.
ReFED (2018). ReFED analysis reveals food waste represents $18.2 billion profit opportunity
for grocery retailers. Available at https://www.refed.com/content-hub/refed-analysis-
reveals-food-waste-represents-18-2-billion-profit-opportunity-for-grocery-retailers.
Rego, L. L., Morgan, N. A., & Fornell, C. (2013). Reexamining the market share–customer
satisfaction relationship. Journal of Marketing, 77(5), 1-20.
Robertson, K., & Parfitt, J. (2018). Guidance for retailers: Why & how to measure food
waste. Available at https://flwprotocol.org/wp-content/uploads/2018/04/FLW-
Standard_Retailer-Guidance_2018-May-16.pdf.
Rysman, M. (2009). The economics of two-sided markets. Journal of economic perspectives,
23(3), 125-43.
Schanes, K., Dobernig, K., & Gözet, B. (2018). Food waste matters – A systematic review of
household food waste practices and their policy implications. Journal of Cleaner

39
Production, 182, 978-991.
Schwarz, G. (1978). Estimating the dimension of a model. The annals of statistics, 6(2), 461-
464.
Sigmund, M., & Ferstl, R. (2019). Panel Vector Autoregression in R with the Package
panelvar. The Quarterly Review of Economics and Finance.
Sims, C. A. (1980). Macroeconomics and reality. Econometrica: Journal of the Econometric
Society, 1-48.
Song, P., Xue, L., Rai, A., & Zhang, C. (2018). The ecosystem of platform: A study of
asymmetric cross-side network effects and platform governance. MIS Quarterly, 42(1),
121-142.
Sridhar, S., Mantrala, M. K., Naik, P. A., & Thorson, E. (2011). Dynamic marketing
budgeting for platform firms: Theory, evidence, and application. Journal of Marketing
Research, 48(6), 929–943.
Srinivasan, S., Rutz, O. J., & Pauwels, K. (2016). Paths to and off purchase: quantifying the
impact of traditional marketing and online consumer activity. Journal of the Academy of
Marketing Science, 44(4), 440-453.
Sriram, S., Manchanda, P., Bravo, M. E., Chu, J., Ma, L., Song, M., Shriver, S., &
Subramanian, U. (2015). Platforms: A multiplicity of research opportunities. Marketing
Letters, 26, 141-152.
Stangherlin, I., Duarte Ribeiro, J., & Barcellos, M. (2019). Consumer behavior towards
suboptimal food products: A strategy for food waste reduction. British Food Journal,
121(10), 2396-2412.
Stöckli, S., Niklaus, E., & Dorn, M. (2018). Call for testing interventions to prevent consumer
food waste. Resources, Conservation & Recycling, 136, 445-462.
Stummer, C., Kundisch, D., & Decker, R. (2018). Platform launch strategies. Business &

40
Information Systems Engineering, 60(2), 167-173.
Taylor, D. G., & Levin, M. (2014). Predicting mobile app usage for purchasing and
information-sharing. International Journal of Retail & Distribution Management, 42(8),
759-774.
Theotokis, A., Pramatari, K., & Tsiros, M. (2012). Effects of expiration date-based pricing on
brand image perceptions. Journal of Retailing, 88(1), 72-87.
Thies, F., Wessel, M., & Benlian, A. (2018). Network effects on crowdfunding platforms:
exploring the implications of relaxing input control. Information Systems Journal, 28(6),
1239-1262.
Tsiros, M., & Heilman, C. M. (2005). The effect of expiration dates and perceived risk on
purchasing behavior in grocery store perishable categories. Journal of Marketing, 69(2),
114-129.
Van Geffen, L., van Herpen, E., Sijtsema, S., & van Trijp, H. (2020). Food waste as the
consequence of competing motivations, lack of opportunities, and insufficient abilities.
Resources, Conservation & Recycling: X, 5, [100026].
Van Heerde, H. J., Leeflang, P. S. H., & Wittink, D. R. (2000). The estimation of pre- and
postpromotion dips with store-level scanner data. Journal of Marketing Research, 37(3),
383-395.
Voigt, S., & Hinz, O. (2015). Network effects in two-sided markets: why a 50/50 user split is
not necessarily revenue optimal. Business Research, 8(1), 139-170.
Wang, X., & Li, D. (2012). A dynamic product quality evaluation based pricing model for
perishable food supply chains. Omega, 40, 906-917.
Waste Wise Products (2017). Food waste: An economic and environmental problem.
Available at https://www.wastewiseproductsinc.com/blog/food-waste/food-waste-an-
economic-and-environmental-problem/.

41
Winsight Grocery Business (2018). Food waste action guide could help grocers go green.
Available at https://www.winsightgrocerybusiness.com/retailers/food-waste-action-guide-
could-help-grocers-go-green.
Winterich, K. P., Nenkov, G. Y., & Gonzales, G. E. (2019). Knowing what it makes: How
product transformation salience increases recycling. Journal of Marketing, 83(4), 21-37.
Yang, Y., See-To, E. W., & Papagiannidis, S. (2019). You have not been archiving emails for
no reason! Using big data analytics to cluster B2B interest in products and services and
link clusters to financial performance. Industrial Marketing Management.
Yoder, B. J., & Pettigrew-Crosby, R. E. (1995). Predicting nitrogen and chlorophyll content
and concentrations from reflectance spectra (400–2500 nm) at leaf and canopy scales.
Remote sensing of environment, 53(3), 199-211.

42
Table 1
Illustrative examples of food waste reduction platforms.
Platform name
Country
Main focus
11th Hour
Singapore
Meals close to expiry date in restaurants
Chowberry
Nigeria
Products close to expiry date in supermarkets
Food Cloud
Ireland
Products which are past their expiry date in
supermarkets
Food for all
United States
Meals close to expiry date in restaurants
Food Loop
Germany
Products close to expiry date in supermarkets
Go Mkt
United States
Meals and products close to expiry date in
local restaurants
Leloca
United States
Products for which there is a food surplus
Last Minute Sotto
Casa
Italy
Products close to expiry date in supermarkets
and food retailers
MyFoody
Italy
Products close to expiry date in supermarkets
NoFoodWasted
The Netherlands
Products close to expiry date in supermarkets
OptiMiam
France
Food and meals close to expiry date at
bakeries and in restaurants
Swipe Shark
Denmark
Products close to expiry date in supermarkets
Too Good To Go
Belgium, Denmark,
France, Germany, the
Netherlands, Norway,
Spain, Switzerland,
United Kingdom
Meals and products close to expiry date in
restaurants and supermarkets
Zéro-Gâchis
France
Products close to expiry date in supermarkets

43
Table 2
Descriptive statistics.
Variables
Mean
SD
Min
Max
Number of last-minute
discounts
63.718
40.325
10
161
Number of distinct
consumer views
17.072
12.618
3
60
Note: Descriptive statistics are based on activity on the digital platform related to 48 retail stores. The data
pertains to the activity of the first 52 weeks after these stores joined the platform.

44
Table 3
Panel unit root tests.
Panel unit root test
Levin-Lin-Chu
(2002)
ADF-Fisher based
test by Choi (2001)
ADF-Fisher based
test by Maddala and
Wu (1999)
Distinct consumer
views
.000
.000
.000
Last-minute
discounts
.000
.000
.000
Notes: Null hypotheses assumes unit root.
No intercept or trend is included in the different tests

45
Table 4
Model and moment selection criteria (MMSC) for lag selection of PVAR.
Lag
MMSC-BIC
MMSC-AIC
MMSC-HQIC
1
-1495
-350
-807
2
-1477
-348
-799
3
-1464
-350
-796
4
-1447
-349
-788
5
-1431
-348
-782
6
-1421
-354
-783
7
-1406
-354
-777
8
-1383
-347
-764
Note: MMSC-BIC, MMSC-AIC and MMSC-HQIC are model and moment selection criteria that resemble the
common Maximum Likelihood based criteria Bayesian Information Criterion (BIC), Akaike Information
Criterion (AIC) and Hannan-Quinn Information Criterion (HQIC).

46
Table 5
Short and long-term cross-side network effects.
Short-term effect
Long-term effect
Retail store to consumer effect
1.29
0
Consumer to retail store effect
8.06
1.10
Note: All non-zero effects are significant (p < .05). Insignificant effects are reported as 0.

47
Table 6
Cross-side network effects: Robustness to alternative specifications.
Robustness check
Consumer to
retail store effect
(Effect 1)
Retail store to
consumer effect
(Effect 2)
Lag 2
19.4+
1.16+
Time fixed effects
6.61+
1.14+
Log transformation of variables
0.63+
0.33+
Notes: We report the cumulative effects which are calculated by taking the sum of significant GIRF values.
+ p < .10

48
Figure 1
Distribution of grocery retail stores joining the digital platform over time.

49
Figure 2
Distribution of last-minute discount views by consumers over time.

50
Figure 3
Generalized impulse response function of CNE.
Note: Dashed lines represent 95% confidence interval.