Measuring Customer Response to Word-of-mouth Messages on Social Media: Development of a Multi-item Scale

Abstract

Word of mouth (WoM) is increasingly becoming a key element of marketing communication and is being leveraged on social media for increasing customer engagement. Literature on WoM on social media (E-WoM) is still relatively scarce. From the existing body of work on WoM, a number of causes and drivers of WoM emerge but there is not much work done in studying the customer’s responses to E-WoM (or E-WoM outcomes) and a comprehensive scale covering all dimensions has not been developed. This article attempts to identify potential outcomes of E-WoM on social media and develop a multi-item scale for measuring the same.

A list of 37 items of customer’s responses to E-WoM was generated based on the literature review and exploratory study. These items were tested through a survey administered to youth across four metro cities of India. EFA and CFA were conducted to identify the relevant dimensions. The research resulted in the development of a comprehensive 30-item scale measuring 7 dimensions of E-WoM outcomes on social media. The seven E-WoM outcomes measured by the scale are awareness and interest creation, information search online, online buzz generation, offline buzz generation, lead generation, liking and trial, and purchase intention.

Keywords

Word of Mouth customer’s response to E-WoM or E-word of mouth outcomes E-WoM scale

Introduction

Word-of-mouth (WoM) communication is defined as ‘informal communication, directed at other consumers, about ownership, usage, or characteristics about particular goods and services and/or their producers/sellers’ (De Matos, & Rossi, 2008), or ‘informal communication between private parties concerning evaluations of goods and services’ (Anderson, 1998), and as ‘a firms’ intentional influencing of consumer-to-consumer communications by professional marketing techniques” (Khan & Shariar, 2011; Kozinets, Valck, Wojnicki & Wilner, 2010).

WoM has a greater impact than other means of communication and is seen as a powerful force in the market place (Buttle, 1998; Samson, 2010; Walter, 2008). It is explored by corporates as the new means of marketing communication (Magnini, 2011; Swanson & Kelley 2001; Sweeney, Soutar, & Mazzarol, 2012). Further, WoM has longer carry-over effects as compared to traditional marketing (Trusov, Bucklin, & Pauwels, 2009).

Internet has increased accessibility, reach and transparency (Kozinets et al., 2010) and social media websites have emerged as influential media platforms (Zhang & Daugherty, 2009), supported by the viral quality of the internet, especially Web 2.0 (Steinman & Hawkins, 2010). Companies are now recognizing the power of social media (Haywood, 1989; Sashi, 2012), which is a phenomenal change in communication (Dotson 2009; Kilian, Hennigs, & Langner, 2012; Heinze & Fletcher, 2011). It offers companies tremendous opportunity to listen to their consumers, to engage them and even influence their conversations (Bruhn, Schoenmueller, & Schafer, 2012). WoM on social media or E-WoM is also easier to stimulate and influence because of high visibility and previously posted comments (Dellarocas & Narayan, 2006). WoM marketing strategies will benefit from research on what categories of customers are recommending the business, what are they recommending and what is prompting them to do so (Stokes, Ali, & Lomax, 2002). It offers companies tremendous opportunity to listen to their consumers, to engage them, and even influence their conversations (Bruhn, Schoenmueller and Schafer, 2012).

This article is an attempt to develop a scale for measuring the probable outcomes of a WoM episode received by a consumer/potential consumer on social media.

Research Gaps and Objectives

Though there is ample literature on WoM, tracing its origin, evolution and significance in the field of marketing communication, literature on E-WoM is still developing. Social media offers an opportunity to trace the impact of WOM activity (Trusov et al., 2009). Extending research on WoM to an online context is recommended (Sweeney et al., 2012).

The existing literature on WoM brings out a number of causes and drivers of WoM but there is not much work done in studying the customer’s response to WoM or the outcomes of WoM since capturing WoM data was difficult in the offline context. More so, expected outcome of WoM also needs to be specifically defined before it can be measured. Past researchers have looked at impact of WoM on different dimensions but usually taking one or two dimensions at the most. Most of the existing research is based on consumers in American and European developed economies. Not much research is done in the Indian consumer context.

WoM has been studied earlier as the traditional communication model, (Cheung & Thadani, 2010) and input-process-output model, (Chan & Ngai, 2011), but these models take into account only a few variables that are linked into a cause and effect relationship. A more extensive review of literature to understand the impact of WoM on social media was undertaken by Srivastava and Sharma (2017) in which an attempt has been made to divide literature into drivers, measures and effects of WoM and identify variables in each of these categories.

Outcomes of WoM or various customer responses have been studied with respect to variables such as behavioural intention (Wee et al., 1995), understanding of WoM communication (Delgadillo & Escalas, 2004), acceptance level of reader, intention to respond, ‘ripple effect of diffusion’ (Huang, Cai, Tsang, & Zhou, 2011), higher volume of WoM or client referrals (File & Prince, 1992) and online purchase increase (Chung, 2010). Empirical findings suggest that online WoM has a strong impact on refraining from buying a product, followed by generating further WoM and buying a recommended product. Further, negative WoM has a greater impact on consumer buying behaviour than positive WoM (Hennig-Thurau & Walsh, 2003). Reading WoM helps consumers save time and risk thus decreasing search time (Hennig-Thurau & Walsh, 2003). The usage range within a product category has an effect on pre-trial intentions, in turn, leading to recommending the product that has been tried (Samson, 2010). Not only does E-WoM lead to further online conversations (Samson, 2010), it also leads to offline advocacy (Graham & Havlena, 2007).

Trusov et al. (2009) proposes that E-WoM among consumers can influence brand image and purchase intention but his study is limited to automobile manufacturers and does not cover other sectors. E-WoM has a strong impact on purchase intentions, especially when consumers use mobile internet to read online recommendations directly at the point of purchase (Jalilvand & Samiei, 2012). According to Graham and Havlena (2007), online WoM creates brand advocacy offline, independent of advertisers’ influence. Valence in WoM content effects purchase intention in both UK and China (Christodoulides, Michaelidou, & Argyriou, 2012). Bruhn et al. (2012) illustrated that corporate weblogs or brand profiles on social networking sites impact brand image.

According to Lim and Chung (2014), quality of WoM messages and source of WOM impacted purchase intentions for online sports retail shopping. The provider of WoM and the receiver’s expertise were found to moderate the effect of message quality on purchase intention. A study by Zhang and Daugherty (2009) linked third-person effect to behavioural consequences related to WoM communication on social media. A WoM campaign may generate greater consumer involvement within a product category, resulting in more information seeking (Giese, Spangenberg, & Crowley, 1996).

Few scholars have made an attempt to study the impact of WoM on sales. Spekman and Dotson (2009) suggest that not only does social media play a role in the sales cycle but also effectively shortens the cycle by targeting customers who are motivated to purchase. File et al. (1992) also proved through their research that positive WoM is a strong factor in the purchase of financial services.

In most of the literature, usually a single driver/influencer of WoM is measured against a single effect of WoM, and most of these studies are not done in the context of online social media.

Walter (2008) first made an attempt to bring together the desired outcomes of WoM and drawing from WOMMA Terminology Framework, classified these into two broad categories—outcomes related to demand generation and outcomes related to further WoM generation. While the aspect of further WoM generation covered outcomes of ‘inquiry likelihood’ and ‘pass along likelihood’, the demand generation aspect covered outcomes of ‘purchase likelihood’ and ‘use likelihood’. Although literature on WoM is abundant, research of E-WoM, particularly Social Media is scarce (Brown, Amanda, & Nick, 2007).

While Walter’s study does provide a basis for defining possible outcomes of WoM, its scope and applicability were limited because it covered only volunteer-initiated WoM among existing social networks and did not cover other types of spontaneous or unseeded WoM on social media. Further, the variables defined by Walter were measured using single-item scales which may not be robust and limit the scope of statistical analysis and interpretation.

The research questions which arise therefore are as follows:

RQ1-Is the classification of outcome measures comprehensive or is there a need to include other variables, particularly in the context of social media

RQ2-Can a comprehensive multidimensional scale be developed for measuring the probable outcomes of an E-WoM episode on the social media?

Methodology

The scale development followed the procedure prescribed in literature (Anderson & Gerbing, 1988; Churchill, 1979; Devellis, 1991). As a first step, an initial set of items were generated based on literature review and exploratory consumer search using qualitative methods like in-depth interviews and focus group discussions. This was followed by item generation and checking of content validity through in-depth interviews, group discussions and expert opinion. Following this, an exploratory factor analysis (EFA) was conducted to define the underlying structure amongst variables and item reduction.

The third and final stage had three-fold objectives: developing model fit and reducing the items further, validation of factors through convergent and discriminant validity, and reliability check of final scale achieved. A confirmatory factor analysis (CFA) was conducted to achieve these.

Sample Selection

The study focuses on youth between the age group of 18–24 years (Nagle & Anand, 2012; Handa & Khare, 2013; Narang, 2011). In India, the Internet usage amongst youth is relatively high as compared to other age groups. As per IMRB and I-Cube 2013 survey, 57 per cent of the active Internet users are college going or young men (IAMAI and KANTAR IMRB, 2013). Moreover, 48 per cent of the mobile Internet users in India are also between 18 to 24 years of age. Samples were drawn from this age group across four metro cities of India, using convenience sampling. It was ensured that samples were taken from diverse areas within the given metros, ensuring a balance across gender and socio-economic backgrounds. A total of 815 respondents were part of the research.

Stage 1

Exploratory Research and Construct Development

Exploratory research was conducted through in-depth interviews across a sample size of 51 respondents. Out of these 25 were girls and 26 were boys. The key purpose of the exploratory research was to gain an understanding of how youth views and reacts to E-WoM on social media.

Primary data was collected through personally administered questionnaires. A standardized interview guide was used. The interview guide was a mix of introductory informal questions, open-ended questions and guided questions to churn out new variables for research and reiterated the significance of other existent variables that have been extracted from the literature. The data was collected at various places that are generally visited by youth: at college canteens, eateries/cafes popular amongst youth, homes, gyms, sports complex, etc. The in-depth interviews were designed to get insights into the pattern of social media usage, types of social media frequented, kinds of WoM messages received, responses to WoM, reasons for different responses, other likely responses, etc. Most respondents in this age group were found to be avid users of the social media and often clicked on or read the messages they received. They received messages related to products, social issues and politics.

Second step in the exploratory research was focus group discussions conducted in groups of 10 for a matching sample of 40 youth, where participants were encouraged to discuss the E-WoM messages they had received and their reactions to these.

The exploratory research helped generate possible actions/outcomes of E-WoM on social media.

Constructs development needs to take into account existing thought (Churchill, 1979; McKenzie, 2003. Walter’s broad classification of WoM Outcomes (2008) into ‘Demand Generation’ and ‘Further WoM Generation’ was taken as the basis for developing a comprehensive construct for defining and measuring E-WoM outcomes on social media. However, considering the previous literature on consumer responses to marketing communication (Barry & Howard, 1990; Lavidge & Steiner, 1961; Smith & Swinyard, 1982;), it was found that Walter’s classification covered only the liking, preference and trial/purchase likelihood aspects while leaving out the response aspects related to generation of awareness and knowledge. Even with respect to each of the response stages covered by Walter (2008), the outcomes defined related to only one dimension at most. In E-WOM communication, in general (and online social networks in particular), awareness and knowledge typically involve separate processes (Coulter & Roggeveen, 2012;De Bruyn & Lilien, 2008).

Through the insights generated through the exploratory research and the previous literature on WoM, it was found that possible outcomes/responses to an E-WoM episode on social media could include creation of awareness, knowledge gained through information received and further information search/inquiry which could be offline or online, sharing information both online and offline, liking, preference leading to trial and purchase intention.

The possible outcomes generated through exploratory research were compared with the outcome variables covered in existing studies to help validate the earlier variables and include any outcome which may have been left out. As a result of this process, a list of 46 items was developed. Respondent statements related to purchase intention were found to be very similar to the existing scale of Grewal, Monroe and Krishnan (1998), and these scale items with slight modification were also included in the final list of items.

To check the degree to which the items generated reflected the outcomes of an E-WoM episode, content validity was studied (Malhotra & Dash, 2010). Expert opinion was ensured through in-depth interviews with marketing experts including 12 academicians and brand managers. They were asked to evaluate each item for relevance, content validity, clarity and conciseness and identify any missing aspects in the constructs. From the list of 46 items, certain items were merged or eliminated to avoid duplication, ambiguity and implicit assumptions (Shimp & Sharma, 1987). Out of a total of 46 items, 37 were retained for conducting factor analysis and scale development. These included items related to further inquiry through different means and buzz generation as well as those relating to preference, trial, purchase intention, referral and usage likelihood.

Stage 2

Factor Analysis

In the second phase, primary data was collected through personally administered questionnaires. Hair, Anderson, Tatham and Black (2006), suggest at least five times as many observations as the number of variables to be analysed. The survey was administered to 336 respondents and a total of a total of 304 valid responses were received for factoring 37 variables.

Respondents were asked to describe a recent E-WoM message received by them on a social media platform and then given the list of 37 statements describing possible responses. They were asked to indicate on a scale of 1 to 7, how likely they were to undertake each of these actions in response to the E-WoM message described by them. Some of these 37 questions included the following:

How likely are you to

read the full message in detail

click on any link embedded in the message

repost the message

post any enquiry online

visit the company website

compare the prices of similar brand online

tag anyone in the message

mention the message to anyone offline

develop a liking for this brand after reading the message

try the brand at least once

purchase this brand in the future

EFA using principal component analysis and oblimin rotation (oblique method of rotation) was used to extract the factors that load onto the latent variables and to define the underlying structure among the variables in the analysis. Oblique rotation was used due to presence of correlated factors (Hair et al., 2006). To check the degree of interrelation among the variables, the Bartlett’s test of sphericity and Kaiser–Meyer–Olkin (KMO) measure of sample adequacy (MSA) were obtained. The communalities were assessed to check whether they are less than 0.5 as that variable would not have sufficient explanation. Finally, factors with significant cross loadings or ones with factor loading of less than 0.5 were eliminated. A second round of EFA was then conducted, after elimination to check the improvement in total variance explained.

In the first round of EFA, the scree plot brought out 7 components (Eigenvalue greater than 1) on to which the 37 items were loading. 7 components explained 68.056 per cent of all variances. The 37 items fulfilled the KMO MSA ( > 0.5; Hair et al., 2006) at 0.918, with a statistical significance of less than 0.001. Anti-image correlation matrix with the MSA was obtained and all variables were found to be within acceptable range ( > 0.5) with the lowest being 0.86 (Hair et al., 2006).

Next, variables with less than absolute value of 0.5 factor loadings were eliminated. These are—1c, 1d, 1f and 2j. One variable (1k) with significant cross loadings was eliminated; 32 items were retained (Table 1).

No further factors had less than significant loadings. 1k shows significant cross loadings (Table 4).

The items loading onto 7 constructs are given in Table 5.

The deleted questions were found to be either answered in some other question or not being relevant. For example, ‘likelihood of reposting the message’ was answered in part by questions like ‘likelihood of clicking on any link embedded in the message’ or ‘likelihood of posting an enquiry online’ or ‘likelihood of searching online for information about the brand’. These 32 items were run through EFA, using principal component analysis and oblimin rotation. These 32 items loaded onto 7 factors. Total variance explained was 70.777 (Table 3). KMO Bartlett’s test and p-value were within range, 0.915 and of less than 0.001 respectively (Table 2). Communalities were above 0.50. All MSA were within range, lowest being 0.862.

Table 1.

Pattern Matrix—EFA

	Component
	1	2	3	4	5	6	7
@1a	0.039	0.092	0.014	–0.799	0.091	–0.011	–0.133
@1b	0.097	0.029	–0.085	–0.747	0.105	0.086	0.036
~~@1c~~	~~0.495~~	~~0.075~~	~~0.096~~	~~–.380~~	~~–.323~~	~~–.012~~	~~–.023~~
~~@1d~~	~~0.323~~	~~0.044~~	~~0.139~~	~~–.369~~	~~–.328~~	~~–.051~~	~~.213~~
@1e	–0.215	0.018	0.221	–0.659	0.178	0.116	0.069
~~@1f~~	~~0.030~~	~~0.009~~	~~0.080~~	~~–.418~~	~~–.082~~	~~.039~~	~~.465~~
@1g	0.129	0.010	0.701	–0.092	–0.013	0.048	–0.013
@1h	0.141	0.026	0.706	0.056	–0.063	0.043	0.079
@1i	–0.054	–0.017	0.865	–0.011	0.004	0.088	0.006
@1j	–0.191	0.070	0.795	–0.050	0.062	–0.006	0.063
~~@1k~~	~~–0.146~~	~~0.075~~	~~0.421~~	~~–.032~~	~~.176~~	~~–.056~~	~~.550~~
@1l	0.254	0.060	0.025	–0.015	0.031	–0.035	0.728
@1m	0.263	0.011	0.016	0.055	–0.013	0.072	0.746
@1n	0.095	0.026	0.271	0.028	0.101	0.093	0.596
@2a	0.783	0.015	0.060	–0.083	0.079	0.019	–0.130
@2b	0.828	–0.024	0.007	–0.114	–0.014	0.021	–0.019
@2c	0.828	0.025	0.017	0.062	0.027	0.012	0.094
@2d	0.608	0.008	0.007	0.009	0.149	0.154	0.082
@2e	0.759	–0.001	0.020	0.023	0.094	0.029	0.139
@2f	0.809	0.027	0.013	0.026	–0.001	0.097	0.049
@2g	0.705	0.068	–0.017	0.030	0.168	–0.018	0.105
@2h	0.253	0.068	0.016	–0.163	0.730	0.000	–0.023
@2i	0.185	0.011	0.053	–0.149	0.757	0.031	0.001
~~@2j~~	~~0.442~~	~~0.052~~	~~0.025~~	~~–0.074~~	~~0.360~~	~~0.084~~	~~0.199~~
@2k	0.234	0.094	0.032	–0.110	0.580	0.047	0.150
@2l	0.682	0.056	–0.080	–0.053	0.137	0.055	0.145
@3a	0.140	0.676	0.231	0.009	0.055	–0.016	–0.255
@3b	0.108	0.701	0.201	–0.002	0.063	0.010	–0.180
@3c	0.224	0.650	0.182	0.084	–0.099	–0.021	–0.064
@3d	–0.083	0.741	0.006	–0.174	0.046	0.021	–0.116
@3e	–0.084	0.832	–0.144	–0.040	–0.039	0.091	0.103
@3f	–0.074	0.779	–0.087	0.019	–0.011	0.042	0.159
@3g	–0.024	0.820	–0.079	–0.004	–0.011	0.054	0.210
@3h	0.023	0.627	0.082	–0.004	0.132	0.108	–0.047
@4a	0.026	0.019	0.031	–0.007	–0.055	0.888	–0.059
@4b	–0.018	0.039	–0.020	–0.038	–0.015	0.847	–0.017
@4c	0.008	–0.014	0.024	0.047	–0.006	0.894	0.008

Source: Authors’ own.

Notes: Extraction Method: Principal Component Analysis.

Rotation Method: Oblimin with Kaiser Normalization.

a. Rotation converged in 11 iterations.

The highlighted numbers are the significant loadings.

Table 2.

KMO and Bartlett’s Test-EFA

Kaiser–Meyer–Olkin Measure of Sampling Adequacy.		0.915
Bartlett’s Test of Sphericity	Approx. Chi-Square	9224.392
	Df	528
	Sig.	0.000

Source: Authors’ own.

Next, a reliability test was conducted using Cronbach’s Alpha for each of the six scales (Table 6). The alpha coefficients were well above 0.70 and therefore acceptable (George & Mallery, 2003).

Stage 3

Confiramtory Factor Analysis

The scale so developed was next validated for its psychometric properties conducting CFA using Lisrel 8.80. A fresh matching sample of 450 respondents was taken to test the 32 items retained in the EFA, out of which 408 valid responses were received. More than 400 data points are acceptable for conducting CFA (Boomsma, 1982, 1985; Schumacker & Lomax, 2010).

Based on EFA, the proposed model had 32 observed variables loading on to 7 latent variables, namely comprehending the message, creating buzz, online enquiry, offline enquiry, liking and trial, purchase intention and approaching the company. Each observed variable was hypothesized to load onto a single latent variable.

After confirming model identification, robust maximum likelihood was used for model estimation and model testing, as this is the recommended method for data with normality issues (Joreskog & Sorbom, 2004.)

Table 3.

Total Variance Explained Table-EFA

Component	Initial Eigenvalues			Extraction Sums of Squared Loadings			Rotation Sums of Squared Loadings^a
Component	Total	% of Variance	Cumulative %	Total	% of Variance	Cumulative %	Total
1	11.531	34.941	34.941	11.531	34.941	34.941	8.009
2	3.848	11.661	46.602	3.848	11.661	46.602	6.649
3	2.405	7.289	53.891	2.405	7.289	53.891	5.257
4	1.646	4.988	58.879	1.646	4.988	58.879	3.746
5	1.558	4.720	63.599	1.558	4.720	63.599	4.941
6	1.265	3.834	67.432	1.265	3.834	67.432	6.199
7	1.104	3.344	70.777	1.104	3.344	70.777	3.295
8	0.921	2.789	73.566
9	0.744	2.255	75.822
10	0.642	1.946	77.768
11	0.580	1.757	79.525
12	0.555	1.682	81.206
13	0.496	1.504	82.710
14	0.472	1.431	84.142
15	0.457	1.385	85.526
16	0.426	1.292	86.818
17	0.407	1.234	88.052
18	0.374	1.132	89.184
19	0.369	1.117	90.302
20	0.358	1.085	91.387
21	0.327	0.990	92.376
22	0.317	0.961	93.338
23	0.299	0.905	94.243
24	0.258	0.782	95.024
25	0.235	0.712	95.736
26	0.228	0.691	96.427
27	0.214	0.648	97.075
28	0.198	0.600	97.675
29	0.193	0.586	98.260
30	0.160	0.486	98.746
31	0.144	0.437	99.183
32	0.138	0.417	99.600
33	0.132	0.400	100.000

Source: Authors’ own.

Notes: Extraction Method: Principal Component Analysis.

a. When components are correlated, sums of squared loadings cannot be added to obtain a total variance.

The highlighted Components 1 to 7 are explaining the total variance.

The Highlighted number 70.777 is explaining the cumulative percentage of total variance explained.

Our proposed model was over identified at 443 degrees of freedom and therefore acceptable for testing (Schumacker & Lomax, 2010).

The following path model was generated for testing (Figure 1).

Suggested modification indices were added to improve the model and fit indices.

Further, item 1j which had a low factor loading of 0.63, after the modification indices were added, was also removed.

Table 4.

Pattern Matrix—EFA 2

	Component
	1	2	3	4	5	6	7
@1a	0.123	0.086	0.048	–0.780	–0.046	–0.044	–0.117
@1b	0.208	0.034	–0.049	–0.775	0.046	–0.001	0.052
@1e	–0.133	0.010	0.247	–0.651	0.086	–0.134	0.058
@1g	0.140	–0.008	0.701	–0.103	0.046	–0.007	–0.022
@1h	0.139	0.015	0.715	0.057	0.049	0.031	0.061
@1i	–0.049	–0.032	0.855	–0.062	0.088	0.015	0.022
@1j	–0.192	0.060	0.768	–0.124	–0.008	–0.031	0.103
~~@1k~~	–~~0.123~~	~~0.072~~	~~0.441~~	~~-0.024~~	~~-0.064~~	~~-0.201~~	~~0.507~~
@1l	0.306	0.076	0.065	–0.029	–0.050	–0.044	0.699
@1m	0.299	0.030	0.054	0.051	0.062	–0.023	0.723
@1n	0.136	0.035	0.286	–0.028	0.079	–0.069	0.586
@2a	0.782	0.006	0.036	–0.154	0.001	–0.018	–0.095
@2b	0.848	–0.027	0.009	–0.143	0.002	0.039	–0.008
@2c	0.855	0.018	0.018	0.001	–0.003	0.024	0.088
@2d	0.595	0.006	–0.034	–0.075	0.137	–0.102	0.119
@2e	0.766	–0.013	0.016	–0.010	0.013	–0.085	0.131
@2f	0.810	0.015	0.025	0.041	0.087	–0.032	0.024
@2g	0.691	0.039	–0.007	0.061	–0.027	–0.215	0.052
@2h	0.070	0.008	–0.060	–0.050	–0.005	–0.903	–0.045
@2i	–0.010	–0.047	–0.032	–0.043	0.025	–0.929	–0.010
@2k	0.120	0.046	0.013	0.002	0.043	–0.702	0.083
@2l	0.679	0.038	–0.059	–0.019	0.044	–0.169	0.101
@3a	0.154	0.623	0.252	0.006	–0.005	–0.066	–0.294
@3b	0.115	0.646	0.237	0.044	0.024	–0.121	–0.246
@3c	0.265	0.606	0.248	0.135	–0.006	0.040	–0.144
@3d	–0.068	0.723	–0.002	–0.206	0.022	–0.044	–0.090
@3e	–0.051	0.830	–0.133	–0.088	0.097	0.056	0.122
@3f	–0.045	0.783	–0.096	–.077	0.045	0.063	0.202
@3g	0.004	0.803	–0.037	0.023	0.064	–0.041	0.170
@3h	–0.042	0.595	0.068	0.044	0.121	–0.248	–0.051
@4a	0.035	0.014	0.031	–0.024	0.885	0.056	–0.046
@4b	–0.033	0.029	–0.016	0 .008	0.848	–0.060	–0.025
@4c	0.005	–0.016	0.016	0.026	0.892	0.008	0.019

Source: Authors’ own.

Notes: Extraction Method: Principal Component Analysis.

Rotation Method: Oblimin with Kaiser Normalization.^a

a. Rotation converged in 11 iterations.

The highlighted numbers are the significant loadings.

This gave us the following Path Model with modification indices (Figure 2).

Goodness of fit (GFI) indices for the model were found to be acceptable (Table 7)—Root mean square root error of approximation (RMSEA) was less than 0.7; GFI was close to 0.90 and comparative fit index (CFI) and normed fit index (NFI) were also above 0.90, indicating good fit (Hair et al., 2006).

A relative Chi square (CMIN ratio) of 1.87 also indicated adequate fit (Byrne, 1989; Carmines & McIver, 1981; Marsh & Hocevar, 1985; Wheaton, Muthen, Alwin, & Summers, 1977).

The next step was assessing Construct Validity or the extent to which the measured items represent the latent constructs (Hair et al., 2006). Construct Validity was assessed via Convergent Validity and Discriminant Validity.

Table 5.

32 Items Loading Onto 7 Components

Scale			Items
(Component 1)	1	1	Share the message online with others
	2	2	Forward the message on the same social media site
	3	3	Repost the Message
	4	4	Post Enquiry online
	5	5	Mention the message in any other post
	6	6	Post/repost the message
	7	7	Discuss the message on social media
	8	8	Contact Company employee/Representative online
(Component 2)	9	1	Develop a greater liking for this brand compared to other such brands
	10	2	Prefer this brand over other such brands
	11	3	Prefer this brand over your current brands
	12	4	Try the product at least once
	13	5	Try the product more than once
	14	6	Use this product along with your current brand
	15	7	Use this brand more than you would use other such brands
	16	8	Recommend this brand to one or more persons
(Component 3)	17	1	Visit the company website
	18	2	Visit online retail sites that sell the product
	19	3	Read Reviews about the product, online
	20	4	Compare price of the product online
(Component 4)	21	1	Read the full message in detail
	22	2	Click on any link embedded in the message
	23	3	Search online about the information about the product
(Component 5)	24	1	Mention about the brand offline
	25	2	Discuss the message with anyone offline
	26	3	Recommend the brand offline
(Component 6)	27	1	Probability of buying the brand mentioned in the message
	28	2	The likelihood that I would purchase the brand
	29	3	Probability that I should consider buying it
(Component 7)	30	1	Contact company employee/representative online
	31	2	Contact company employee/representative offline
	32	3	Visit show room or shop of the company

Source: Authors’ own.

Table 6.

Reliability Test

Component Number	Cronbach’s Alpha
1	0.924
2	0.907
3	0.841
4	0.752
5	0.859
6	0.866
7	0.854

Source: Authors’ own.

Table 7.

Fit Indices

Chi square	753.02
Degrees of freedom	403
CMIN Ratio	1.87
GFI	0.86
CFI	0.99
NFI	0.97
RMSEA	0.046

Source: Authors’ own.

Convergent Validity (Table 8) was demonstrated as first, factor loadings or standardized lambdas were more than 0.7 or close to 0.7. Second, Average Variance Extracted (AVE) or the average squared factor loadings was more than 0.5. Third, the Construct Reliability (CR), which is the ratio of squared sum of factor loadings for each construct and the squared sum of factor loadings for each construct and the error variance term for each construct was greater than 0.7 for all seven latent variables.

Figure 1.

Source: Authors’ own.

Figure 2.

Source: Authors’ own.

Discriminant Validity or the extent to which a particular construct varies from other constructs in the model is confirmed as the AVE is greater than the squared correlation among all seven corresponding constructs (Fornell & Cha, 1981; Fornell & Larcker, 1981; Hair et al., 2006). These inter-construct correlations are given in Table 9.

The research identified seven specific outcomes of E-WoM on social media, each of which can be measured through the multi-item sub-scales. The outcomes and the sub-scales are listed ahead (see Table 10):

Awareness and Interest Creation (Component 5): If a WoM message is likely to be read by the receiver, it will create awareness and if found to be relevant could also lead to interest generation. The sub-scale for this outcome comprises three items from likelihood of reading the message in detail and seeking further information to clicking on embedded links. This would be the first step in any further engagement by the receiver.

Information Search Online (Component 3): Beyond initial interest, another outcome of E-WoM could be further search for information which is easily done online. This outcome is measured by a 5-item sub-scale covering aspects like the likelihood of visiting the company’s website, visiting online retail sites, comparing prices and reading online reviews.

Online Buzz Generation (Component 1): A key advantage of E-WoM over other forms of communications is the ease with which the message can be shared and discussed with many receivers and this quality is enhanced in case of social media networks. Therefore, one of the important outcomes of E-WoM would necessarily be creation of online buzz. The 8-item sub-scale comprising items such as likelihood of forwarding of messages, online discussions, sharing on different sites and so on can be used as a measure of this outcome.

Offline Buzz Generation (Component 4): Consumer can share or use information received online on offline media/platforms, thus leading to creation of offline buzz. This outcome can be measured by 3-item sub-scale incorporating the likelihood of the receiver mentioning the brand in an offline context, discussing it offline or even recommending it offline.

Lead Generation (Component 7): E-WoM facilitates two-way interaction between sender and receiver and also between original source and receiver. Therefore, if the receiver is likely to initiate contact with the company as a result of E-WoM message, it is a very valuable source of leads for a firm. A 3-item sub-scale measuring the likelihood of consumer contacting the company online or offline and visiting its showroom/office for further query or discussion can be an effective measure of this outcome.

Liking and Trial (Component 2): Marketing communication aims to generally move the consumer from awareness to liking and trial which can form the prelude to purchase. An 8-item sub–scale can measure this outcome of E-WoM by understanding likelihood of the E-WoM leading to a preference of the concerned brand over other brands, along with the likelihood of trial and usage of the product.

Purchase Intention (Component 6): One of the most important outcomes desired by marketers in most cases is the ability of marketing communication to move consumers along the response hierarchy all the way to purchase or at least purchase intention. This outcome of demand generation likelihood can be measured by a 3-item sub-scale evaluating the probability of considering the brand for purchase and likelihood of buying the brand. The initial statements for this scale were based on an existing purchase intention scale (Grewal et al., 1998) and have been validated by this research.

Table 8.

Convergent Validity

	Measured variable	Standardized Lambda	Error	CR	AVE
Component 1	1	0.74	0.45	0.83	0.64
	2	0.8	0.35
	3	0.86	0.26
	4	0.71	0.49
	5	0.85	0.28
	6	0.87	0.25
	7	0.77	0.4
	8	0.79	0.37
Component 2	1	0.73	0.47	0.71	0.54
	2	0.77	0.4
	3	0.69	0.53
	4	0.71	0.49
	5	0.74	0.45
	6	0.69	0.53
	7	0.8	0.35
	8	0.72	0.48
Component 3	1	0.81	0.35	0.78	0.59
	2	0.73	0.47
	3	0.76	0.41
Component 4	1	0.76	0.43	0.74	0.56
	2	0.78	0.39
	3	0.7	0.51
Component 5	1	0.9	0.19	0.89	0.73
	2	0.91	0.17
	3	0.75	0.44
Component 6	1	0.87	0.25	0.85	0.67
	2	0.78	0.4
	3	0.81	0.35
Component 7	1	0.91	0.16	0.81	0.7
	2	0.91	0.18
	3	0.67	0.65

Source: Authors’ own.

Table 9.

Discriminant Validity

	Component 1	Component 2	Component 3	Component 4	Component 5	Component 6	Component 7	AVE
Component 1	1	0.16	0.1444	0.3364	0.2025	0.1156	0.3844	0.64
Component 2	0.4	1	0.2601	0.1764	0.2401	0.3136	0.0841	0.54
	7.41
Component 3	0.38	0.51	1	0.1296	0.2601	0.1225	0.1936	0.59
	6.35	9.17
Component 4	0.58	0.42	0.36	1	0.2704	0.1225	0.1849	0.73
	14.96	7.54	6.03
Component 5	0.45	0.49	0.51	0.52	1	0.1369	0.1156	0.56
	7.68	8.42	8.41	9.72
Component 6	0.34	0.56	0.35	0.35	0.37	1	0.0676	0.67
	6.93	12.23	5.86	6.62	6.61
Component 7	0.62	0.29	0.44	0.43	0.34	0.26	1	0.7
	14.7	4.72	7.7	8.5	5.45	4.55

Source: Authors’ own.

Note: Values below the diagonal are correlation estimates. Values in italics are the t-values. Values above the diagonal are squared correlations, which are compared with the AVE values, given in bold.

Table 10.

Labelling the Constructs

Scale			Items
(Component 1) Online Buzz	1	1	@2a Share the message online with others
	2	2	@2b Forward the message on the same social media site
	3	3	@2c Forward the message on another social media site
	4	4	@2d Post Enquiry online
	5	5	@2e Mention the message in any other post
	6	6	@2f Post/repost the message
	7	7	@2g Discuss the message on social media
	8	8	@2l Contact Company employee/Representative online
(Component 2) Liking and Trial	9	1	@3a Develop a greater liking for this brand compared to other such brands
	10	2	@3b Prefer this brand over other such brands
	11	3	@3c Prefer this brand over your current brands
	12	4	@3d Try the product at least once
	13	5	@3e Try the product more than once
	14	6	@3f Use this product along with your current brand
	15	7	@3g Use this brand more than you would use other such brands
	16	8	@3h Recommend this brand to one or more persons
(Component 3) Information Search Online	17	1	@1g Visit the company website
	18	2	@1h Visit online retail sites that sell the product
	19	3	@1i Read Reviews about the product, online
	20	4	@1j Compare price of the product online
(Component 4) Offline Buzz	24	1	@2h Mention about the brand offline
	25	2	@2i Discuss the message with anyone offline
	26	3	@2k Recommend the brand offline
(Component 5) Awareness and Interest	21	1	@1a Read the full message in detail
	22	2	@1b Click on any link embedded in the message
	23	3	@1e Search online about the information about the product
(Component 6) Purchase Intention	27	1	@4a Probability of buying the brand mentioned in the message
	28	2	@4c The likelihood that I would purchase the brand
	29	3	@4b Probability that I should consider buying it
(Component 7) Lead Generation	30	1	@1l Contact company employee/representative online
	31	2	@1m Contact company employee/representative offline
	32	3	@1n Visit show room or shop of the company

Source: Authors’ own.

Contribution and Limitations

This research has filled a major gap in academic literature on measuring the impact of E-WoM on social media, by defining the possible outcomes and developing a comprehensive E-WoM outcome scale to measure the impact of E-WoM on desired outcomes. The seven outcomes of E-WoM identified by this research are in line with existing marketing theory on stages of the consumer response hierarchy which defines the expected path of moving the consumers from awareness to purchase. At the very least the E-WoM message should be read by the recipient creating awareness, and if the message is interesting and relevant it should also result in creating an interest leading to search for further information, discussions within and outside the social network, initiation of contact with the company/brand representatives, development of liking and brand preference, leading finally to purchase intention. For both academicians and practitioners, this is a powerful tool which can help measure effectiveness of E-WoM campaigns both in terms of likely strength of impact and the dimensions which would be impacted. Desired outcomes could vary depending upon the communication objectives and therefore the sub-scales can provide clarity for specific objectives.

The outcomes defined in this study cover the outcomes related to generation of WoM as well as demand-related outcomes specified by Walter. However, this study adds value by further refining the outcome definition by making these more specific and also by adding new dimensions to make it more comprehensive. An additional benefit of the scale developed by this study is that it can be easily mapped onto the response hierarchy models of communication and stages of the purchase decision process, providing for easier linkage between marketing communication objectives and outcomes.

This research study has provided initial evidence for the reliability and validity of the E-WoM outcomes scale in the context of social media for Indian youth. Limitations of this study are that it is based on convenient sampling and that is limited to four largest metros in India. Further validation could be done for its applicability for respondents in smaller towns and also across different countries. Further research could look at validating/modifying the scale for all age groups.

Future researchers may also try to establish the linkage between the drivers and outcomes of E-WoM on social media and analyse the factors impacting these.

References

Anderson

E. W.

(1998). Customer satisfaction and word of mouth. Journal of Service Research, 1(1), 5–17.

Anderson

J. C.

& Gerbing

D. W.

(1988). Structural equation modelling in practice: A review and recommended two step approach. Psychology Bulletin, 103(3), 411–423.

Barry

T. E.

& Howard

D. J.

(1990). A review and critique of the hierarchy of effects in advertising. International Journal of Advertising, 9(2), 121–135.

Boomsma

(1982). Robustness of LISREL against small sample sizes in factor analysis models. In Joreskog

K. G.

& Wold

(Eds), Systems under indirection observation: Causality, structure, and prediction part I (pp. 149–173). Amsterdam: North Holland.

Boomsma

(1985). Nonconvergence, improper solutions, and starting values in LISREL maximum likelihood estimation. Psychometrika, 50, 229–242.

Brown

Amanda

J. B.

& Nick

(2007). Word of mouth communication within online communities: Conceptualizing the online social network. Journal of Interactive Marketing, 21(3), 2–20.

Bruhn

Schoenmueller

& Schafer

D. B.

(2012). Are social media replacing traditional media in terms of brand equity creation? Management Research Review, 35(9), 770–790.

Buttle

F. A.

(1998). Word of mouth: Understanding and managing referral marketing. Journal of Strategic Marketing, 6, 241–254.

Byrne

B. M.

(1989). A Primer of LISREL: Basic applications and programming for confirmatory factor analytic models. New York, NY: Springer-Verlag.

10.

Carmines

E. G.

& McIver

J. P.

(1981). Analyzing models with unobserved variables. In Bohrnstedt

G. W.

& Borgatta

E. F.

(Eds), Social measurement: Current issues (pp. 65–115). Beverly Hills, CA: SAGE Publications.

11.

Chan

Y. Y. Y.

& Ngai

E. W. T.

(2011). Conceptualizing electronic word of mouth activity: An input-process-output perspective. Marketing Intelligence & Planning, 29(5), 488–516.

12.

Cheung

C. M. K.

& Thadani

D. R.

(2010). The effectiveness of electronic word-of-mouth communication: A literature analysis. 23rd Bled eConference eTrust: Implications for the Individual, Enterprises and Society, Bled, Slovenia, June 20–23.

13.

Christodoulides

Michaelidou

& Argyriou

(2012). Cross-national differences in e-WOM influence. European Marketing Journal, 46(11/12), 1689–1707.

14.

Chung

C. A. K.

(2010). Social media gratification and attitude toward social media marketing messages: A study of the effect of social media marketing messages on online shopping value (p. 581). Proceedings of the Northeast Business & Economics Association.

15.

Churchill

G. A.

Jr. (1979). A paradigm of developing better measures of marketing constructs. Journal of Marketing Research, 16(1), 64–73.

16.

Coulter

K. S.

& Roggeveen

(2012). Like it or not: Consumer responses to word-of-mouth communication in on-line social networks. Management Research Review, 35(9), 878–899.

17.

De Bruyn

& Lilien

G. L.

(2008). A multi-stage model of word-of-mouth influence through viral marketing. International Journal of Research in Marketing, 25, 151–163.

18.

De Matos

C. A.

& Rossi

C. A. V.

(2008). Word-of-mouth communications in marketing: A meta-analytic review of the antecedents and moderators. Journal of the Academy of Marking Science, 36. doi: 10.1007/s11747-008-0121-1

19.

Delgadillo

& Escalas

J. E.

(2004). Narrative word-of-mouth communication: Exploring memory and attitude effects of consumer storytelling. Advances in Consumer Research, 31(1), 186–192.

20.

Dellarocas

& Narayan

(2006). A statistical measure of a population’s propensity to engage in post-purchase online word-of-mouth. Statistical Science, 21(2), 277–285.

21.

File

K. M.

Judd

B. B.

& Prince

R. A.

(1992). Interactive marketing: The influence of participation on positive word-of-mouth and referrals. Journal of Services Marketing, 6(4), 5–14.

22.

File

K. M.

& Prince

R. A.

(1992). Positive word-of-mouth: Customer satisfaction and buyer behaviour. International Journal of Bank Marketing, 10(1), 25–29.

23.

Fornell

& Cha

(1981). Partial least squares. In Bagozzi

R. P.

(Ed.), Advanced methods of marketing research (pp. 52–78). Cambridge, MA: Blackwell.

24.

Fornell

& Larcker

D.F.

(1981). Evaluating structural equation models with unobservable variables and measurement error. Journal of marketing research, 18(1), 39–50.

25.

George

& Mallery

(2003). SPSS for Windows step by step: A simple guide and Reference 11.0 update (4th ed.). Boston: Allyn & Bacon.

26.

Giese

J. L.

Spangenberg

E. R.

& Crowley

(1996). Effects of product-specific word-of-mouth communication on product category involvement. Marketing Letters, 7(2), 187–199.

27.

Graham

& Havlena

(2007). Finding the missing link: Advertising’s impact on word of mouth, web searches and site visits. Journal of Advertising Research, 47(4), 427–435.

28.

Grewal

Monroe

K. B.

& Krishnan

(1998). The effects of price-comparison advertising on buyers’ perceptions of acquisition value, transaction, value, and behavioural intentions. Journal of Marketing, 62(April), 46–59.

29.

Hair

J. F. J.

Anderson

R. E.

Tatham

R. L.

& Black

W. C.

(2006). Multivariate data analysis (6th ed.). Upper Saddle River, NJ: Prentice Hall.

30.

Handa

& Khare

(2013). Gender as a moderator of the relationship between materialism and fashion clothing involvement among Indian youth. International Journal of Consumer Studies, 37(1), 112–120.

31.

Haywood

K. M.

(1989). Managing word of mouth communications. Journal of Services Marketing, 3(2), 55–67.

32.

Heinze

& Fletcher

(2011). Can we make higher education relevant to the needs of the Search & Social Media Marketing industry? In 6th Education in a Changing Environment Conference, Creativity and Engagement in Higher Education, 6–8 July, Salford, Greater Manchester.

33.

Hennig-Thurau

& Walsh

(2003). Electronic word-of-mouth: Motives for and consequences of reading customer articulations on the Internet. International Journal of Electronic Commerce, 8(2), 51–74.

34.

Huang

Cai

Tsang

A. S. L.

& Zhou

(2011). Making your online voice loud: The critical role of WOM information. European Journal of Marketing, 45(7), 1277–1297.

35.

IAMAI and KANTAR IMRB. (2013). I-cube survey: Internet in India. Retrieved 10 July 2017 , from http://imrbint.com/site/reports/i-cube

36.

Jalilvand

M. R.

& Samiei

(2012). The effect of electronic word of mouth on brand image and purchase intention. Marketing Intelligence & Planning, 30(4), 460–476.

37.

Joreskog

K. G.

& Sorbom

(2004). LISREL 8.7 for Windows (Computer Software), Lincolnwood, IL: Scientific Software International.

38.

Khan

Md. R. T.

& Shahriar

F. M.

(2011). Online social media as a driver of buzz marketing: Who’s riding? Annual Conference on Innovations in Business and Management, London, UK.

39.

Kilian

Hennigs

& Langner

(2012). Do millennials read books or blogs? Introducing a media usage typology of the Internet generation. Journal of Consumer Marketing, 29(2), 114–124.

40.

Kozinets

R. V.

Valck

K. De

Wojnicki.

A. C.

& Wilner

S. J. S.

(2010). Networked narratives: Understanding word-of-mouth marketing in online communities. Journal of Marketing, 74(2), 71–89.

41.

Lavidge

R. J.

& Gary

A. S.

(1961). A model for predictive measurements of advertising effectiveness. Journal of Marketing, 25(October), 59–62.

42.

Lim

B. C.

& Chung

C. M. Y.

(2014). Word-of-mouth: The use of source expertise in the evaluation of familiar and unfamiliar brands. Asia Pacific Journal of Marketing and Logistics, 26(1), 39–53.

43.

Magnini

V. P.

(2011). The implications of company-sponsored messages disguised as word-of-mouth. Journal of Services Marketing, 25(4), 243–251.

44.

Malhotra

N. K.

& Dash

(2010). Marketing research: An applied approach. New Delhi: Dorling Kindersely.

45.

Marsh

H. W.

& Hocevar

(1985). Application of confirmatory factor analysis to the study of self-concept: First-and higher order factor models and invariance across groups. Psychological Bulletin, 97, 568–582.

46.

Nagle

Y. K.

& Anand

(2012). Empathy and personality traits as predictors of adjustment in Indian youth. Industrial Psychiatry Journal, 21(2), 125.

47.

Narang

(2011). Examining the role of various psychographic characteristics in apparel store selection: A study on Indian youth. Young Consumers, 12(2), 133–144.

48.

Samson

(2010). Product usage and firm-generated word of mouth: Some results from FMCG product trials. International Journal of Market Research, 52(4), 459–482.

49.

Sashi

C. M.

(2012). Customer engagement, buyer-seller relationships, and social media. Management Decision, 50(2), 253–272.

50.

Schumacker

R. E.

& Lomax

R. G.

(2010). A beginner’s guide to structural equation modeling (3rd ed.). New York, NY: Routledge.

51.

Shimp

T. A.

& Sharma

(1987). Consumer ethnocentrisms: Construction and validation of the CETSCALE. Journal of Marketing Research, 24(3), 280–289.

52.

Smith

R. E.

& Swinyard

W. R.

(1982). Information response models: An integrated approach. Journal of Marketing, 46(1), 81–93.

53.

Spekman

R. E.

& Dotson

(2009). Using social media in B2B context (Technical Note M-0778, 1–15). Charlottesville, VA: Darden Business.

54.

Srivastava

& Sharma

R. W.

(2017). Developing a model for studying the antecedents and effects of word of mouth (wom) and e-wom marketing. Jindal Global Business Review Journal, 6(1), 1–19.

55.

Steinman

M. L.

& Hawkins

(2010). When marketing through social media, legal risks can go viral. Intellectual Property & Technology Law Journal, 22(8). 1–9.

56.

Stokes

Ali

S. S.

& Lomax

(2002). Shaping up word of mouth marketing strategy: The case of an independent health club. Journal of Research in Marketing and Entrepreneurship, 4(2), 119–133.

57.

Swanson

S. R.

& Kelley

S. W.

(2001). Service recovery attributions and word-of-mouth intentions. European Journal of Marketing, 35(1/2), 194–211.

58.

Sweeney

J. C.

Soutar

G. N.

& Mazzarol

(2012). Word of mouth: Measuring the power of individual messages. European Journal of Marketing, 46(1/2), 237–257.

59.

Trusov

Bucklin

R. E.

& Pauwels

(2009, September). Effects of word-of-mouth versus traditional marketing: Findings from an Internet social networking site. Journal of Marketing, 73(5), 90–102.

60.

Walter

C. J.

(2008). The role of disclosure in organized word-of-mouth marketing programs. Journal of Marketing Communications, 14(3), 225–241.

61.

Wheaton

Muthen

Alwin

D. F.

& Summers

G. F.

(1997). Assessing reliability and stability in panel models. Sociology Methodology, 8, 84–136.

62.

Zhang

& Daugherty

(2009). Third-person effect and social networking: Implications for online marketing and word-of-mouth communication. American Journal of Business, 24(2), 53–64.