Evaluating Consumers’ Adoption of Mobile Technology for Grocery Shopping: An Application of Technology Acceptance Model

Introduction

India still practices the culture where the merchants deliver groceries and household staples to consumers’ residence (AC Nielsen, 2015 April). The similar concept of home delivery shopping is re-emerging across the world but with a twist of use of mobile phones. Mobile applications (mobile apps) are found to be the most significant way for Indians to access e-commerce (Livemint, 2016). Mobile has become an important channel for shoppers as they research and purchase products (eMarketer, 2016 January). According to AC Nielsen Report (2015 April), the number of Indians opting for shopping on mobile apps is on rise. The number of smartphone owners using mobile apps of electronic commerce companies has gone up from 1680 in May 2014 to 4320 in May 2015 (AC Nielsen, 2015 April). Internet and Mobile Association of India (IAMAI), 2015, reported that the number of users using internet on mobile is almost 58.5 per cent of the entire Internet accessed in India during March 2015, which is expected to cross 80 per cent by June 2016. This increase in purchase and use of smartphone can be credited to the reducing average selling price (ASP) of smartphones (Neeraj, 2016). In 2015, the consumers’ share of using Internet using has increased from 54 per cent to 64 per cent as compared to previous year.

India is soon to overtake the USA and become the world’s second-largest market for smartphones excee ding 200 million smartphone users (eMarketer, 2016). The various sectors have received an overwhelming response in fashion industry, consumer durable, electronic items, etc., but m-shopping is not so popular and is in nascent stage in the grocery sector. Electronic grocery (e-grocery) allows consumers to purchase grocery online through various mediums such as cash on delivery and online payments without any time and location constraints. The e-grocery vendors provide home delivery of various products including staples, edible oil, beverages, milk and dairy products, baby and child care, cereals and breakfast, snacks, etc., which are considered as essentials and are bought in high quantities (E-Grocery Market in India, 2015). One-quarter of global respondents are already ordering grocery products online for home delivery, and more than half (55%) are willing to use it in the future (AC Nielsen, 2015 April). The continued growth in smartphone subscriptions is expected to lead to an accelerated growth in data usage while monthly mobile data consumption is expected to increase 18-fold by 2020 over current levels (Livemint, 2016).

Review of Literature

Shopping through mobile apps is gaining high importance in the field of marketing and retailing (Wang, Malthouse & Krishnamurthi, 2015). This is a comparative new area of study (Annexure 1), but still it poses an immense growth in future and literature is found limited in this field (Groß, 2015; Holmes, Byrne & Rowley, 2013). It was found during literature review that most studies were conducted in USA, Taiwan, China, etc., and are found very less in other parts of the world (Groß, 2015). This in-depth review of literature has identified that the studies related to the adoption mobile phone towards grocery shopping have not been conducted for Indian consumers.

Mobile commerce (MC), most popularly known as m- commerce, is the use of mobile phone for making purchases. Wong and Sheng (2012) defined m-commerce as ‘any monetary transactions related to purchases of goods or services through internet-enabled mobile phones or over the wireless telecommunication network’. Mobile device seems to assist the consumer for doing his shopping (Yang, 2010). There are two major theory considered for the study—theory of reasoned action (TRA) and technology acceptance model (TAM). The TAM was designed on the basis of TRA (Fishbein & Ajzen, 1975) to predict the likelihood that a new technology will be adopted with in a group of individuals or organizations (Davis, Bagozzi & Warshaw, 1989). The aim of TAM was to provide a model to the researchers that can help them analyse the cause of rejection or acceptance of technology. The TAM was meant to explain the behaviour of computer usage and was found to be well suited for modelling computer acceptance (Davis, 1989).

With the advancement of Internet on mobile, the consumers are shifting towards MC or the use of mobile phones for shopping. Customers use mobile devices because the technology provides convenient access, which leads them to incorporate m-shopping into their habitual routines (Wang et al., 2015). Shopping through mobile is examined in various field. Intention to buy tourism products using mobile studied was on Korean consumers, and results show that perceived usefulness and site attachment were important mediators of tourism shopping through apps (Kim, Chung, Lee & Preis, 2016). Similar study was conducted on Hong Kong consumers to test the use of MC which showed that attitude of users towards MC services was found to be impactful (Fong & Wong, 2015).

The TAM has been found to successfully explain adoption of numerous technologies that range from software package to different online services. Moon and Domina (2015) conducted an exploratory study focused on USA and South Korean consumers’ intention to use fashion mobile applications (FMAs) to purchase fashion products and how that intention is affected by cultural differences. Attitude towards using FMAs was the most significant predictor of intention to purchase fashion products through FMAs for both South Korean and American participants. Hu, Chau, Sheng and Tam (1999) used TAM to explain physicians’ decisions to accept telemedicine technology in the healthcare context. Mckechnie, Winklhofer and Ennew (2006) studied TAM to understand consumers behaviour towards use of Internet in area of financial services. Ko, Kim and Lee (2009) examined the potential of a consumer technology adoption model in fashion products. Park (2009) tested adoption of e-learning by students based on the TAM. Willman (2015) used TAM 3 (extension of TAM) to compare the Western and Eastern culture towards using of retail shopping mobile apps and how consumers were motivated to use them. It was concluded that Western consumers do not use retail app or use it very less while on the counterpart, Eastern consumers use it more frequently. Amaro and Duarte (2015) proposed a model to find out the reasons that affect the intentions of consumer to purchase travel online using constructs from TAM, innovation diffusion theory (IDT) and theory of planned behaviour (TPB) constructs.

OPEN IN VIEWER

Figure 1.

Source: Authors’ compilation.

Venkatesh and Morris (2000) extended TAM and tested in three different organizations among 246 employees using three measurements taken over a 3-month period. The proposed model was strongly supported at all points of measurement, and explained up to 60 per cent of the variance was explained. Venkatesh and Davis (2000) developed and tested a theoretical extension of the TAM that explains perceived usefulness and usage intentions in terms of social influence and cognitive instrumental processes. The study was conducted on accepting of technology in doing work by floor supervisor, people in financial services, accounting services and investment banking. The study incorporated additional theoretical constructs spanning social influence processes (subjective norm, voluntariness and image) and cognitive instrumental processes (job relevance, output quality, result demonstrability and perceived ease of use). Li and Huang (2009) extended TAM model by adding perceived risk in existing TAM model. The results suggest the need for consi-deration of perceived risk as an antecedent in the technology acceptance. Tseng and Hsu (2010) applied electronic word of mouth (eWOM) communications and TAM towards consumers intention towards EeePC (innovative technology). Ajzen (2011) clarified behavioural intentions using vari ables perceived behavioural control, perceived self-efficacy, perceived controllability, subjective norm, normative belief, attitude and intention. Wong & Sheng (2012) analysed the various reasons that motivated consumers to adopt m-shopping. Al-rahmi & Othman (2013) used TAM to analyse the student’s perception towards use of social media for collaborative learning by testing TAM Model for students in Malaysia.

OPEN IN VIEWER

Figure 2.

Source: Adapted from technology acceptance model (Davis, 1989).

OPEN IN VIEWER

Table 1.

Definition of Constructs

Construct	Model	Definitions in Context to Study	Reference
Perceived usefulness (U)	TAM	The extent to which potential users expect that mobile app will benefit their grocery shopping.	Davis (1989)
Perceives ease of use (EOU)		The degree to which a consumer believes that using a mobile app would be free of effort.
Behavioural intention (BI)	TRA	How likely one is to buy grocery using mobile application	Fishbein and Ajzen (1975)
Attitude (A)		Feeling of an individual (positive or negative) about buying grocery through mobile application

Source: Authors’ compilation.

Wu and Wang (2005) presented an extended TAM that integrates IDT, perceived risk and cost into the TAM to investigate what determines user MC acceptance. Extensive literature reviews were conducted by Surendran (2012) and Chuttur (2009). An extensive survey of extant related studies was synthesized into online shopping acceptance model (OSAM) to explain consumer acceptance of online shopping by Zhou, Dai and Zhang (2007). Shankar and Balasubramanian (2009) found in their literature review that marketers would have to work diligently to enhance customer receptiveness to mobile marketing in the broader population and, in the short run, need to focus on customers who are receptive to these initiatives.

The conceptual model is based on TAM as shown in Figure 2, and the constructs are explained in Table 1.

Research Methodology

Data Analysis

The PLS-SEM was used for testing of hypothesis as it has few assumptions about data distribution (Hair, Hult, Ringle & Sarstedt, 2013; Hair, Ringle & Sarstedt, 2011; Hair, Sarstedt, Pieper & Ringle, 2012). The PLS does not hold assumptions of normality for data as it uses method of bootstrapping for estimating parameters (Gefen, Rigdon & Straub, 2011; Henseler, Ringle & Sarstedt, 2012). Normality was therefore not checked. It is also considered better for testing constructs with fewer variables. Since the constructs have been measured by using three items or four items, this characteristic seems to be relevant. The PLS is more advanced technique as it means testing regression equation multiple times. The constructs were reflective in nature, and the items were predefined. The PLS-SEM is considered to be a soft modelling approach as it has delivered good results with less assumptions. The PLS is used for data analysis over covariance based structural modeling (CB-SEM) as the aim of the study was to identify key driver constructs not theory testing or comparison (Hair et al., 2011). Also, the software PLS-SEM website clarifies use of PLS-SEM over CB-SEM using the TAM model  (https://www.smartpls.com/documentation/learn-pls-sem-and-smartpls/pls-sem-compared-with-cbsem).

OPEN IN VIEWER

Table 2.

Constructs and Its Measures

Constructs	Notation	Items	Developer
Perceived usefulness	PU1	I believe that the use of a mobile device would make my grocery shopping process more effective.	Fröhlke and Pettersson 2015
	PU2	I believe that the use of a mobile device would make my grocery shopping process more convenient.
	PU3	I think that I would save time by using a mobile device while shopping for groceries.
	PU4	I believe that, in general, using a mobile device in my grocery shopping process would have been useful.
Perceived ease of use	PEOU1	It is easy to purchase grocery through mobile shopping.	Fröhlke and Pettersson (2015)
	PEOU2	It is easy to learn how to purchase through mobile shopping.
	PEOU3	It is easy to use mobile shopping.
	PEOU4	I think it would be easy to use a mobile device in the grocery shopping process.
Attitude towards mobile app	A1	I believe it would be advantageous to use my mobile device in my grocery shopping process.	Fishbein and Ajzen (1975)
	A2	I think it would be a good idea to use a mobile device when shopping for groceries.
	A3	I think it would be positive to be able to use my mobile device when shopping for groceries.
Intention to use mobile app	BI1	I intend to keep using of mobile app for buying groceries in the future.	Al-rahmi & Othman (2013)
	BI2	I intend to increase the use of my mobile app in the future.
	BI3	I intend to recommend my friends to using of mobile app for grocery shopping in the future.

Source: Authors’ compilation.

Results

Socio-demographic Characteristics

The demographic profile of respondents is subcategorized as respondent’s gender, age, marital status, occupation and location. A brief description of respondent’s demographic characteristics is presented in Table 3. There are 52.1 per cent females as compared to 47.9 per cent male respondents. In terms of age, most significant respondents lie in the age group of 18–29 constituting about 70.2 per cent of the respondents. The data were skewed as 56.4 per cent of respondents were married. The respondents belong to various occupations showing diversity, but the respondents majorly were salaried. In terms of state of residence, the large share of responses was from national capital region (NCR) constituting about 30.9 per cent of responses.

OPEN IN VIEWER

Table 3.

Descriptive Statistics of Respondents’ Characteristics

Variable	Category	Frequency	Percentage
Gender	Female Male	201 145	58.1 41.9
Age	18–29 30–39 40–49	246 78 22	71.1 22.5 6.4
Marital status	Bachelor Married	205 141	59.2 40.8
Occupation	Private job Government job Student Homemaker Self-employed	160 08 120 32 26	46.2 2.3 34.6 9.2 7.5
State	Assam Delhi Haryana Madhya P Maharashtra NCR Punjab Rajasthan Uttar Pradesh Uttarakhand	8 15 2 4 52 101 59 8 82 14	2.3 4.3 0.6 1.2 15 29.2 17.1 2.3 23.7 4
Total		346	100

Source: Authors’ compilation.

Model Assessment

As the model was based on developed theory, therefore, PLS-SEM was used for evaluation. The SEM is considered to be ‘the dominant multivariate technique’ (Hair, Black, Babin, Anderson & Tatham, 2013). In PLS-SEM, the model is evaluated in two steps. The first step consists of assessment of measurement which includes evaluating reliability and validity. The second step includes assessment of structure which conducts the evaluation of hypothesis.

The path diagram was constructed using the PLS tools and were connected as hypothesized in the theory as shown in Figure 1. At the first step, PLS algorithm was calculated. All the factors showed significant loadings. Further bootstrapping was done to check that the test on sample is the true representation of population. Bootstrapping is the process to estimate t-values of item loadings for outer model and path coefficients for inner model. Bootstrapping was run on 500 subsamples (Hair et al., 2014) at two tail-test. The values of t were found to be >1.96 at 0.5 level of significance.

Measurement Model

Assessment of Measurement (Outer model)

The first step involves identification of constructs as formative or reflective constructs. All the four variables of study are identified as reflective constructs. The reflective constructs are measured by determining indicator reliability, convergent validity, internal consistency and discrimi nant validity (Hair et al., 2011; Henseler, Ringle & Sinkovics, 2009). The reliability of items was measured by Cronbach’s Alpha (Table 4). The value of Cronbach’s alpha seems to be robust as these values are higher than 0.7. Many researchers (Hair et al., 2011; Henseler et al., 2009) consider composite reliabilities as more suitable for PLS- SEM than Cronbach’s Alpha. The composite reliability of constructs ranges from 0.92 to 0.82 (Table 4). The values between 0.70 and 0.95 are considered ‘satisfactory to good’ (Hair et al., 2014). To evaluate convergent validity, average variance extracted (AVE) of each variable was calculated, as suggested by Fornell and Larcker (1981). The AVE for all items shows convergent validity as the value exceeds 0.50 (Table 4). The factor loadings were greater than 0.5 and, hence, found significant. These values were also found significant during bootstrapping. Discriminant validity was used to assess if a construct is more strongly related to its own measure than with any other constructs by examining the overlap in variance by comparing the AVE of each construct with the squared correlations among construct (Fornell & Larcker, 1981). The diagonal elements represent AVE of construct, and other values represent the correlation between constructs (Table 5). The square root of each construct’s AVE was found to be higher than its correlations with any other construct. Therefore, it can be concluded that the constructs share high variances among themselves as compared to other constructs, thus, supporting the adequate discriminant validity of the scales (Henseler et al., 2009). The results show that the constructs fit well to the model and the reliability and validity is firm.

OPEN IN VIEWER

Table 4.

Measurement Statistics of Constructs

Construct	Mean	SD	Indicator loading	t-Value	Composite Reliability	Cronbach’s Alpha	AVE
Perceived usefulness PU1 PU2 PU3 PU4 Perceived ease of use PEOU1 PEOU2 PEOU3 PEOU4 Attitude A1 A2 A3 Behavioural intention BI1 BI2 BI3	2.598 2.468 2.208 2.373 2.324 2.092 1.939 2.298 2.327 2.457 2.431 2.616 2.023 2.419	1.31 1.09 1.31 1.01 1.06 1.21 1.17 1.07 1.12 1.23 1.17 1.09 1.07 1.15	0.916 0.891 0.844 0.845 0.739 0.813 0.725 0.894 0.885 0.920 0.906 0.871 0.547 0.892	77.251 65.110 42.219 57.228 24.744 29.629 19.569 75.392 50.123 92.098 78.810 43.821 7.243 79.250	0.93 0.82 0.87 0.92	0.88 0.73 0.81 0.897	0.81 0.61 0.63 0.76

Source: Authors’ compilation.

Note: *t-values are obtained by conducting bootstrapping on 500 samples and at 0.05% level of significance.

Test of Collinearity

Collinearity is the group of points present on a single line. It is a phenomenon in which two or more predictor variables in a multiple regression model are highly correlated, meaning that one can be linearly predicted from the others with a substantial degree of accuracy. The collinearity is measured by VIF values. The values >5 are indicator of collinearity among the indicators. In the given model, all the VIF values were less than 5 which shown absence of collinearity (Table 6).

OPEN IN VIEWER

Table 5.

Discriminant Validity of the Constructs: Correlations between Constructs (Fornell–Larcker Criteria, 1981)

	A	BI	PEOU	PU
A	0.904
BI	0.650	0.786
PEOU	0.677	0.554	0.795
PU	0.765	0.677	0.631	0.875

Source: Authors’ compilation.

OPEN IN VIEWER

Table 6.

Test of Collinearity

	A	BI	PEOU	PU
A		2.415
BI
PEOU	1.662			1.00
PU	1.662	2.415

Source: Authors’ compilation.

OPEN IN VIEWER

Table 7.

Results of PLS-SEM

Hyp.	Hypotheses	t-Value	p-Value	Beta	Result
H1	PEOU is positively related to PU	18.881	0.000	0.631	Supported
H2	PEOU is positively related to attitude	7.935	0.000	0.323	Supported
H3	PU is positively related to attitude	16.12	0.000	0.562	Supported
H4	Attitude is positively related to intention	4.720	0.000	0.317	Supported
H5	PU is positively related to intention	6.262	0.000	0.435	Supported
H6	Attitude mediates the relationship between PU and BI	Partial mediation (Table 8)			Supported
H7	PU mediates the relationship between PEOU and BI	Partial mediation (Table 8)			Supported

Source: Authors’ compilation.

Assessment of Structure

The outer model provided acceptable results for reliability and validity. The inner model was examined to assess the relationship between variables by testing hypothesis. The analysis of internal structure supported all five hypotheses as shown in Table 7 and Figure 3 as the t-values are found to be significant (t > 1.96).

The strongest relationship was found between perceived ease of use and perceived usefulness ( β = 0.631, p = 0.000). This means that if the consumer finds the use of mobile phone easy for purchasing grocery, he is more likely to find it useful as well. The second strongest relationship was found between perceived usefulness and attitude ( β = 0.562, p = 0.000). It indicates that if the consumer finds the mobile technology to be useful for his grocery shopping, he will develop a favourable attitude towards it. The hypothesis regarding perceived usefulness and behavioural intention was also supported ( β = 0.435, p = 0.000). This means that if a user of mobile app perceives mobile app as useful, he will have positive intention towards the purchase of grocery using it. The relationship of perceived ease of use and attitude was found significant ( β = 0.323, p =0.000). This indicates that if the consumer finds the technology to be easy for usage, he develops a positive attitude towards it. The next hypothesis examined the relationship between attitude and behavioural intention which was also supported ( β = 0.317, p = 0.000). This indicates that if a consumer develops a positive attitude for cause of a technology, he may intend to use it in near future. Mediation effects were analysed by measuring direct and indirect effect (Table 8). According to Hair, Ringle & Sarstedt (2013), value adjustment factor (VAF) is measured by division of indirect effect by total effect. If VAF < 20 per cent, no mediation, if VAF > 20 per cent but <80 per cent, partial mediation and if VAF > 80 per cent, full mediation. The results show that perceived usefulness partially mediates the relationship between perceived ease of use and behavioural intention, thus supporting the sixth hypothesis. The seventh hypothesis was also supported showing that attitude mediates the relationship between usefulness and intention to use mobile for grocery shopping.

Further, cross tabulation was conducted to draw inferences about impact of demographical factors on consumers usage of mobile phones. The results of demographical analysis were as under:

Females preferred to use mobile apps over males for doing mobile shopping but in case of grocery shopping, males prefer to use mobile apps.

The younger generation (age 18–29) prefers shopping through mobile for both general and grocery items in comparison to other age groups.

The respondents engaged in private job constituted the largest segment of people shopping using mobile, followed by students. However, in case of grocery shopping, the largest portion was constituted by students followed by housewife.

The consumers were indifferent of marital status for shopping through mobile apps, but in case of m-shopping for grocery items, married people agreed to use it more as compared to unmarried people.

The descending order of preference of shopping through mobile is summarized as unmarrried female → married male → married female → unmarried male.

The descending order of preference of grocery shopping through mobile is summarized as married female → unmarried male → married male → unmarried male.

Conclusions and Implications

Results of the study demonstrate the application of TAM in accessing acceptance of mobile shopping for grocery items in India. The results are in agreement with the previous studies such as use of TAM in online grocery shopping (Kurnia & Chien, 2003), understanding of students’ intention to use e-learning (Park, 2009). All of the constructs proposed in TAM and their relationships were found to be relevant in this study. This study will enrich the literature in the field of electronic commerce-enabled adoption of technology using TAM which is currently still limited (Gefen & Straub, 2000; Kurnia & Chien, 2003). The TAM model was successful in explaining adoption of technology in the various field, yet this model was not tested in grocery sector (Park, 2009). The contribution of research can be summarized as a validation of the TAM given by Davis in 1989, in a completely new empirical context in terms of country (India), population (sample is a mix of student, salaried and home makers), area (grocery shopping) and technology used (mobile).

OPEN IN VIEWER

Table 8.

Mediation Results

Hyp	Relationship	Direct Effect	Indirect Effect	Total Effect	VAF	Results
H6	PEOU → PU → BI	0.579	0.427	1.006	0.424	Partial mediation
H7	PU → A → BI	0.677	0.502	1.179	0.426	Partial mediation

Source: Authors’ compilation.

OPEN IN VIEWER

Figure 3.

Source: Authors’ compilation.

OPEN IN VIEWER

Table 9.

R-square Values

	R-Square	R-Square Adjusted
A	0.649	0.647
BI	0.500	0.467
PU	0.398	0.396

Source: Authors’ compilation.

It was found that tier II city consumers prefer to shop using mobile app in comparison to tier I consumers. This result is in with agreement to AC Nielsen Report (2015 July) that time spent by users in non-metros (116 min/month) was found higher than in metros (109 min/month).

The study was able to test the adoption of mobile apps (technology) for purchasing grocery items in the context of Indian grocery retailing Industry as all the relationships pertaining to TAM were found significant. It was found that 90.8 per cent (314) respondents agreed to do online shopping using mobile apps, but only 36.7 per cent (127) respondents said to use mobile apps for grocery shopping.

These results show that the mobile shopping from grocery is not very well adopted by the people although the technology is not a problem. The e-retailers need to analyse the hidden reasons that stops consumers from buying grocery using apps. The results are consistent with the market trends that such grocery shopping apps are not in profits, even PepperTap (one newly launched mobile grocery shopping apps) was shut down.

Limitations and Future Implications

The study has few limitations which are stated as under. The sample was selected based on purposive and snowball sampling, and therefore precautions should be taken before generalizing the results. The study was conducted in Indian context and cross-cultural issues were not addressed. The study can be further tested in other countries as well. According to Hofstede’s cultural dimension theory, the culture can affect individual behaviour. The aim of study was to assess the usage of mobile apps in grocery shopping, so few variables of TAM were not included. This study can be further extended by testing actual behaviour and subjective norm.