2022, Tsai, Deli Fee Is Less Considered Than Deli Perfor

Received: 1 November 2021 Revised: 19 April 2022 Accepted: 20 April 2022
DOI: 10.1111/ijcs.12816
ORIGINAL ARTICLE
Which food delivery platforms are winning the restaurant food

delivery wars? Analysis from a consumer perspective
Pei-Hsuan Tsai1 | Wei-Hung Hsiao2 | Chih-Jou Chen1
1
Department of Marketing and Logistics
Management, National Penghu University of Abstract
Science and Technology, Magong City, Taiwan The emergence of the COVID-19 pandemic dramatically lowered the foodservice
2
Department of Distribution Management,
industry's income overnight. Conversely, the practical measure of remaining at home
National Taichung University of Science and
Technology, Taichung City, Taiwan to deal with the pandemic's impact has boosted the online food delivery business. In
this study, a consumer perspective was adopted and an adapted version of the
Correspondence
Pei-Hsuan Tsai, Department of Marketing and extended unified theory of acceptance and use of technology (UTAUT2) was used
Logistics Management, National Penghu
alongside multi-attribute decision-making methods (DEMATEL, DANP and modified
University of Science and Technology,
No. 300, Liuhe Road, Magong City 880011, VIKOR) to construct a model for evaluating and selecting a food delivery platform
Penghu County, Taiwan.
(FDP). The results of the INRM (influential network relation map) revealed that the
Email: pei.hsuan0616@gmail.com;
peihsuan0616@gms.npu.edu.tw first dimension to be improved upon and adjusted should be security, followed by
effort expectancy, performance expectancy, social influence, facilitating conditions,
Funding information
Ministry of Science and Technology (MOST) of hedonic motivation and habit. The DANP influential weights suggested that habits
Taiwan, Grant/Award Number: MOST
were the most important dimension, followed by hedonic motivation, while perfor-
109-2410-H-346-001
mance expectancy was the least important. According to the results of the gap analy-
sis, the first dimension that required improvement was performance expectancy,
followed by effort expectancy, facilitating conditions, security, social influence, habits
and hedonic motivation. It is expected that the findings of this study can serve as a
reference for consumers selecting FDPs to better satisfy their dining needs. The
novel model is discussed in terms of theoretical, practical and managerial
implications.
KEYWORDS
consumer preferences, DANP, DEMATEL, food delivery platform, modified UTAUT2 model,
modified VIKOR
1 | I N T RO DU CT I O N risen to become a distinctive dining option for consumers (Cheng

et al., 2021). With the services provided by FDPs, consumers can eas-
The outbreak of the COVID-19 pandemic drastically reduced the pri- ily have their favourite foods delivered to their homes. Due to their
mary income of the foodservice industry overnight. In addition to rely- ease of operation and convenience, these food delivery services have
ing on take-out and delivery services to compensate for such negative fundamentally altered consumer behaviour and daily consumption
impacts, businesses also explored potential options that could ‘reduce habits (Cho et al., 2019; Kumar & Shah, 2021).
their distance from the consumer dining table at home’ (Kim The global delivery market has seen significant growth since the
et al., 2021; Lock, 2020; Neise et al., 2021). outbreak of COVID-19, and the experts predict that this trend is likely
From 2020, the practical use of the measure of staying at home to continue. It has been estimated that by 2025, the global online
to cope with the impact of the pandemic has accelerated the online food delivery sector will grow to USD 200 billion, with a compound
food delivery sector. In Taiwan, food delivery platforms (FDPs) have annual growth rate (CAGR) of 10.3%. The Asia-Pacific region accounts
Int J Consum Stud. 2023;47:155–176. wileyonlinelibrary.com/journal/ijcs © 2022 John Wiley & Sons Ltd. 155
14706431, 2023, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/ijcs.12816 by National Cheng Kung University, Wiley Online Library on [31/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
156 TSAI ET AL.
for more than 50% of the total global food delivery market and is the theoretical gap in the research area. To fill this gap, we further explore
fastest-growing region for the sector in the world. Taiwan is a key the key factors that influence consumer selection in FDPs and
market for Foodpanda and Uber Eats, with expected revenue of $1 uncover the preferences of potential consumers. This study combined
billion generated each year. The combined daily orders of the two the extended UTAUT2 model and MADM methods to evaluate fac-
platforms have reached nearly 600,000 in Taiwan. tors that affect consumer choice, and compared and contrasted the
The COVID-19 pandemic has also had a considerable impact on primary findings with those of previous research (Huang &
the online food delivery sector over the last 2 years. The reduction of Kao, 2015).
on-site services caused by the pandemic has led to the bankruptcy of The main contributes of this study are fourfold: a theoretical basis
countless restaurants, while consumers have grown unprecedentedly was first established regarding previous research; consumer inter-
more dependent on food delivery services. In the second quarter of views on use experience were conducted to identify the key factors
2020 alone, downloads of food delivery service apps increased three- that influence consumer FDP choices; the decision-making trial and
fold. After Taiwan implemented a three-level national epidemic alert evaluation laboratory (DEMATEL) modelling approach was adopted to
system in May 2021, daily orders on FDPs increased by more than uncover the causal relationships between the evaluation dimensions
40%. It is expected that such growth will continue as the pandemic and criteria; and a DEMATEL-based analytic network process (DANP)
fundamentally changes the consumer use of food delivery services. was used to calculate the weights of the evaluation dimensions and
Statistically, the growth and demand for food delivery services in criteria, and the modified Vlse Kriterijumska Optimizacija
Taiwan are considerable. Instead of emerging as a short-term phe- Kompromisno Resenje (modified VIKOR) was employed to conduct a
nomenon, food delivery continues to thrive. gap analysis. The findings were expected to serve as a practical refer-
Recent studies on FDPs have tended to utilize statistical analysis ence for consumers to better identify which FDPs satisfy their needs.
(Fakfare, 2021; Hwang et al., 2021; Mehrolia et al., 2021; Wang The research questions of the study are as follows: (1) What are
et al., 2019) and qualitative research (Farah et al., 2021), while few the key factors that influence consumer FDP choice? (2) Could a
have made use of multi-attribute decision-making (MADM) methods, model be developed that determines the causal relations and impacts
forming a methodological gap in the research area. Unlike classic sta- of these factors? (3) How would the influential weights of factors in
tistical approaches, which require a minimum of 200 samples to vali- such a model be calculated? (4) How can FDPs formulate improve-
date the causal association between variables (Hoe, 2008), MADM ment strategies to influence consumer choice?
methods only require 10 samples (Tsai, 2020; Tsai & Chen, 2021) to
produce useful research (Tsai & Chen, 2021). The average sample size
ranged between 10 and 30 specialists in relevant studies who used 2 | LI T E RA T U R E RE V I E W
MADM methods (Horng et al., 2018). This study follows the big sam-
ple principle (n > 30) based on consumers' perspective. Therefore, we In this section, the theories adopted in the study are explained and a
sampled 418 consumers (Tsai et al., 2020). Selecting an FDP involves comprehensive and in-depth review of past literature was conducted
a decision-making process taking into consideration multiple attri- to summarize emerging trends in the evaluation methodology used in
butes. To ensure accurate investigation and analysis, this study modelling consumer FDP preferences. The definition of FDPs,
attempted to apply MADM methods to resolve the implied challenges. corresponding theories and studies are herein subsequently explored.
MADM has been rigorously developed (Mardani et al., 2015; Sipahi &
Timor, 2010) and increasingly applied to the study of consumer pref-
erences (Kwak et al., 2005; Lin et al., 2016), to ensure that modelled 2.1 | Food delivery platform industry
processes are aligned with reality.
In addition, past research has focused on the development and Due to the impact of the COVID-19 pandemic, FDPs have become an
stability of mobile catering service apps (Wang et al., 2019), consumer important dining option for consumers worldwide. With the services
behaviour and loyalty in using online food delivery services provided by various mobile apps, consumers can easily have their
(Suhartanto et al., 2019), and consumer purchase and use intentions favourite food delivered to them at home. Due to its ease and conve-
of food delivery service apps (Ramos, 2022; Roh & Park, 2019; nience, food delivery services have shifted diners' consumer behav-
Tandon et al., 2021). Current theories being applied in research iour and daily consumption habits.
include the theory of consumption values (TCV) (Tandon et al., 2021), According to Statista (2019), the market size of the global online
uses and gratifications theory (UGT) (Ray et al., 2019), the expectation food delivery sector has exceeded USD 100 billion with a market size
confirmation model (ECM), the task-technology fit (TTF) model with of more than 970 million consumers. It is expected that the growth
the unified theory of acceptance and use of technology (UTAUT) rate of the industry will reach 10% over the next 5 years and new
model (Zhao & Bacao, 2020), unified theory of acceptance and use of players continue to enter the market as it grows. Consequently, FDP
technology 2 (UTAUT2) (Ramos, 2022), the technology acceptance apps have gained cross-national and cross-cultural recognition and
model (TAM) and the attention-interest-desire-action (AIDA) model acceptance and have greatly influenced and changed consumer habits
(Song et al., 2021). There is a lack of studies exploring the extended and industry structure. Furthermore, according to a report by
UTAUT2 model with MADM methods to analyse FDPs, forming a MoneyDJ, the annual turnover of Taiwan's catering market in 2018
TSAI ET AL. 157
grew to approximately NT$ 470 billion, with a steady annual growth Kaohsiung, and it boasts a network of more than 8000 restaurants.
rate. However, according to estimates from relevant research institu- The Uber Eats market share has grown rapidly in the following entry.
tions, the food delivery market only accounted for approximately 5% Raj Beri, the general manager of Uber Eats Asia Pacific, Europe, Mid-
of the catering market overall; hence, future growth potential is signif- dle East and Africa, stated that ‘Uber Eats is changing the way people
icant. Currently, the largest players in Taiwan include Foodpanda, eat’, implying that the company helps consumers to use their mobile
Honestbee, Foodomo and Uber Eats, all of which entered the market devices to make food orders and receive their food swiftly. On
early. A summary of the market leaders is exhibited in Table 1. 11 September 2019, Uber Eats exceeded 1 billion deliveries world-
The German FDP Foodpanda has, till date, been operating in wide, echoing the growth of the global food delivery market. The
Taiwan for 7 years. Throughout 2019, the number of orders, users, development and performance of small- and medium-sized restau-
and restaurants in its network grew substantially. Foodpanda, under rants have also played an important role in market growth. It can be
its brand name as well as that of Delivery Hero, operates in more than expected that food delivery will become an indispensable part of daily
200 cities and eight nations in Asia, and is the world's largest deli food dining.
delivery group. From its entry into the Taiwan market in 2012, Although Taiwan's food delivery market is worth approximately
Foodpanda has enlisted tens of thousands of restaurants with delivery NT$ 27 billion, competition has become increasingly fierce in recent
services covering breakfast to midnight snacks. Compared to 2018, years. The increasing demands of consumers and competition from
Foodpanda's daily orders have increased by 25 times, its network of new players have forced platforms with poor management out of
restaurants has increased twofold, and the number of active users has business; Honestbee and Foodtoall are two examples. Therefore,
increased twentyfold. After expanding to the cities of Changhua and developing methods to survive the intense competitive rivalry and
Chiayi in August 2020, Foodpanda's services covered 11 cities (Taipei, attracting consumers to increase revenue and market share and main-
New Taipei, Keelung, Hsinchu, Miaoli, Taichung, Changhua, Chiayi, tain sustainable operations is a major concern for FDPs, while defining
Tainan and Kaohsiung) with plans to further expand coverage soon. methods of choosing between the increasing number of FDPs has
Among the cities covered, New Taipei City accounts for the most sub- become a key concern for consumers. It is for those reasons that the
stantial proportion of its total orders (30%), followed by Taipei (25%) present research has set out to determine how FDP business models
and Taichung (15%). Foodpanda's future development strategies are can improve and consumer choice can be supported.
‘continuously improving operational efficiency’ and ‘actively expan-
ding the presence in Taiwan’. It also upgraded its AI algorithm to
increase the accuracy of prediction of meal preparation and delivery 2.2 | UTAUT2 model
times, as well as providing the optimal allocation of delivery personnel
and routes, accounting for the delivery person's historical speed, effi- The UTAUT model was proposed by Venkatesh et al. (2003). It inte-
ciency, and familiarity of the location and real-time traffic conditions, grates rational choice theory and the technology acceptance model
maximizing delivery efficiency and minimizing consumer waiting time. (TAM) proposed by previous literature (Venkatesh et al., 2003). Spe-
Additionally, the company launched the option for consumers to col- cifically, the model integrates eight research theories, and it has been
lect their meals, which allows consumers to order food online and pick empirically validated to explain more than 70% of technology use
it up at the desired outlet, with no extra delivery fees. behaviour, making it more effective than any previous theory or
Uber Eats has a global presence in 83 cities and officially entered model alone (Tamilmani et al., 2021). The theories employed and the
the Taiwan market at the end of 2016. Its service offering covers structure of the model are listed in Table 2.
10 counties and cities, including Keelung, Taipei, New Taipei,
Taoyuan, Hsinchu, Taichung, Changhua, Chiayi, Tainan, and
TABLE 2 The UTAUT framework and the eight theories included

TABLE 1 Major FDPs in Taiwan
Years Authors Theoretical model
Registered Year # of 1962 Rogers Innovation Diffusion Theory
Platform country/ entering restaurants (IDT)
operator founding year Taiwan cooperated
1975 Fishbein and Ajzen Theory of Reasoned Action (TRA)
Foodpanda Germany/2012 2012 10,000+
1985 Ajzen Theory of Planned Behaviour
Foodomo Taiwan/2012 2015 800+ (TPB)
Uber Eats USA/2014 2016 8000+ 1986 Bandura Social of Cognitive Theory (SCT)
Yo-Woo Taiwan/2014 2016 1000+ 1986/ Davis/Venkatesh Technology Acceptance Model
Deliveroo UK/2013 2018 900 2000 and Davis (TAM/TAM2)
Honestbee Singapore/2012 2015 3000+ 1991 Thompson et al. Model of PC Utilization (MPCU)
Foodtoall Taiwan/2017 2018 1000+ 1992 Davis et al. Motivational Model(MM)

1995 Taylor and Todd Combined TAM and TPB (C-
Note: Honestbee ended its operations on 1 July 2019 and Foodtoall ended
TAM-TPB)
its operations on 15 August 2019.
158 TSAI ET AL.
The UTAUT model considers the decisive effect of four factors— When the motives for first use and increased frequency are examined,
performance expectancy, effort expectancy, social influence, and facil- it is discovered that fewer interaction possibilities are the most impor-
itating conditions—and is mainly used to predict information technol- tant motivating factors (Market Intelligence and Consulting Institute
ogy user intentions and behaviours. In addition, age, gender, [MIC], 2020). Therefore, this study modified the UTAUT2 model by
experience and voluntariness of use were found to moderate the removing price value and adding food order security to fit the current
effects of the decisive factors (Venkatesh et al., 2003). Brown and study context.
Venkatesh (2005) and Venkatesh et al. (2003) defined performance Hence, the dimensions of the model used in this study included
expectancy as the degree to which consumers believe that applying performance expectancy, effort expectancy, social influence, facilitat-
the technology is conducive to performing an activity, effort expec- ing conditions, hedonic motivation, consumer habits and consumer
tancy as the ease of use associated with the technology, social influ- security.
ence as the degree to which consumers consider their significant
others when deciding to use the technology, and facilitating condi-
tions as the context in which consumers believe that the resources 2.3 | Key factors that influence consumer choices
exist to support the use of the technology. Several prior studies have of FDPs
used the UTAUT model, assessing online ticket purchasing (Escobar-
Rodríguez & Carvajal-Trujillo, 2014), M-payment adoption (Al-Saedi This section further defines the evaluation criteria of the seven
et al., 2020), mobile banking (Jadil et al., 2021) and online shopping dimensions in terms of consumer FDP choice.
(Erjavec & Manfreda, 2022).
With the rapid advancement of information technology and the
development of new technological devices, Venkatesh et al. (2012) 2.3.1 | Performance expectancy
proposed the extended UTAUT (UTAUT2) model. In addition to the
four factors in UTAUT, factors that affect consumer acceptance of Performance expectancy refers to the extent to which an individual
the technology, such as hedonic motivation, price value and habits, believes that the use of the devices and services helps to improve the
were also introduced into the model, to broaden the comprehensive- performance of a given task. When performance expectancy is high,
ness of the framework (Tamilmani et al., 2021). As the original UTAUT individuals are more willing to use corresponding associated technol-
model only considers external motivations and overlooks the internal ogy (Alalwan, 2020; Ramos, 2022; Venkatesh et al., 2003; Venkatesh
motivations of consumers, UTAUT2 was designed to account for both et al., 2012). Accordingly, for the present study, performance expec-
external and internal motivations. Hence, although the framework of tancy was defined as the degree to which consumers believe that the
UTAUT2 is more complex than that of the original model, it better use of FDP apps provides more convenience or is more efficient than
explains consumer adoption behaviours (Tamilmani et al., 2021). Simi- other means. The evaluation criteria of this dimension included (1) the
larly, despite being a relatively new model, UTAUT2 has been applied FDP provided direct discounts aligned with consumer expectations
to several applications, such as influencing customers' intention to (PE1) (Alalwan, 2020; Ray & Bala, 2021; Sheehan et al., 2019); (2) using
adopt and actual adoption of internet banking (Alalwan et al., 2018), the FDP app reduces waiting time and increases convenience when
acceptation of online games in mobile devices (Ramírez-Correa obtaining the food (PE2) (Alalwan, 2020; Roh & Park, 2019); and
et al., 2019), continuance usage intention for food delivery apps (3) the FDP offers indirect discounts in the way of codes/coupons
(Ramos, 2022) and adoption of augmented reality technology (AR) in that meet consumer expectations (PE3) (Alalwan, 2020; Ray &
education (Faqih & Jaradat, 2021). Bala, 2021).
According to previous studies, one of the major factors influenc-
ing food delivery app acceptance is the security of personal informa-
tion (Ramesh et al., 2021). Cheng et al. (2021) examined the internet 2.3.2 | Effort expectancy
big data of online food delivery generated in the first and second
stages, and discovered 17 critical factors, the most significant of Effort expectancy refers to the ease associated with the use of new
which was security. Cho et al. (2019) proposed that security should products and systems (Alalwan, 2020; Ramos, 2022; Venkatesh
always be prioritized in all food ordering apps. This study explored the et al., 2003; Venkatesh et al., 2012). Accordingly, in this study, effort
factors affecting consumer FDP choices based on previous research. expectancy was defined as the degree to which consumers believe
Accounting for consumer rights and personal information protection, that the FDP app was easy to use and understand. The evaluation
the security of online food orders matters in FDP business models criteria of this dimension included (1) the interface of the FDP app
and is a critical consumer selection criterion. In terms of user motiva- was seen as easy to navigate (EE1) (Alalwan, 2020; Ramesh
tion during the pandemic, most users (50.4%) wanted to save time et al., 2021; Roh & Park, 2019), (2) using the FDP app to order meals
spent going out and queuing, while 39.4% were apprehensive because was seen as easy and straightforward (EE2) (Alalwan, 2020; Ramesh
of the pandemic, thereby reducing the chance of going out and con- et al., 2021; Roh & Park, 2019), and (3) an excessively complex design
tacting people, while discounts (38.7%), bad weather/do not want to of the FDP app affected consumer usage (EE3) (Alalwan, 2020;
go out, and a variety of meals were also important considerations. Ramesh et al., 2021; Roh & Park, 2019).
TSAI ET AL. 159
2.3.3 | Social influence (Alalwan, 2020; Venkatesh et al., 2003; Venkatesh et al., 2012). This
study defined habits as the consumers' decisions to continue using
Social influence refers to the degree to which the acceptance and use of the FDP app through experiential learning. The evaluation criteria of
new products/services/systems is dependent on the influence of their this dimension included (1) using the FDP app to order meals has
significant others (Alalwan, 2020; Okumus et al., 2018; Venkatesh become a habit for the consumers (UH1) (Alalwan, 2020; Ramesh
et al., 2003; Venkatesh et al., 2012). Accordingly, for the present study, et al., 2021); (2) consumers felt regularly and strongly motivated to
social influence was defined as the degree to which consumers perceive use the FDP app to order meals (UH2) (Alalwan, 2020; Ramesh
that their significant others recommend the FDP apps. The evaluation et al., 2021); and (3) consumers intend to use the FDP app to order
criteria of this dimension included (1) the attitudes of the consumer's rel- future meals (UH3) (Alalwan, 2020; Ramesh et al., 2021).
atives and friends affected the use of the FDP app (SI1) (Alalwan, 2020;
Ramesh et al., 2021), (2) the support received by relatives and friends to
use the FDP app (SI2) (Alalwan, 2020; Ramesh et al., 2021), and the 2.3.7 | Security
degree to which the general society believed that the use of the FDP
app was in line with current trends (SI3) (Alalwan, 2020; Ramesh According to the ‘Mobile Payment Consumer Survey’ conducted by the
et al., 2021). MIC (2017), security is the most important among the top five consumer
decision factors when considering mobile payment systems. Hence,
security is a dimension that should not be ignored. This study defined
2.3.4 | Facilitating conditions security as the degree to which the FDP operator is able to ensure the
security of the consumer (Cheng et al., 2021). The evaluation criteria of
Facilitating conditions refer to the extent to which individuals perceive the the dimension included (1) the security of transactional information in
existence of support when using a new product/service (Alalwan, 2020; the FDP app (SEC1) (Ramesh et al., 2021); (2) the security of personal
Okumus et al., 2018; Venkatesh et al., 2003; Venkatesh et al., 2012). information on the FDP app (SEC2) (Ramesh et al., 2021); and (3) the
Accordingly, for this study, facilitating conditions was defined as the extent security of the mobile operating system (SEC3).
to which consumers are aware of the resources and support available The 7 dimensions (performance expectancy, effort expectancy,
when using the FDP app. The evaluation criteria of this dimension included social influence, facilitating conditions, hedonic motivation, habits and
(1) consumers have perceptual behavioural control when using the FDP security) and 21 evaluation criteria were proposed based on the litera-
app to order food (FC1) (Alalwan, 2020; Ramesh et al., 2021); (2) when ture review and consumer interviews. The definitions of each criterion
faced with problems using the FDP app, consumers can seek support from are exhibited in Table 3.
a frequently asked questions (FAQ) knowledge bank (FC2) (Alalwan, 2020;
Ramesh et al., 2021); and (3) compatibility and stability of the FDP app
(FC3) (Alalwan, 2020; Ramesh et al., 2021). 3 | METHODOLOGY
For the present study, the selection of FDPs was evaluated from the
2.3.5 | Hedonic motivation perspective of consumers. First, relevant theories and past studies
were reviewed to obtain the main research framework. By referring to
Hedonic motivation refers to the degree to which individuals obtain the previous studies and user experiences, the influential indicators of
pleasure when using a new service/product (Alalwan, 2020; consumer FDP choices were obtained, and the DEMATEL method
Venkatesh et al., 2003; Venkatesh et al., 2012). In this study, hedonic was used to determine the causal relationships between the evalua-
motivation was defined as the degree of pleasure and enjoyment that tion dimensions and selection criteria (first questionnaire, distributed
consumers feel when using the FDP app. The evaluation criteria of online). The INRM of the evaluation dimensions and criteria was then
this dimension included (1) feelings of satisfaction when using the established, and the influential weights for each evaluation criterion
FDP app (HE1) (Alalwan, 2020; Ramesh et al., 2021; Venkatesh were calculated using the DANP method. Last, the modified VIKOR
et al., 2012); (2) feelings of modest enjoyment when using the FDP method was introduced to conduct a gap analysis on each evaluation
app (HE2) (Alalwan, 2020; Hwang et al., 2021; Ramesh et al., 2021; dimension and criterion (second questionnaire, performance-focused).
Venkatesh et al., 2012); and (3) perceptions of extreme enjoyment The results were expected to assist consumers FDP choices. The
when using the FDP app (HE3) (Alalwan, 2020; Ramesh et al., 2021; development process of the decision-making model is illustrated in
Venkatesh et al., 2012). Figure 1.
2.3.6 | Habits 3.1 | DEMATEL
Habits refer to the impact using the new service/product has on con- DEMATEL was developed by the Battelle Memorial Institute of
sumer behaviour through consumers' existing experiences and habits Geneva between the years 1972 and 1976 (Fontela & Gabus, 1976;
160 TSAI ET AL.
TABLE 3 Descriptions of the evaluation criteria for FDPs
Dimensions Criteria Definitions

Performance expectancy (PE) The FDP provided direct discounts aligned with The FDP occasionally launches a variety of
consumer expectations (PE1) preferential activities to meet consumer
expectations
Using the FDP app reduces waiting time and Using the FDP app reduces waiting time and
increases convenience when obtaining the food increases the convenience of obtaining meals
(PE2)
The FDP offers indirect discounts in the way of The FDP occasionally launches time-limited
codes/coupons that meet consumer discounts/promotions to meet consumer
expectations (PE3) expectations
Effort expectancy (EE) The interface of the FDP app was seen as easy to The design of the interface of the FDP app is
navigate (EE1) easy for consumers to navigate
Using the FDP app to order meals was seen as easy Consumers can use the FDP app to order meals
and straightforward (EE2) easily and straightforwardly
An excessively complex design of the FDP app The design of the FDP app is too complicated
affected consumer usage (EE3) and negatively affects consumers use
Social influence (SI) The attitudes of the consumer's relatives and Relatives and friends influence consumer
friends affected the use of the FDP app (SI1) willingness to use the FDP app
The support received by relatives and friends to A positive evaluation from relatives and friends
use the FDP app (SI2) affects the consumer's willingness to use the
FDP app
The degree to which the general society believed The public generally believes that using the FDP
that the use of the FDP app was in line with app is aligned with current trends, which, in
current trends (SI3) turn, affects individual consumer willingness to
use the app
Facilitating conditions (FC) perceptual behavioural control of the consumers Consumers are aware of their control over the
when using the FDP app to order food (FC1) resources and opportunities needed when
ordering meals using the FDP app
When faced with problems using the FDP app, When encountering issues using the FDP app,
consumers can seek support from a frequently consumers can easily find the FAQs to resolve
asked questions (FAQ) knowledge bank (FC2) them
The compatibility and stability of the FDP app The degree of compatibility of the FDP app
(FC3) ensures its stability
Hedonic motivation (HE) The feelings of satisfaction when using the FDP Using the FDP app is pleasurable
app (HE1)
The feelings of modest enjoyment when using the Using the FDP app is enjoyable
FDP app (HE2)
The perceptions of extreme enjoyment when using Using the FDP app is fun
the FDP app (HE3)
Use habit (UH) Using the FDP app to order meals has become a Due to the COVID-19 pandemic, using the FDP
habit for the consumers (UH1) app to order meals has become a habit for
consumers
Consumers felt regularly and strongly motivated to Due to the COVID-19 pandemic, consumers are
use the FDP app to order meals (UH2) regularly inclined to use the FDP app to order
meals
Consumers would intend to use the FDP app to Due to the COVID-19 pandemic, consumers feel
order future meals (UH3) they will use the FDP app to order more meals
Security (SEC) The security of transactional information in the The FDP operator ensures the security of
FDP app (SEC1) transactional information through the app
The security of personal information on the FDP The FDP operator ensures the security of
app (SEC2) consumers' personal information when using
the app
The security of the mobile operating system (SEC3) The FDP operator ensures the security of the
app when used on a mobile operating system
TSAI ET AL. 161
Construct an evaluation model for

consumer FDP choice Literature review
The UTAUT2 Model
Determine the evaluation indicators Consumer experience interviews

Establish the INRM
Develop the research framework based on
Phase One
the evaluation dimensions and criteria
obtained in the previous stage
Establish the research framework Design and distribute the causality

questionnaire
Determine the distribution targets

(consumers)
Collect the responses of the first survey and

Construct the INRM use DEMATEL to define the causal
relationships and INRM
Use the DANP to calculate the influential

Calculate the influential weights, conduct gap analysis, and
weights
Determine the influential weights based
on the results of the DEMATEL Design & distribute the gap analysis
formulate decision-making strategies
analysis questionnaire
Determine the distribution targets
Phase Two
(consumers)
Conduct a gap analysis
Collect the responses of the second survey

and use the modified VIKOR to analyse
gaps
Formulate improvement strategies based

on the results of the INRM, influential
weights, and ranking analysis
FIGURE 1 Research design framework
Gabus & Fontela, 1973). The method utilizes expert knowledge to competitiveness (Tsai, 2020) and retail stores (Tsai et al., 2022). The
clarify causal relationships between variables. In addition to conver- empirical results of the above research have collectively confirmed
ting the causal relationships between criteria into a clear structural that the direct-relation matrix of DEMATEL effectively illustrates the
model, the DEMATEL method aims to obtain the interdependences causal structure of complex problems.
across multiple criteria and the degree to which such dependencies In summary, DEMATEL is an effective method for
exist (Altuntas & Gok, 2021; Tsai et al., 2020; Tsai & Chen, 2021). establishing and analysing structural models and is generally used to
From its conception, DEMATEL has been widely applied in vari- identify the relationships in social phenomena and to reveal the inter-
ous areas, including business intelligence (Mavi & Standing, 2018), lu, 2016; Tsai &
dependences between factors (Gölcük & Baykasog
green marketing (Tsai et al., 2020), enhancing service industry Chen, 2021).
162 TSAI ET AL.
The calculation process of DEMATEL comprises of six main steps: the positive and negative ideal solutions. The VIKOR method has been
(1) generating the individual direct-relation matrix, (2) calculating the widely applied to different areas, such as wealth management and
average direct-influence relation matrix, (3) examining the consistency, bank evaluation (Chang & Tsai, 2016), evaluating aviation industry ser-
(4) calculating the normalized average direct-influence relation matrix, vice quality (Gupta, 2018), evaluating service business (Sofiyabadi
(5) calculating the total-influence relation matrix, and (6) establishing et al., 2016), trust barriers to online shopping (Rasty et al., 2021) and
the influential network relation map (INRM). The detailed steps are dis- evaluating TV game consoles (Tsai & Chen, 2021).
cussed by Peng and Tzeng (2013) and Tsai and Chen (2021). However, in terms of uncertainties in the market environment
(such as the impact of the COVID-19 pandemic and international mar-
ket fluctuations), past studies have tended to focus on ranking prob-
3.2 | DANP lems (Chang & Tsai, 2016; Tsai & Chang, 2013), rather than resolving
issues in the market. To effectively tackle this over-emphasis on rank-
Derived from the analytic hierarchy process (AHP) method, the analytic ing, Peng and Tzeng (2013) suggested that the focus of the model
network process (ANP) proposed by Saaty (1996) illustrates relations of should be shifted from ‘ranking’ to ‘improving’ the problems by iden-
criteria in the form of a network. Specifically, ANP adds dependence tifying performance gaps that can be used to assist managers and
and feedback mechanisms to AHP to solve dependencies among multi- decision makers to formulate improvement strategies and enhance
ple criteria. The traditional AHP method assumes that each evaluation competitive advantage (Lin et al., 2016; Peng & Tzeng, 2013;
attribute is independent and does not affect one another. However, in Tsai, 2020). Therefore, this study adopted the method proposed by
reality, such assumptions are rarely applicable as attributes are more Peng and Tzeng (2013) (modified VIKOR) to conduct a gap analysis of
likely to have inter-influence relationships. ANP solves nonlinear and the evaluation criteria.
The formula used for the gap analysis

aspired worst
complex hierarchical relationships, and its decision-making process bet- was rkj ¼ f aspired
j f kj = f j fj .
ter reflects real-life situations than AHP; hence, ANP has been widely
and increasingly used in recent years (Rao, 2021).
Ou Yang et al. (2008) contended that ANP assumes that each 4 | EMP I RICAL R ES UL TS
group of clusters has the same degree of influence, which overlooks
the fact that different clusters should have varying degrees of influ- 4.1 | Questionnaires and data collection
ence. In response, they proposed a method combining DEMATEL and
ANP (DEMATEL-based ANP, or DANP). The DANP uses DEMATEL to DEMATEL questionnaire was designed with reference to the litera-
identify the degree of influence between clusters, ascertains the ture review, UTAUT2 model and interviews on the consumer experi-
‘dynamic importance’ implied by the total-influence matrix to the sup- ence of using FDPs. A five-point scale was used to measure the
ermatrix of the ANP, and then multiplies the matrix until the result importance of the dimensions and criteria. Currently, the majority of
converges and stabilizes. The method is similar to the approach used previous studies were based on 10 to 30 responses from a DEMATEL
in ANP where the influence and importance of the criteria are questionnaire (Tsai, 2020; Tsai & Chen, 2021). However, given the
obtained through questionnaire surveys. research subjects of this study represent a significant population of
It is apparent that, in addition to considering the differing degrees Taiwanese consumers, to ensure satisfactory reliability this study col-
of influence across clusters, the DANP is simpler than the traditional lected 418 responses, which meet standard requirements for large-N
approach of combining DEMATEL and ANP that requires a DEMATEL sampling better.
questionnaire to determine the influence of the attributes and to The performance-gap questionnaire used for the study was
establish a network hierarchy, than requiring an ANP questionnaire to the same as that of the DEMATEL questionnaire. The overall question-
be used to identify the weights. Hence, for the present study, the naire format was drawn from the performance-gap questionnaire
DANP was employed to simplify the survey process and to evaluate developed by Peng and Tzeng (2013). The participants were asked
the different levels of influence across criteria (Rao, 2021; Tsai & to rate the importance and satisfaction for each dimension and cri-
Chen, 2021). terion. A 11-point scale was used to measure the importance in the
The total-influence matrix obtained in step 5 of DEMATEL was performance-gap questionnaire (‘0’ = ‘least important’ and
used to calculate the influential weights of the evaluation dimensions ‘10’ = ‘most important’). The survey participants were Taiwanese
and criteria. The detailed steps of the DANP are discussed by Peng consumers, and a total of 422 responses were collected.
and Tzeng (2013) and Tsai and Chen (2021).
4.2 | DEMATEL analysis

3.3 | Gap analysis with modified VIKOR
4.2.1 | Process of DEMATEL and result analysis
The VIKOR proposed by Opricovic (1998) is an optimal compromise
programming approach to multiple-criteria decision making (MCDM). After collecting the responses to DEMATEL, the analysis process was
The general idea is to seek the optimal (compromise) solution among conducted in accordance with Section 3.1.
TSAI ET AL. 163
T A B L E 4 Average direct-influence
Dimensions PE EE SI FC HE UH SEC
relations matrix A of dimensions
PE 0.000 2.249 2.278 2.385 2.431 2.481 2.414
EE 2.282 0.000 2.323 2.419 2.395 2.517 2.407
SI 2.306 2.352 0.000 2.419 2.435 2.438 2.376
FC 2.376 2.354 2.407 0.000 2.416 2.411 2.438
HE 2.328 2.356 2.483 2.409 0.000 2.481 2.445
UH 2.368 2.440 2.517 2.486 2.488 0.000 2.431
SEC 2.364 2.347 2.450 2.452 2.526 2.469 0.000
TABLE 5 Normalized matrix D of

Dimensions PE EE SI FC HE UH SEC
dimensions
PE 0.000 0.152 0.154 0.161 0.164 0.168 0.163
EE 0.154 0.000 0.157 0.163 0.162 0.170 0.163
SI 0.156 0.159 0.000 0.163 0.165 0.165 0.161
FC 0.161 0.159 0.163 0.000 0.163 0.163 0.165
HE 0.157 0.159 0.168 0.163 0.000 0.168 0.165
UH 0.160 0.165 0.170 0.168 0.168 0.000 0.164
SEC 0.160 0.159 0.166 0.166 0.171 0.167 0.000
TABLE 6 Total influence matrix T of

Dimensions PE EE SI FC HE UH SEC Di
dimensions
PE 5.639 5.798 5.924 5.966 6.010 6.048 5.947 41.332
EE 5.809 5.702 5.964 6.005 6.046 6.087 5.983 41.596
SI 5.803 5.832 5.821 5.999 6.041 6.076 5.975 41.549
FC 5.833 5.858 5.988 5.885 6.067 6.102 6.005 41.738
HE 5.865 5.893 6.027 6.060 5.963 6.142 6.041 41.991
UH 5.943 5.974 6.107 6.143 6.186 6.078 6.119 42.551
SEC 5.902 5.929 6.062 6.099 6.146 6.178 5.936 42.252
Rj 40.794 40.986 41.893 42.157 42.461 42.711 42.006
Note: Any value greater than threshold value is presented in bold (The threshold θ = 5.980).
Step 1: Generating the individual direct-relation matrix than the threshold value (α = 0.05), indicating that the responses of
The step 1 was used to generate the individual direct-relation matrix. the 418 participants reached consistency and consensus.
The participants were asked to refer to the attached definition of each
criterion and to rate their perceived influence using a 5-point scale Step 4: Calculating the normalized average direct-influence relation
(‘0’ = ‘no influence’ and ‘4’ = extremely high influence). The greater matrix
the rating, the greater the degree of influence between the two evalu- In this step, the matrix A obtained in step 2 was normalized to obtain the
ation criteria. normalized direct-influence relation matrix D. After confirming the con-
sistency of the average direct-influence relation matrix, the sum of the
Step 2: Calculating the average direct-influence relation matrix vertical and horizontal criteria was compared, and the maximum value
The mean ratings given by the participants per evaluation criterion was selected and divided by the value of each criterion, respectively, to
was then calculated to generate the average direct-relation matrix A. yield the normalized average direct-influence relation matrix (Table 5).
The results are presented in Table 4.
Step 5: Calculating the total-influence relation matrix
Step 3: Examining the consistency The total-influence relation matrix T was obtained (Table 6). The
In this step, the consistency of the responses provided by the par- threshold value was calculated using the arithmetic mean of the tij in
P n
n P T. Specifically, the sum of all tij was divided by 49, and the threshold
H H
1
ticipants was tested using nðn1 Þ aij aH1
ij =aij 100%. The
i¼1 j¼1 value θ = 5.980 was ascertained. The threshold value was then used
results showed that the consistency ratio was 0.023, which is smaller to determine the existence of the influential relationship between the
164 TSAI ET AL.
two dimensions. For example, PE was found to influence SI, yet did causality values of SI, FC, HE and UH were 0.344, 0.420, 0.470
not influence EE. and 0.161, respectively (all smaller than 0); therefore, SI, FC, HE and
UH were the effects and were influenced by the cause dimensions.
Step 6: Establishing the INRM The strength of influence of HE, UH and SEC were the highest
In this step, the values in the columns and rows of T were summed. (84.452, 85.262 and 84.258, respectively), suggesting that the influ-
The sums of the rows and columns were represented as Di and Rj, ence between HE, UH and SEC and other dimensions were stronger;
respectively. The strength of the relationship (Di + Rj) and causality in other words, they had more mutual influential relationships with
(Di Rj) between the dimensions were calculated, the results of which other dimensions.
are shown in Table 7. An INRM was then generated based on Table 7 to illustrate the
The causality results showed that the causality values of PE, EE and relationships between the dimensions (Figure 2). The causality results
SEC were 0.538, 0.609 and 0.246, respectively (all greater than 0); there- indicated the direction of the arrows. A single arrow indicated a one-
fore, PE, EE and SEC were the causes and influences of other evaluation way influence from a given dimension toward other single dimensions,
dimensions, rather than being influenced by other dimensions. The (such as PE to UH), while a double arrow indicated a two-way influ-
ence between two dimensions (such as SEC and HE).
As is shown in Figure 2, the seven dimensions were related to
TABLE 7 Strength of influence and causality between evaluation
one another. Specifically, there were arrows directed from SEC to SI,
dimensions
FC, HE and UH, and the relationships were strong; thus, when SEC
Dimensions Di Rj Di + Rj Di Rj
changed, SI, FC, HE and UH tended to change accordingly; hence,
PE 41.332 40.794 82.126 0.538 SEC had the greatest degree of influence. Based on the protection of
EE 41.596 40.986 82.582 0.609 consumer rights and personal information, security is the priority for
SI 41.549 41.893 83.442 0.344 FDP operators as well as the greatest concern of consumers when
FC 41.738 42.157 83.895 0.420 choosing an FDP (Kumar & Shah, 2021; Ray & Bala, 2021; Yang
HE 41.991 42.461 84.452 0.470 et al., 2021).
UH 42.551 42.711 85.262 0.161 The results also showed that the overall influence of UH was most
SEC 42.252 42.006 84.258 0.246 prominent, indicating that UH was a core evaluation criterion. The cau-
sality analysis showed that UH was an effective factor; therefore, to
Note: (1) Di + Rj refers to the strength of the influential relationships
improve UH, it was necessary to improve the factors that influenced
between the dimensions. (2) Di Rj refers to the degree of causality
between dimensions. A positive value indicates that the dimension is the UH. Improving SEC was found to be most effective, as it influenced SI,
cause, and a negative value indicates that the dimension is the effect. FC, HE and UH, and could also, therefore, indirectly improve UH.
FIGURE 2 The INRM of the evaluation dimensions

TSAI ET AL. 165
TABLE 8 Average direct influence matrix (A) about evaluation T A B L E 9 Normalized average direct influence matrix (D) about
criteria evaluation criteria
Dimensions/criteria Dimensions/criteria
Performance expectancy (PE) PE1 PE2 PE3 Performance expectancy (PE) PE1 PE2 PE3
PE1 0.000 2.560 2.555 PE1 0.000 0.495 0.494
PE2 2.555 0.000 2.615 PE2 0.494 0.000 0.506
PE3 2.557 2.569 0.000 PE3 0.495 0.497 0.000
Effort expectancy (EE) EE1 EE2 EE3 Effort expectancy (EE) EE1 EE2 EE3
EE1 0.000 2.569 2.545 EE1 0.000 0.491 0.486
EE2 2.617 0.000 2.620 EE2 0.500 0.000 0.500
EE3 2.579 2.608 0.000 EE3 0.492 0.498 0.000
Social influence (SI) SI1 SI2 SI3 Social influence (SI) SI1 SI2 SI3
SI1 0.000 2.404 2.488 SI1 0.000 0.480 0.497
SI2 2.390 0.000 2.517 SI2 0.478 0.000 0.503
SI3 2.383 2.407 0.000 SI3 0.476 0.481 0.000
Facilitating conditions (FC) FC1 FC2 FC3 Facilitating conditions (FC) FC1 FC2 FC3
FC1 0.000 2.438 2.545 FC1 0.000 0.476 0.497
FC2 2.536 0.000 2.581 FC2 0.495 0.000 0.503
FC3 2.433 2.510 0.000 FC3 0.475 0.490 0.000
Hedonic motivation (HE) HE1 HE2 HE3 Hedonic motivation (HE) HE1 HE2 HE3
HE1 0.000 2.447 2.476 HE1 0.000 0.494 0.500
HE2 2.450 0.000 2.462 HE2 0.495 0.000 0.497
HE3 2.469 2.483 0.000 HE3 0.499 0.501 0.000
Use habit (UH) UH1 UH2 UH3 Use habit (UH) UH1 UH2 UH3
UH1 0.000 2.455 2.464 UH1 0.000 0.492 0.494
UH2 2.483 0.000 2.495 UH2 0.498 0.000 0.500
UH3 2.469 2.517 0.000 UH3 0.495 0.505 0.000
Security (SEC) SEC1 SEC2 SEC3 Security (SEC) SEC1 SEC2 SEC3
SEC1 0.000 2.572 2.653 SEC1 0.000 0.483 0.498
SEC2 2.636 0.000 2.670 SEC2 0.495 0.000 0.502
SEC3 2.620 2.694 0.000 SEC3 0.492 0.506 0.000
4.2.2 | DEMATEL analysis of evaluation criteria the existence of an influential relation between the two given
criteria. Next, the values in the columns and rows of T were summed
The DEMATEL analysis of the evaluation criteria was based on that up and represented by Di and Rj, respectively. The strength of the
outlined in Section 3.1. The consistency ratio of the criteria of relationship (Di + Rj) and causality (Di Rj) between the criteria
PE = 0.003; that of the evaluation criteria of EE = 0.002; that of the were calculated, the results of which were shown in Table 11. An
criteria of SI = 0.003; that of the criteria of FC = 0.003; that of the INRM was then generated based on Table 11 to illustrate the rela-
criteria of HE = 0.003; that of the criteria of UH = 0.003 and that of tionships between the criteria (Figure 3). Causality was indicated by
the criteria of SEC = 0.002. Hence, it can be seen that the responses the direction of arrows.
reached consistency and consensus. The average direct-influence
relation matrix, normalized average direct-influence relation matrix, Performance expectancy (PE)
total-influence relation matrix, the strength of the relationship, and As is shown in Figure 3, PE2 had the greatest degree of influence and
causality of the criteria are exhibited in Tables 8–11, respectively. strength of influence, suggesting that it was the most important factor
Detailed analysis and the INRM results of the criteria for each dimen- among the PE criteria. According to MIC, consumers choose food
sion are outlined in the following paragraphs. delivery for three main reasons: ‘avoiding the need to go out’, ‘saving
The total-influence relation matrix T was obtained (Table 10), time’, and ‘discounts on delivery fees’. The first two reasons repre-
and the threshold was calculated as 52.371 (PE), 30.040 (EE), sent the true needs of consumers, which FDPs serve through their
11.363 (SI), 14.787 (FC), 71.708 (HE), 66.955 (UH) and 42.559 service offerings. Discounts, on the other hand, were not a primary
(SEC), respectively. The threshold value was then used to determine factor that attracted consumers. Therefore, PE2 and PE1 influenced
166 TSAI ET AL.
TABLE 10 Total influence matrix (T) about evaluation criteria a core factor. Therefore, marketing efforts should focus on the
Dimensions/criteria impression that the use of FDP apps is a public trend, thereby pro-
moting SI2 and influencing SI1, and indirectly stimulating consumer
Performance expectancy (PE) PE1 PE2 PE3
willingness to use the platform.
PE1 51.831 52.275 52.550
PE2 52.533 52.317 52.930
Facilitating conditions (FC)
PE3 52.243 52.358 52.301
As is shown in Figure 3, FC2 had the greatest degree of influence
Effort expectancy (EE) EE1 EE2 EE3
and strength of influence, suggesting that it was the most important
EE1 29.634 29.890 29.842 factor among the FC criteria. In terms of the strength of the rela-
EE2 30.438 30.030 30.317 tionship, FC3 was found to have the greatest overall influence, indi-
EE3 30.242 30.171 29.792 cating that it was a core factor. This implies that it is necessary to
Social influence (SI) SI1 SI2 SI3 improve the compatibility and stability of FDPs and thus to promote
SI1 11.051 11.436 11.742 FC2 and influence FC1, indirectly stimulating consumer willingness
SI2 11.396 11.133 11.766 to use the platform.
SI3 11.217 11.279 11.248

Facilitating conditions (FC) FC1 FC2 FC3
Hedonic motivation (HE)
As is shown in Figure 3, HE3 had the greatest degree of influence and
FC1 14.416 14.700 15.055
strength of influence, suggesting that it was the most important factor
FC2 15.007 14.638 15.325
among the HE criteria. In terms of the strength of the relationship,
FC3 14.662 14.631 14.646
HE2 was found to have the greatest overall influence, indicating that
Hedonic motivation (HE) HE1 HE2 HE3
it was a core factor. Hence, by improving the enjoyment of use of
HE1 70.704 71.150 71.221
FDPs, HE3 and HE1 are both influenced, which indirectly stimulates
HE2 70.920 70.704 71.104 consumer willingness to use the platform.
HE3 71.311 71.428 71.162
Use habit (UH) UH1 UH2 UH3 Habits (UH)
UH1 66.274 66.775 66.669 As is shown in Figure 3, UH2 had the greatest degree of influence and
UH2 67.145 66.985 67.211 strength of influence, suggesting that it was the most important factor
UH3 67.209 67.386 66.942 among the UH criteria. In terms of the strength of the relationship,
Security (SEC) SEC1 SEC2 SEC3 UH1 was found to have the greatest overall influence, indicating that
it was a core factor. Therefore, by encouraging consumers to use the
SEC1 41.896 42.276 42.585
FDPs to the point that doing so becomes a habit, UH2 and UH3 are
SEC2 42.663 42.387 43.026
influenced, which indirectly stimulates consumer willingness to use
SEC3 42.701 42.762 42.732
the platform.
Note: Any value greater than threshold value is presented in bold
performance expectancy (θ = 52.371), effort expectancy (θ = 30.040),
Security (SEC)
social influence (θ = 11.363), facilitating conditions (θ = 14.787), hedonic
motivation (θ = 71.078), use habit (θ = 66.955), and security (θ = 42.559). As is shown in Figure 3, SEC2 had the greatest degree of influence
and strength of influence, suggesting that it was the most impor-
tant factor among the SEC criteria. In terms of the strength of the
consumer willingness to use FDPs, and converged on the decision relationship, SEC3 was found to have the greatest overall influ-
direction of PE3. ence, indicating that it was a core factor. Therefore, it is necessary
to enhance the security between the mobile phone operating sys-
Effort expectancy (EE) tem and the platform in order to promote SEC2 and influence
As is shown in Figure 3, EE2 had the greatest degree of influence and SEC1, which indirectly stimulates consumer willingness to use the
strength of influence, suggesting that it was the most important factor platform.
among the EE criteria. Therefore, it is necessary to improve EE1 in
order to improve EE2 and EE3, ensuring a more user-friendly interface
to promote consumer willingness to engage with the FDP. 4.3 | DANP analysis
Social influence (SI) The total-influence matrix T obtained in DEMATEL was used to calcu-
As is shown in Figure 3, SI2 had the greatest degree of influence and late the influential weights of the evaluation criteria, and the detailed
strength of influence, suggesting that it was the most important factor results are as follows.
among the SI criteria. In terms of the strength of the relationship, SI3 The main purpose of the DANP was to determine the influential
was found to have the greatest overall influence, indicating that it was weights, so as to identify the most influential factors when consumers
TSAI ET AL. 167
T A B L E 1 1 Strength of influence and

Dimensions/criteria Di Rj D i + Rj Di Rj
causality between evaluation criteria
Performance expectancy (PE) 41.332 40.794 82.126 0.538
PE1 156.656 156.607 313.262 0.049
PE2 157.780 156.950 314.730 0.830
PE3 156.901 157.780 314.682 0.879
Effort expectancy (EE) 41.596 40.986 82.582 0.609
EE1 89.365 90.314 179.679 0.949
EE2 90.785 90.090 180.876 0.695
EE3 90.205 89.951 180.156 0.254
Social influence (SI) 41.549 41.893 83.442 0.344
SI1 34.228 33.664 67.892 0.564
SI2 34.295 33.847 68.142 0.447
SI3 33.745 34.756 68.501 1.012
Facilitating conditions (FC) 41.738 42.157 83.895 0.420
FC1 44.171 44.085 88.256 0.086
FC2 44.970 43.968 88.938 1.002
FC3 43.939 45.026 88.965 1.087
Hedonic motivation (HE) 41.991 42.461 84.452 0.470
HE1 213.075 212.935 426.010 0.140
HE2 212.727 213.281 426.009 0.554
HE3 213.901 213.487 427.388 0.414
Use habit (UH) 42.551 42.711 85.262 0.161
UH1 199.718 200.627 400.346 0.909
UH2 201.340 201.145 402.485 0.194
UH3 201.537 200.822 402.359 0.715
Security (SEC) 42.252 42.006 84.258 0.246
SEC1 126.756 127.260 254.016 0.504
SEC2 128.076 127.424 255.500 0.651
SEC3 128.194 128.342 256.537 0.148
evaluate and choose an FDP. The data collected through the ques- two criteria equally important. Among the criteria of SEC, SEC3 was
tionnaire survey were used for the calculation and the results ranked considered the most important criterion; however, the weights of
according to the weights are shown in Table 12. It can be seen from SEC1 and SEC2 did not differ significantly from that of SEC3, indicat-
the table that the rank of importance of the dimensions from high to ing that consumers considered the other two criteria equally
low was UH, HE, FC, SEC, SI, EE and PE. important.
Among the criteria of PE, PE3 was considered the most important According to the calculated influential weights exhibited in
criterion, while PE1 was considered the least important. Among the Table 12, UH1, UH2, UH3 and FC3 were the most influential fac-
criteria of EE, EE1 was considered the most important criterion, while tors in terms of consumer FDP choice. In addition, the specific
EE2 and EE3 were considered the least important. Among the criteria influential weights were generally similar between the dimensions
of SI, SI3 was considered the most important criterion, while SI1 was and criteria, suggesting that the differences in the importance of
considered the least important. Among the criteria of FC, FC3 was the evaluation factors proposed by this study were not
considered the most important criterion, while FC2 was considered substantial.
the least important. Among the criteria of HE, HE3 was considered
the most important criterion; however, the weights of HE1 and HE3
did not differ significantly from that of HE3, indicating that consumers 4.4 | Modified VIKOR analysis
considered the other two criteria equally important. Among the
criteria of UH, UH2 was considered the most important criterion; Table 13 shows the gap values of the evaluation dimensions and
however, the weights of UH1 and UH3 did not differ significantly criteria. PE was found to have the largest gap value, suggesting
from that of UH2, indicating that consumers considered the other that PE had the greatest gap between actual and desired
168 TSAI ET AL.
FIGURE 3 The INRM of each criterion for the food delivery platforms
performance. PE1 and SI1 were found to have the largest gap 5 | DI SCU SSION AND I MPLICATIO NS
values among the criteria, followed by FC1. Since the gap values
represent the difference between actual and desired perfor- Using a thorough review of the literature and interviews on con-
mance, dimensions and criteria with larger gap values required sumer user experience, this study constructed a consumer evalua-
greater improvement. Therefore, PE, PE1, SI1 and FC1 should tion model for the selection of FDPs. This study's theoretical,
be the first priority for FDPs when implementing improvement practical, and managerial implications are written in the following
measures. sections.
TSAI ET AL. 169
TABLE 12 Ranking of influential weights of the evaluation dimensions and criteria
Dimensions Original weights Ranking Criteria Original weights Ranking Overall weights Ranking
Performance expectancy (PE) 0.139 7 PE1 0.332 3 0.046 20–21
PE2 0.333 2 0.046 20–21
PE3 0.335 1 0.047 14–19
Effort expectancy (EE) 0.140 6 EE1 0.334 1 0.047 14–19
EE2 0.333 2–3 0.047 14–19
EE3 0.333 2–3 0.047 14–19
Social influence (SI) 0.143 5 SI1 0.329 3 0.047 14–19
SI2 0.331 2 0.047 14–19
SI3 0.340 1 0.049 1–5
Facilitating conditions (FC) 0.144 3 FC1 0.331 2 0.048 6–13
FC2 0.330 3 0.048 6–13
FC3 0.338 1 0.049 1–5
Hedonic motivation (HE) 0.145 2 HE1 0.333 2–3 0.048 6–13
HE2 0.333 2–3 0.048 6–13
HE3 0.334 1 0.048 6–13
Use habit (UH) 0.146 1 UH1 0.333 2–3 0.049 1–5
UH2 0.334 1 0.049 1–5
UH3 0.333 2–3 0.049 1–5
Security (SEC) 0.143 4 SEC1 0.332 3 0.048 6–13
SEC2 0.333 2 0.048 6–13
SEC3 0.335 1 0.048 6–13
5.1 | Theoretical implications substantial impact on customers' propensity to utilize food delivery
apps. This further demonstrates that SEC are essential considerations
This study uses customers on FDPs as the research object, investi- evaluated by customers in the online ordering environment, having a
gates the key elements that influence their FDPs selection behaviour, significant impact on their usage intention and behaviours.
and fills gaps in the research on consumers' continuous usage inten- Second, from influential weights, the DANP's major goal was
tions and behaviours. To put forward a research model, we changed to discover the influential weights to identify the most relevant
the original UTAUT2 theoretical model, incorporated a safety dimen- aspects when consumers assess and chose an FDP. The empirical
sion, and employed a hybrid MADM model (DEMATEL, DANP and evidence reveals that both use habit and hedonic motivation have
modified VIKOR) that affects consumers' usage of FDPs on food a significant influence on usage intention, with use habit having a
delivery platforms. Previous research claimed that a structural equa- significant effect. This finding contributes to prior studies on food
tion model (SEM) is a hypothesis regarding the structure of interac- delivery users by confirming the importance of use habit and pro-
tions among measured variables in a given population, and that SEM viding an empirical basis for investigating consumers' food delivery
necessitates a high sample size (MacCallum & Austin, 2000). app usage intention. We also analysed the study model's gap
DEMATEL, on the other hand, just requires 10 samples to provide rel- values and discovered that both performance expectation and
evant research (Tsai, 2020; Tsai & Chen, 2021). In relevant research effort expectancy had the largest role in the difference between
that employed DEMATEL, the average sample size was between actual and intended performance.
10 and 30 specialists (Tsai & Chen, 2021). This study adheres to the Third, we used the modified UTAUT2 model to explain con-
big sample approach based on consumer perceptions, and as a conse- sumers' choice and intention to utilize food delivery apps, which
quence, 418 customers were sampled. Therefore, we claim that the broadens the scope of UTAUT2 research and offers a theoretical
DEMATEL-based influential network technique is more appropriate in foundation for future studies on consumers' propensity to reuse food
determining the causal relationship based on consumers' perceptions. delivery apps. By highlighting this conclusion, we not only assist to
Without the predetermined route framework, a new analytic frame- promote food delivery research, extend it in food delivery platforms,
work based on the population of interest, in this case the consumers, and fill gaps in past studies, but we also enable food delivery enter-
may be developed. In accordance with the INRM of the evaluation prises establish various business and marketing strategies for
dimensions, the study's findings suggest that the SEC factor has a consumers.
170 TSAI ET AL.
TABLE 13 Gap values and rankings of the evaluation dimensions TABLE 14 Proposal for improvement strategies
and criteria
Program Strategy for improvement
Rank of Rank of
F1: Influential network SEC ! EE ! PE ! SI ! FC ! HE!UH
Dimensions/criteria Gaps dimensions criteria of dimensions
Performance 0.371 7 F2: Influential network PE: PE2 ! PE1 ! PE3
expectancy (PE) of criteria EE: EE2 ! EE3 ! EE1
PE1 0.356 20–21 SI: SI2 ! SI1 ! SI3
PE2 0.313 7–8 FC: FC2 ! FC1 ! FC3
PE3 0.309 5 HE: HE3 ! HE1 ! HE2
Effort expectancy (EE) 0.359 6 UH: UH2 ! UH3 ! UH1
EE1 0.315 10–11 SEC: SEC2 ! SEC3 ! SEC1
EE2 0.306 3–4 F3: Sequence of PE ! EE ! FC ! SEC ! SI ! UH ! HE
dimensions to rise to
EE3 0.306 3–4
desired level (by gap
Social influence (SI) 0.338 3 value, from high to
SI1 0.356 20–21 low)
SI2 0.329 16 F4: Sequence of criteria PE: PE1 ! PE2 ! PE3
to rise to desired EE: EE1 ! EE2 ! EE3
SI3 0.304 2
level within individual SI: SI1 ! SI2 ! SI3
Facilitating conditions 0.340 4–5 criteria (by gap value, FC: FC1 ! FC2 ! FC3
(FC) from high to low) HE: HE1 ! HE3 ! HE2
FC1 0.341 19 UH: UH3 ! UH1 ! UH2
FC2 0.326 14 SEC: SEC1 ! SEC2 ! SEC3
FC3 0.313 7–8
Hedonic motivation 0.318 1
(HE)
prioritized when FDPs implement improvement measures. Although
HE1 0.315 10–11 this study adopted a consumer perspective, the FDPs and consumers
HE2 0.303 1 are closely associated. A stable and sound FDP provides intermediary
HE3 0.310 6 functions to assist in handling the diverse set of services for the
Use habit (UH) 0.322 2 enlisted restaurants, to offer consumers improved meal ordering
UH1 0.334 17 experiences daily (see Table 14).
UH2 0.314 9 FDPs operators may benefit from the study's findings by better
UH3 0.336 18 understanding consumers' wants and providing better service; the
ultimate goal is to boost sales volume, broaden the scope of business,
Security (SEC) 0.340 4–5
and enable consumers to utilize FDPs more successfully. Based on the
SEC1 0.328 15
gap analysis results, FDPs operators may become aware of their
SEC2 0.321 13
shortcomings and take suitable efforts to remedy them, allowing them
SEC3 0.317 12
to retain existing clients and attract new ones.
Note: A smaller gap value indicates less improvement was needed, while a
larger gap value indicates more improvement is required; the factors are
ranked in ascending order based on gap values.
5.3 | Managerial implications
The findings of this study have significant management implications

5.2 | Practical implications for FDPs in terms of service enhancement and the creation of appro-
priate consumer-attraction tactics.
DEMATEL results highlighted that security was the most influential First, in this section, the relationships between the evaluation
factor among all evaluation dimensions, which echoed the findings of dimensions are outlined. According to Table 7 and Figure 2, the
past studies that consumers' first priority when selecting FDPs was strength of influence of SEC was the greatest among the dimensions
the security of the app (Kumar & Shah, 2021; Ray & Bala, 2021; Yang with positive causality values, indicating that SEC had the greatest
et al., 2021); hence, FDP operators should consider security a primary impact on consumer FDP choice and was a primary factor for resolv-
factor when designing and improving app interfaces. The gap analysis ing core issues. The results were consistent with the findings of Cheng
aimed to identify the dimensions that required immediate improve- et al. (2021) and Hong et al. (2021), who discovered that by enhancing
ment. The results showed that performance expectancy and effort security, the FDP operator can assure customer security. Specifically,
expectancy had the largest gap values, indicating that they should be when platforms make changes in SEC, it follows that SI, FC, HE and
TSAI ET AL. 171
UH tend to change accordingly. This finding echoed the past studies Among the PE criteria, PE3 was considered the most important
that claim consumer willingness to use an FDP is impacted signifi- criterion. Due to the ongoing pandemic, Taiwan was under Level
cantly by the security of the system (Cheng et al., 2021; Hong 3 national alert during this study. Many local businesses voluntarily
et al., 2021; Kumar & Shah, 2021). suspended indoor services and only offered delivery and take-out ser-
The strength of influence of EE was the second-largest among vices. As a result, FDPs became an important driving force for the pro-
the dimensions with positive causality values, suggesting that EE had tection of consumers as well as a means of survival for the catering
the second-largest effect on consumer FDP choice. The results were industry. Many FDPs launched discounts/promotional codes, and
consistent with the findings of Ramos (2022), who affected the con- some offered surprises every day to stimulate consumer use (Ramesh
sumers' ongoing usage behaviour. Specifically, when platforms make et al., 2021).
changes in EE, the SI, FC, HE and SEC tend to change accordingly. In Among the EE criteria, EE1 was considered the most important
addition to the impact of the COVID-19 pandemic, consumer inten- criterion. Ease of use is a user-centred design concept. The focus is
tion to use FDP apps was associated with ease of learning and of use. to ensure that the product design meets user daily habits and
Therefore, improving EE criteria was of great importance. We suggest needs. In terms of the FDP apps, the aim is to ensure that con-
that FDPs improve user interfaces, so that consumers can order food sumers do not experience stress or frustration during the meal
more easily and effortlessly, which should increase the willingness to selection process and that they can efficiently make an order with
use the app (Alalwan, 2020; Ramos, 2022). minimal effort (Belarmino et al., 2021; Ramesh et al., 2021). The
The third-most influential dimension was PE, and its causality results were consistent with the findings of Okumus et al. (2018).
value was positive, indicating that PE also plays an important role in Among the SI criteria, SI3 (0.340) was considered the most impor-
affecting consumer FDP choice (Ramos, 2022). These findings were tant criterion. ‘Staying at home’ to cope with the pandemic acceler-
consistent with the findings of Ramos (2022) and Zanetta ated the growth of the food delivery sector. Ordering meals online
et al. (2021), which influenced consumers' continuous usage behav- became another everyday choice for Taiwanese. Individuals chose
iour. Hence, delivering the belief to consumers that they could ben- to use the FDPs for four primary reasons: first, diversity in choice;
efit from using the app and that doing so was more efficient than second, there is no need to physically seek out restaurants; third,
other food ordering processes was important. It is also suggested promotional discounts; fourth, solving daily life problems. Therefore,
that FDPs provide discounts, implement measures that reduce we suggest that FDP operators influence public viewpoints to stim-
waiting time and ensure getting the meals is an effortless process, ulate consumer willingness to use their platforms (Ramesh
and occasionally offer discounts/promotional codes to stimulate et al., 2021). These findings were consistent with the findings of
consumer demand. In this sense, the importance of improving PE Okumus et al. (2018) and Zhao and Bacao (2020).
cannot be understated (Alalwan, 2020; Ramesh et al., 2021; Among the FC criteria, FC3 was considered the most important
Ramos, 2022). criterion. By upgrading the user interface (UI) and user experience
UH was found to have the greatest strength of influence and its (UX) design and stability of the system, FDPs could provide better ser-
causality value was negative; indicating that UH was a core factor that vices and user experiences to facilitate consumer willingness to use
affected consumer choice and should be improved upon urgently. The their platforms (Ramesh et al., 2021). These findings were consistent
results were consistent with the findings of Zanetta et al. (2021), who with the findings of Okumus et al. (2018) and Yeo et al. (2017).
affected the consumers' continuous usage intention for FDP apps. Among the HE criteria, HE3 was considered the most important crite-
Due to the pandemic, people are forced to remain indoors, which has, rion. These findings were consistent with the findings of Yeo
in turn, triggered increased use of FDP apps and driven the develop- et al. (2017). The founders of FDPs believe that ordering food should
ment of food delivery services. For that reason, FDP operators should be a simple, fast, and fun experience, which served as foundational
prioritize measures that effectively embed the use of their services concepts in platforms such as Uber Eats and Foodpanda. By integrat-
into consumer life, so that ordering food online becomes a regular ing nearby high-quality restaurants and providing delivery services,
consumer habit (Alalwan, 2020; Ramesh et al., 2021). To achieve this FDPs allow consumers to order meals online and enjoy the experience
goal, we suggest that FDP operators start by improving security fac- of eating at home while saving the trouble of dining out so that they
tors and continue to target effort expectancy, performance expec- have a better dining experience, which further stimulates future use
tancy, and social influence, while facilitating conditions that generate (Ramesh et al., 2021).
hedonic motivation, so that using FDP becomes an indispensable Among the UH criteria, UH2 was considered the most important
component of consumer life. criterion. The pandemic has forced many people to remain at home,
Second, the evaluation criteria weights obtained through DANP which has, in turn, boosted the app-based food delivery sector and
were used to identify the most influential factors when consumers driven the development of corresponding services. In recent years,
evaluate and choose an FDP. Table 12 shows that the most important consumers have become increasingly used to food delivery services
factors were UH, HE, FC, SEC, SI, EE and PE. The original weights (Ramesh et al., 2021; Zanetta et al., 2021). Among the SEC criteria,
were 0.146, 0.145, 0.144, 0.143, 0.143, 0.140 and 0.139, SEC3 was considered the most important criterion. Based on the prin-
respectively. ciple of protecting consumer rights and personal information, security
172 TSAI ET AL.
is the first priority of FDP operators as well as one of the greatest parties. Since different parties may have varied insights, the
concerns of consumers when selecting FDPs (Cheng et al., 2021; corresponding evaluation model may also be different. Second, in this
Hong et al., 2021). study, DEMATEL was used to identify the causal relations and direc-
Lastly, for the present study, performance values and gap values tion of influence for the dimensions. We suggest that future studies
of the criteria were calculated using modified VIKOR; the results should further verify the findings of this study using partial least
served as the basis for improving the factors that affect consumer squares structural equation modelling (PLS-SEM). Third, past research
FDP choice. It can be seen from Table 13 that the gap value of PE employed the UTATU2 model to investigate the factors that impacted
was the largest and the performance value was the smallest, indicating consumers' inclination to continue using food delivery apps during
that the performance expectancy of the FDPs needed further COVID-19 quarantine. The findings indicate that pricing value affects
improvement. From a practical point of view, although the FDPs customers' continued usage behaviour (Ramos, 2022). Future studies
invested effort in performance expectancy, consumers perceived that may incorporate an essential component, price value, to investigate
the benefits and efficiency of the process appeared to be limited. the major elements influencing customer choice of FDPs. Finally,
Therefore, performance expectancy should be the first priority for COVID 19 has resulted in structural changes in many areas of life and
improvement. Second, EE was found to have the second-highest gap business (Chopdar, Paul, & Prodanova, 2022; Gordon-Wilson, 2022;
value and the second-lowest performance value, suggesting that Kursan Milakovic, 2021; Nayal et al., 2022; Rayburn et al., 2022;
consumer-perceived effort expectancy of the FDPs was low. From a Sharma et al., 2021; Yap et al., 2021). For instance, online transactions
practical point of view, the design of the FDP app was believed to be have become part of people's life in many sectors including
too complicated, which affected ease of use and caused high user risk; healthcare, financial services, food delivery, etc. (Chakraborty &
hence, we suggest that FDPs improve factors related to effort expec- Paul, 2022; Chopdar, Paul, Korfiatis, & Lytras, 2022; Purohit
tancy. The gap values, sorted in descending order, were FC, SEC, SI, et al., 2022; Siddiqi et al., 2022). As an outcome, we need new theo-
UH and HE; hence, we suggest that FDPs improve the dimensions ries, methods and paradigms to carry research studies in the post-
according to the gaps by effectively allocate corresponding resources. pandemic era to analyse the new processes, patterns and problems.
Table 13 also showed that HE2 had the largest performance value Following prior calls (Paul & Bhukya, 2021), we feel the need for
and SI1 had the smallest performance value. Larger performance developing new frameworks and models to carry out future studies in
values indicated better performance of the evaluation criterion. Given this area.
that the PE1 and SI1 had the largest gap values, they should be a more
distant focus for FDPs when designing improvement measures,
followed by FC1 and UH3. A larger gap value suggests worse perfor- 6 | CONC LU SIONS
mance of the evaluation criterion and should be prioritized in
improvement measures. Therefore, we suggest that the factors should Currently, there are many FDPs in Taiwan. Determining ways to select
be improved in the following sequence: PE1–SI1–FC1–UH3–UH1– the most suitable service platform is of great interest for consumers;
SI2–SEC1–FC2–SEC2–SEC3–EE1–HE1–UH2–PE2–FC3–HE3–PE3– therefore, assisting consumers to choose a suitable FDP was the main
EE2–EE3–SI3–HE2. research motivation, purpose, and focus of the present study. The
conclusions are as follows. First, referring to the UTAUT2 model, this
study extracted 7 dimensions (performance expectancy, effort expec-
5.4 | Limitations and future research tancy, social influence, facilitating conditions, hedonic motivation,
habits and security), and 21 evaluation criteria. Second, an evaluation
This study aimed to explore the key factors that influence consumer model was then constructed that demonstrates the influences on con-
preferences in selecting FDPs and on that basis to present possible sumer FDP selection. The results showed that performance expec-
marketing strategies for improvement prioritization. Although the tancy, effort expectancy, and security were causal factors. The
research process strove to be objective and rigorous, few limitations findings are similar to those of Ramos (2022) and Cheng et al. (2021),
remain. First, the food delivery market is broad. The present study who both stressed the impacts on consumers' ongoing usage behav-
took a consumer perspective and lacked viewpoints from experts in iour. Specifically, security was found to influence social influence,
the field; hence, the conclusions may not be comprehensive. Second, facilitating conditions, hedonic motivation and daily habits; effort
since the research subjects were consumers with experience using expectancy was found to influence facilitating conditions, hedonic
food delivery apps, the perceptions of consumers without such expe- motivation, habits and security; and performance expectancy was
riences were not included. As a result, the findings may not fully found to have the strongest influence among all the dimensions that
reflect the preferences of all consumers. affected consumer choice.
The following suggestions were proposed for future research in Third, the weights of the evaluation criteria obtained through DANP
the related fields. First, this study mainly focused on consumers. were used to identify the most influential factors when consumers evalu-
Future research is suggested to extend the research subject and ated and selected an FDP. The results showed that the consumers con-
include platform operators, delivery personnel and government sidered integration into daily habits as an important factor, followed by
departments into the analysis to explore the perception of different (in descending order) hedonic motivation, facilitating conditions, security,
TSAI ET AL. 173
social influence, effort expectancy and performance expectancy. In addi- Bandura, A. (1986). Social foundations of thought and action: A social cogni-
tion, PE3, EE1, SI3, FC3, HE3, UH2 and SEC3 appeared to be perceived tive theory. Prentice-Hall, Inc.
Belarmino, A., Raab, C., Tang, J., & Han, W. (2021). Exploring the motiva-
as the most important criteria among the corresponding dimensions.
tions to use online meal delivery platforms: Before and during quaran-
Finally, this study employed a modified VIKOR to calculate the gap tine. International Journal of Hospitality Management, 96, 102983.
values of the evaluation dimensions and criteria and the ranking of the Brown, S. A., & Venkatesh, V. (2005). Model of adoption of technology in
gaps of the dimensions in descending order, as follows: PE, EE, FC, SEC, households: A baseline model test and extension incorporating house-
hold life cycle. MIS Quarterly, 29, 399–426.
SI, UH, and HE; while the criteria ranking was as follows: PE1, SI1, FC1,
Chakraborty, D., & Paul, J. (2022). Healthcare apps' purchase intention: A
UH3, UH1, SI2, SEC1, FC2, SEC2, SEC3, EE1, HE1, UH2, PE2, FC3, consumption values perspective. Technovation, 102481. https://doi.
HE3, PE3, EE2, EE3, SI3 and HE2. It can be seen that performance org/10.1016/j.technovation.2022.102481
expectancy and effort expectancy had the largest gaps vis-à-vis con- Chang, S. C., & Tsai, P. H. (2016). A hybrid financial performance evalua-
tion model for wealth management banks following the global financial
sumer expectations and, hence, should be prioritized for improvement.
crisis. Technological and Economic Development of Economy, 22(1),
21–46.
ACKNOWLEDGEMEN TS Cheng, C. C., Chang, Y. Y., & Chen, C. T. (2021). Construction of a service
The authors acknowledge the helpful comments from the editor, asso- quality scale for the online food delivery industry. International Journal
of Hospitality Management, 95, 102938.
ciate editor and the four anonymous reviewers for their comments
Cho, M., Bonn, M. A., & Li, J. (2019). Differences in perceptions about food
and suggestions which enhanced the quality of the manuscript. We
delivery apps between single-person and multi-person households.
also acknowledge the assistance and support from the Ministry of Sci- International Journal of Hospitality Management, 77, 108–116.
ence and Technology (MOST) of Taiwan, under grant MOST Chopdar, P. K., Paul, J., & Prodanova, J. (2022a). Mobile shoppers'
109-2410-H-346-001. response to Covid-19 phobia, pessimism and smartphone addiction:
Does social influence matter? Technological Forecasting and Social
Change, 174, 121249.
FUND ING INFORMATION Chopdar, P. K., Paul, J., Korfiatis, N., & Lytras, M. D. (2022b). Examining
The authors acknowledge the assistance and support from the Minis- the role of consumer impulsiveness in multiple app usage behavior
try of Science and Technology (MOST) of Taiwan, under grant MOST among mobile shoppers. Journal of Business Research, 140, 657–669.
Davis, F. D. (1986). A technology acceptance model for empirically testing
109-2410-H-346-001.
new end-user information systems: Theory and results (Doctoral disserta-
tion). Massachusetts Institute of Technology.
CONF LICT OF IN TE RE ST Davis, F. D., Bagozzi, R. P., & Warshaw, P. R. (1992). Extrinsic and intrinsic
No conflict of interest has been declared by the authors. motivation to use computers in the workplace. Journal of Applied Social
Psychology, 22(14), 1111–1132.
Erjavec, J., & Manfreda, A. (2022). Online shopping adoption during
DATA AVAI LAB ILITY S TATEMENT COVID-19 and social isolation: Extending the UTAUT model with herd
Data sharing is not applicable to this article as no new data were cre- behavior. Journal of Retailing and Consumer Services, 65, 102867.
ated or analysed in this study. Escobar-Rodríguez, T., & Carvajal-Trujillo, E. (2014). Online purchasing
tickets for low cost carriers: An application of the unified theory of
acceptance and use of technology (UTAUT) model. Tourism Manage-
ORCID ment, 43, 70–88.
Pei-Hsuan Tsai https://orcid.org/0000-0003-2376-2000 Fakfare, P. (2021). Influence of service attributes of food delivery applica-
Wei-Hung Hsiao https://orcid.org/0000-0002-1938-5444 tion on customers' satisfaction and their behavioural responses: The
IPMA approach. International Journal of Gastronomy and Food Science,
Chih-Jou Chen https://orcid.org/0000-0001-6030-8662
25, 100392.
Faqih, K. M. S., & Jaradat, M.-I. R. M. (2021). Integrating TTF and UTAUT2
RE FE R ENC E S theories to investigate the adoption of augmented reality technology
Ajzen, I. (1985). From intentions to actions: A theory of planned behavior. in education: Perspective from a developing country. Technology in
In J. Kuhl & J. Beckmann (Eds.), Action control (pp. 11–39). SSSP Society, 67, 101787.
Springer Series in Social Psychology. Farah, M. F., Ramadan, Z., & Kanso, J. (2021). Satisfying the online food
Alalwan, A. A. (2020). Mobile food ordering apps: An empirical study crave: The case of online food aggregators. International Journal of
of the factors affecting customer e-satisfaction and continued inten- Consumer Studies. https://doi.org/10.1111/ijcs.12768
tion to reuse. International Journal of Information Management, 50, Fishbein, M., & Ajzen, I. (1975). Belief, attitude, intention and behavior: An
28–44. introduction to theory and research. Addison-Wesley.
Alalwan, A. A., Dwivedi, Y. K., Rana, N. P., & Algharabat, R. (2018). Examin- Fontela, E., & Gabus, A. (1976). The DEMATEL observer, DEMATEL 1976
ing factors influencing Jordanian customers' intentions and adoption report. Geneva, Switzerland.
of internet banking: Extending UTAUT2 with risk. Journal of Retailing Gabus, A., & Fontela, F. (1973). Perceptions of the world Problematique:
and Consumer Services, 40, 125–138. Communication procedure, communicating with those bearing collective
Al-Saedi, K., Al-Emran, M., Ramayah, T., & Abusham, E. (2020). Developing responsibility, DEMATEL report no. 1. Battelle Geneva Research Center.
a general extended UTAUT model for M-payment adoption. Technol- Gölcük, _I., & Baykasog lu, A. (2016). An analysis of DEMATEL approaches
ogy in Society, 62, 101293. for criteria interaction handling within ANP. Expert Systems with Appli-
Altuntas, F., & Gok, M. S. (2021). The effect of COVID-19 pandemic on cations, 46, 346–366.
domestic tourism: A DEMATEL method analysis on quarantine deci- Gordon-Wilson, S. (2022). Consumption practices during the COVID-19
sion. International Journal of Hospitality Management, 92, 102719. crisis. International Journal of Consumer Studies, 46(2), 575–588.
174 TSAI ET AL.
Gupta, H. (2018). Evaluating service quality of airline industry using hybrid Neise, T., Verfürth, P., & Franz, M. (2021). Rapid responding to the
best worst method and VIKOR. Journal of Air Transport Management, COVID-19 crisis: Assessing the resilience in the German restaurant
68, 35–47. and bar industry. International Journal of Hospitality Management, 96,
Hoe, S. L. (2008). Issues and procedures in adopting structural equation 102960.
modeling technique. Journal of Applied Quantitative Methods, 3(1), Okumus, B., Ali, F., Bilgihan, A., & Ozturk, A. B. (2018). Psychological fac-
76–83. tors influencing customers' acceptance of smartphone diet apps when
Hong, C., Choi, H., Choi, E.-K., & Joung, H.-W. (2021). Factors affecting ordering food at restaurants. International Journal of Hospitality Man-
customer intention to use online food delivery services before and agement, 72, 67–77.
during the COVID-19 pandemic. Journal of Hospitality and Tourism Opricovic, S. (1998). Multicriteria optimization of civil engineering systems.
Management, 48, 509–518. Faculty of Civil Engineering (in Serbian).
Horng, J. S., Liu, C. H. S., Chou, S. F., Tsai, C. Y., & Hu, D. C. (2018). Devel- Ou Yang, Y. P., Shieh, H. M., & Tzeng, G. H. (2008). A novel hybrid MCDM
oping a sustainable service innovation framework for the hospitality model combined with DEMATEL and ANP with applications. Interna-
industry. International Journal of Contemporary Hospitality Management, tional Journal of Operations Research, 5(3), 160–168.
30(1), 455–474. Paul, J., & Bhukya, R. (2021). Forty-five years of International Journal
Huang, C. Y., & Kao, Y. S. (2015). UTAUT2 based predictions of factors of Consumer Studies: A bibliometric review and directions for
influencing the technology acceptance of phablets by DNP. Mathemat- future research. International Journal of Consumer Studies, 45(5),
ical Problems in Engineering, 2015, 603747. 937–963.
Hwang, J., Kim, J. J., & Lee, K. W. (2021). Investigating consumer innova- Peng, K. H., & Tzeng, G. H. (2013). A hybrid dynamic MADM model for
tiveness in the context of drone food delivery services: Its impact on problem-improvement in economics and business. Technological and
attitude and behavioral intentions. Technological Forecasting and Social Economic Development of Economy, 19(4), 638–660.
Change, 163, 120433. Purohit, S., Arora, R., & Paul, J. (2022). The bright side of online consumer
Jadil, Y., Rana, N. P., & Dwivedi, Y. K. (2021). A meta-analysis of the behavior: Continuance intention for mobile payments. Journal of Con-
UTAUT model in the mobile banking literature: The moderating role of sumer Behaviour. https://doi.org/10.1002/cb.2017
sample size and culture. Journal of Business Research, 132, 354–372. Ramesh, R., Prabhu, S. V., Sasikumar, B., Devi, B. S. K., Prasath, P., &
Kim, K., Bonn, M. A., & Cho, M. (2021). Clean safety message framing as Kamala, S. P. R. (2021). An empirical study of online food delivery ser-
survival strategies for small independent restaurants during the vices from applications perspective. Materials Today: Proceedings.
COVID-19 pandemic. Journal of Hospitality and Tourism Management, https://doi.org/10.1016/j.matpr.2021.05.500
46, 423–431. Ramírez-Correa, P., Rondán-Cataluña, F. J., Arenas-Gaitán, J., & Martín-
Kumar, S., & Shah, A. (2021). Revisiting food delivery apps during COVID- Velicia, F. (2019). Analysing the acceptation of online games in mobile
19 pandemic? Investigating the role of emotions. Journal of Retailing devices: An application of UTAUT2. Journal of Retailing and Consumer
and Consumer Services, 62, 102595. Services, 50, 85–93.
Kursan Milakovic, I. (2021). Purchase experience during the COVID-19 Ramos, K. (2022). Factors influencing customers' continuance usage inten-
pandemic and social cognitive theory: The relevance of consumer vul- tion of food delivery apps during COVID-19 quarantine in Mexico.
nerability, resilience, and adaptability for purchase satisfaction and Britsh Food Journal, 134(3), 833–852.
repurchase. International Journal of Consumer Studies, 45(6), 1425– Rao, S. H. (2021). A hybrid MCDM model based on DEMATEL and ANP
1442. for improving the measurement of corporate sustainability indicators:
Kwak, N. K., Lee, C. W., & Kim, J. H. (2005). An MCDM model for media A study of Taiwan high speed rail. Research in Transportation Business &
selection in the dual consumer/industrial market. European Journal of Management, 41, 100657.
Operational Research, 166(1), 255–265. Rasty, F., Mirghafoori, S. H., Ardekani, S. S., & AJdari, P. (2021). Trust bar-
Lin, C. L., Shih, Y. H., Tzeng, G. H., & Yu, H. C. (2016). A service selection riers to online shopping: Investigating and prioritizing trust barriers in
model for digital music service platforms using a hybrid MCDM an intuitionistic fuzzy environment. International Journal of Consumer
approach. Applied Soft Computing, 48, 385–403. Studies, 45(5), 1030–1046.
Lock, S. (2020). Daily year-on-year impact of COVID-19 on global restau- Ray, A., & Bala, P. K. (2021). User generated content for exploring factors
rant dining. Statista https://www.statista.com/statistics/1103928/ affecting intention to use travel and food delivery services. Interna-
coronavirus-restaurant-visitation-impact/ tional Journal of Hospitality Management, 92, 102730.
MacCallum, R. C., & Austin, J. T. (2000). Applications of structural equation Ray, A., Dhir, A., Bala, P. K., & Kaur, P. (2019). Why do people use food
modeling in psychological research. Annual Review of Psychology, 51, delivery apps (FDA)? A uses and gratification theory perspective. Jour-
201–226. nal of Retailing and Consumer Services, 51, 221–230.
Mardani, A., Jusoh, A., Nor, K. M. D., Khalifah, Z., Zakwan, N., & Rayburn, S. W., McGeorge, A., Anderson, S., & Sierra, J. J. (2022). Crisis-
Valipour, A. (2015). Multiple criteria decision-making techniques and induced behavior: From fear and frugality to the familiar. International
their applications—A review of the literature from 2000 to 2014. Eco- Journal of Consumer Studies, 46(2), 524–539.
nomic Research, 28(1), 516–571. Rogers, E. M. (1962). Diffusion of innovation (1st ed.). Free Press of
Mavi, R. K., & Standing, C. (2018). Cause and effect analysis of business Glencoe.
intelligence (BI) benefits with fuzzy DEMATEL. Knowledge Manage- Roh, M., & Park, K. (2019). Adoption of O2O food delivery services in
ment Research & Practice, 16(2), 245–257. South Korea: The moderating role of moral obligation in meal prepara-
Mehrolia, S., ALagarsamy, S., & Solaikutty, V. M. (2021). Customers tion. International Journal of Information Management, 47, 262–273.
response to online food delivery services during COVID-19 outbreak Saaty, T. L. (1996). Decision making with dependence and feedback: The ana-
using binary logistic regression. International Journal of Consumer Stud- lytic network process. RWS Publications.
ies, 45(3), 396–408. Sharma, G. D., Thomas, A., & Paul, J. (2021). Reviving tourism industry
MIC. (2020). https://mic.iii.org.tw/news.aspx?id=573 post-COVID-19: A resilience-based framework. Tourism Management
MIC (2017). https://mic.iii.org.tw/news.aspx?id=457 Perspectives, 37, 100786.
Nayal, P., Pandey, N., & Paul, J. (2022). Covid-19 pandemic and consumer- Sheehan, D., Hardesty, D. M., Ziegler, A. H., & Chen, H. (2019). Consumer
employee-organization wellbeing: A dynamic capability theory reactions to price discounts across online shopping experiences. Jour-
approach. Journal of Consumer Affairs, 56(1), 359–390. nal of Retailing and Consumer Services, 51, 129–138.
TSAI ET AL. 175
Siddiqi, U. I., Akhtar, N., & Islam, T. (2022). Restaurant hygiene attributes amid the COVID-19 crisis. International Journal of Hospitality Manage-
and consumers' fear of COVID-19: Does psychological distress mat- ment, 94, 102855.
ter? Journal of Retailing and Consumer Services, 67, 102972. Yap, S. F., Xu, Y., & Tan, L. (2021). Coping with crisis: The paradox of tech-
Sipahi, S., & Timor, M. (2010). The analytic hierarchy process and analytic nology and consumer vulnerability. International Journal of Consumer
network process: An overview of applications. Management Decision, Studies, 45(6), 1239–1257.
48(5), 775–808. Yeo, V. C. S., Goh, S. K., & Rezaei, S. (2017). Consumer experiences,
Sofiyabadi, J., Kolahi, B., & Valmohammadi, C. (2016). Key performance indi- attitude and behavioural intention toward online food delivery
cators measurement in service business: A fuzzy VIKOR approach. Total (OFD) services. Journal of Retailing and Consumer Services, 35,
Quality Management & Business Excellence, 27(9–10), 1028–1042. 150–162.
Song, H., Ruan, W. J., & Jeon, Y. J. J. (2021). An integrated approach to the Zanetta, L. D., Hakim, M. P., Gastaldi, G. B., Jucá Seabra, M., Rolim, P. M.,
purchase decision making process of food-delivery apps: Focusing on Nascimentio, L. G. P., Medeiros, C. O. M., & da Cunha, D. T. (2021).
the TAM and AIDA models. International Journal of Hospitality Manage- The use of food delivery apps during the COVID-19 pandemic in
ment, 95, 102943. Brazil: The role of solidarity, perceived risk, and regional aspects. Food
Statista (2019). Global online food delivery report. https://www.statista. Research International, 149, 110671.
com/outlook/374/100/online-food-delivery/worldwide#market- Zhao, Y., & Bacao, F. (2020). What factors determining customer continu-
revenue ingly using food delivery apps during 2019 novel coronavirus pan-
Suhartanto, D., Ali, M. H., Tan, K. H., Sjahroeddin, F., & Kusdibyo, L. demic period? International Journal of Hospitality Management, 91,
(2019). Loyalty toward online food delivery service: The role of e- 102683.
service quality and food quality. Journal of Foodservice Business
Research, 22(1), 81–97.
Tamilmani, K., Rana, N. P., Wamba, S. F., & Dwivedi, R. (2021). The AUTHOR BIOGRAPHIES
extended unified theory of acceptance and use of technology
(UTAUT2): A systematic literature review and theory evaluation. Inter-
national Journal of Information Management, 57, 102269. Pei-Hsuan Tsai is an Assistant Professor of Department of Mar-
Tandon, A., Kaur, P., Bhatt, Y., Mäntymäki, M., & Dhir, A. (2021). Why do keting and Logistics Management, National Penghu University of
people purchase from food delivery apps? A consumer value perspec- Science and Technology. He received his PhD in Management
tive. Journal of Retailing and Consumer Services, 63, 102667.
from the National Taiwan University of Science and Technology
Taylor, S., & Todd, P. A. (1995). Understanding information technology
usage: A test of competing models. Information Systems Research, 6(2), (NTUST). He has published several papers in international journals
144–176. including journals such as Journal of Retailing and Consumer Ser-
Thompson, R. L., Higgins, C. A., & Howell, J. M. (1991). Personal comput- vices, Evaluation and Program Planning, Journal of Policy Modeling,
ing: Toward a conceptual model of utilization. MIS Quarterly, 15(1),
SAGE Open, Curator: The Museum Journal, Technological and Eco-
125–143.
Tsai, P. H. (2020). Strategic evaluation criteria to assess competitiveness nomic Development of Economy, IMA Journal of Management Math-
of the service industry in Taiwan. Journal of Policy Modeling, 42, 1287– ematics, European Journal of Operational Research and Service
1309. Industries Journal. His research interests include multiple-attribute
Tsai, P. H., & Chang, S. C. (2013). Comparing the Apple iPad and non-
decision-making, management sciences, consumer decision
Apple camp tablet PCs: A multicriteria decision analysis. Technological
and Economic Development of Economy, 19(Suppl 1), S256–S284. behaviour, green marketing, retailing store and marketing
Tsai, P. H., & Chen, C. J. (2021). Entertainment in retailing: Challenges and strategy.
opportunities in the TV game console industry. Journal of Retailing and
Consumer Services, 60, 102503. Wei-Hung Hsiao is an Assistant Professor in the Department of
Tsai, P. H., Lin, G. Y., Zheng, Y. L., Chen, Y. C., Chen, P. Z., & Su, Z. C. Distribution Management at National Taichung University of Sci-
(2020). Exploring the effect of Starbucks' green marketing on con- ence and Technology. He received a PhD in Information and man-
sumers' purchase decisions from consumers' perspective. Journal of
agement from National Chung Cheng University. He has
Retailing and Consumer Service, 56, 102162.
Tsai, P. H., Tang, J. W., & Chen, C. J. (2022). Partnerships that go places: published several papers in International Journal of Information
How to successfully market products from vendor partners at retail Management, International Journal of Mobile Communications, Jour-
stores from the vendors' perspective. Journal of Retailing and Consumer nal of Enterprise Information Management, Cyberpsychology, Behav-
Services, 64, 102806.
iour and Social Networking, Behaviour and Information Technology,
Venkatesh, V., & Davis, F. D. (2000). A theoretical extension of the tech-
nology acceptance model: Four longitudinal field studies. Management Systems Research and Behavioural Science, and others. His primary
Science, 46(2), 186–204. research interests include smart retail, smart logistic and con-
Venkatesh, V., Morris, M. G., Davis, G. B., & Davis, F. D. (2003). User sumer behaviour.
acceptance of information technology: Toward a unified view. MIS
Quarterly, 27(3), 425–478. Chih-Jou Chen is a Professor at the Department of Marketing and
Venkatesh, V., Thong, J. Y. L., & Xu, X. (2012). Consumer acceptance and Logistics Management, National Penghu University of Science
use of information technology: Extending the unified theory of accep-
and Technology in Taiwan. He has extensive teaching, training
tance and use of technology. MIS Quarterly, 36(1), 157–178.
Wang, Y. S., Tseng, T. H., Wang, W. T., Shih, Y. W., & Chan, P. Y. (2019). and consulting experience in service innovation management,
Developing and validating a mobile catering app success model. Inter- new product development and supply chain management. His
national Journal of Hospitality Management, 77, 19–30. research interests include intelligent agent technologies, market
Yang, F. X., Li, X., Lau, V. M. C., & Zhu, V. Z. (2021). To survive or to thrive?
competition and strategy, services marketing and management.
China's luxury hotel restaurants entering O2O food delivery platforms
176 TSAI ET AL.
He has authored numerous papers and his research has appeared

How to cite this article: Tsai, P.-H., Hsiao, W.-H., & Chen, C.-J.
in the Journal of Retailing and Consumer Services, Computers in
(2023). Which food delivery platforms are winning the
Human Behavior, International Journal of Organizational Analysis,
restaurant food delivery wars? Analysis from a consumer
Information and Management, Journal of Business Ethics, Journal of
perspective. International Journal of Consumer Studies, 47(1),
Business and Industrial Marketing, Management Decision, IEEE
155–176. https://doi.org/10.1111/ijcs.12816
Access and conference proceedings.

2022, Tsai, Deli Fee Is Less Considered Than Deli Perfor

Uploaded by

Copyright:

Available Formats

2022, Tsai, Deli Fee Is Less Considered Than Deli Perfor

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

2022, Tsai, Deli Fee Is Less Considered Than Deli Perfor

Uploaded by

Copyright:

Available Formats

Received: 1 November 2021 Revised: 19 April 2022 Accepted: 20 April 2022

Which food delivery platforms are winning the restaurant food

Pei-Hsuan Tsai1 | Wei-Hung Hsiao2 | Chih-Jou Chen1

1 | I N T RO DU CT I O N risen to become a distinctive dining option for consumers (Cheng

TABLE 2 The UTAUT framework and the eight theories included

Foodtoall Taiwan/2017 2018 1000+ 1992 Davis et al. Motivational Model(MM)

2.3.6 | Habits 3.1 | DEMATEL

TABLE 3 Descriptions of the evaluation criteria for FDPs

Dimensions Criteria Definitions

Construct an evaluation model for

The UTAUT2 Model

Determine the evaluation indicators Consumer experience interviews

Develop the research framework based on

Establish the research framework Design and distribute the causality

Determine the distribution targets

Collect the responses of the first survey and

Use the DANP to calculate the influential

Determine the distribution targets

Collect the responses of the second survey

Formulate improvement strategies based

FIGURE 1 Research design framework

4.2 | DEMATEL analysis

TABLE 5 Normalized matrix D of

TABLE 6 Total influence matrix T of

FIGURE 2 The INRM of the evaluation dimensions

SI3 11.217 11.279 11.248

T A B L E 1 1 Strength of influence and

TABLE 12 Ranking of influential weights of the evaluation dimensions and criteria

The findings of this study have significant management implications

He has authored numerous papers and his research has appeared

You might also like

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.