Essay: Network usage and potential scale

Network Usage and Potential Scale

Abstract

In the modern information economy, who is to be targeted as a propagator of information will generally decide the way businesses go ahead and allocate their resources to connect with consumers. Network Usage and Potential Scale (NUPS) is an attempt to quantify such relative potential of an individual to act as a ‘hub’ by incorporating Usage and Motive along with Networking Potential. A typical scale development methodology has been employed for the same which includes two surveys and statistical tests to progressively assess the relevance of questions constituting these surveys. The final NUPS score is a sum total of the scores assigned to the questions in the scale.

Introduction

NUPS is an acronym for Network Usage and Potential Scale. NUPS is essentially a survey tool or instrument which considers every individual who has any form of presence on the web as a ‘resource’ and tries to gauge his/her usefulness in any attempt to propagate information. The need to connect and network among ourselves has been ever-existent. Just as the proliferation of tools to do so effectively has helped us in putting it to various uses, identifying trends in this behaviour is also important. The ‘tools’ that we talk about are characterised by their ease of use, availability and ability to deter migration to other similar ones. Web has invaded our communication set up like never before and has been revolutionary in every sense. It then is logical to identify the reasons for this and ask why we have accepted it with such alacrity. Is it just because everyone else is there or is it because of the increasingly personalised domains that are made available almost every day to us?

What then is the exact ‘motive’ with which we use the virtual world? The purpose of our project is to identify a model which can help us map ‘usage patterns’ to individual/group ‘traits’ and then finally lead us to calibrate and map it onto a comparative scale.

Literature Review

Information is power in today’s world and hence, Networking has been heavily researched upon as a phenomenon as well as a tool. Group behaviour has always been deemed to be different from Individual behaviour and has been suitably exploited especially in advertising. Also, Business and Social are the two aspects of the way networks are currently employed and accessed. A Social Network is a social structure made of individuals, which are tied toby ether one or many specific types of interdependence, such as friendship, kinship, financial exchange, dislike or knowledge. Proliferation of web-based tools to engage oneself in such ‘communities’ has helped individuals find similarities and connect unabashedly. ‘Social Networking Sites’ (SNS) have been special in this regard. While Blogs, Forums, Virtual Worlds and such other platforms do exist, the growth and acceptance of SNS’s has been mind boggling. The related field of study is termed as Social Analytics. Need to Belong, Internet Self-Efficacy, Collective Self Esteem, and Need for Cognition have been cited as four primary reasons for huge participation of teenagers in SNS’s (Gangadharbhat, 2008). Research also indicates that student’s use of SNS’s is not only for personal socialization and leisure but also as a platform for more serious activities (Shalija Agarwal, 2009). A study of how young adults communicate among themselves in three contexts, social networking sites (SNSs), instant messaging (IM), and face-to-face showed that people often used the Web, especially SNSs, to connect and reconnect with friends/ family members. Hence, there was overlap between the online and offline networks of the participants. However, the overlap was not perfect; the pattern suggested that adults use different online contexts to strengthen various aspects of their offline connections. An article also talks about how web-based social networking sites are not just for teenagers anymore.

A study suggests that individuals may also use such tools for Career Management. The use of networking techniques to develop and maintain relationships with influential people who can help one get a career direction or work on the web is highly prevalent. Five different types of networking behaviours used in the study were: maintaining external contacts, socializing, engaging in professional activities, participating in community activities, and increasing internal visibility (Forret, 2004). A paper examining the role of electronic mail in organizational communication found that people use EMS more often to communicate with their bosses rather than their subordinates. The results also revealed ample information about the users’ attitude, personality, buying patterns and networking behavior (Sproull, 1986).

Informal groups (‘Cliques’) have more homogeneous opinions and share many common traits. But, each member of the clique would have almost similar knowledge. To find new information, members of the clique look and search for newer sources and this is known as ‘the strength of weak ties’. This helps us understand the motivation of such individuals to seek more of ‘networks’ in spite of apparently being ‘satisfied’ in their cliques (Granovetter, 1983).

There are certain quantitative measures also available to map an individual’s networking characteristics. Social Networking Coefficient is a numerical value in this regard. It represents both the size of an individual’s social network and his/her ability to influence the same. This is primarily employed in Viral Marketing strategies. The concept of Alpha User is also on similar lines. He/She is essentially a human networking hub. The primary behavior exhibited by such an individual is to keep the social network connected and updated. It can be compared to the concept of an ‘Early Adopter’ in traditional marketing. Past information distribution frequency within their network, group memberships, recognition, leadership roles, participation in Social Networking activities, editing/publication/contributing to non-electronic media and editing/publication/contributing to electronic media are a few of the variables used to calculate an individual’s SNP. Some algorithms and coefficients initially used to develop search algorithms focused on propagation of links in continuation and are found to be useful now in Social Analytics.

The measure of the interrelatedness of the neighbours of an object is called a Clustering coefficient. It indicates the ratio of existing links of the neighbours of an object to each other to the maximum possible such links. The hubs and authorities algorithm identifies objects that are more likely to be Alpha Users (i.e. authorities) and objects that are more likely to point to many such authorities (i.e. hubs). Betweenness, Closeness, Cohesion and Radiality are other important measures used in Social Analytics.

Above discussion suggests a historical bias towards ‘potential’ mapping in networks and derives mainly from the Network Theory which is a focal point of modern sociology. Our team felt the need for a tool which could include the ‘Usage’ aspect of studying networking behaviours. This is supposed to include Usage rate, patterns (if any) and reasons for the same. This instrument is targeted at providing a more comprehensive mapping of an individual as a specific ‘target’ in a group displaying similar characteristics. Specific implications and potential applicability has been discussed later in the document.

Attributes and Dimensions of NUPS

As discussed above, our final instrument intends to capture the effects of Usage and Reasons (specific motive) along with the already deeply researched Network Potential to arrive at the relative applicability of a particular individual for his intended use as an information propagator. These three leading ‘factors’ were represented by a group of questions in our initial survey and have been detailed below.

Usage

Amount and pattern of usage is an important antecedent. This is expected to have a direct proportionality with the final score on our scale. An individual having a greater rate of usage (which may in turn be dependent on age and personality) will obviously be a prime target for information propagation. The type of information (i.e. category) will also be decided by the direction of his exposure on the web in terms of using various facilities available to him. This has been captured throughout the development of our survey. As for the pattern of usage, for example, younger people generally tend to remain online (either being active or inactive) for longer periods in a day and level of activity keeps varying, while office workers generally log in during evenings for 1-2 hours and remain highly active during that period. Similarly, a gregarious individual would comment and reply to a large number of posts while a reserved individual may remain selective in this activity and that too involuntarily.

Reasons (Motive)

What could be the basic drive in an individual to expose himself to the virtual world? An answer to this is obviously important to decide the type and amount of information directed at him in a planned effort (like advertising a product online). We have then tried to separately capture this aspect in a category of questions in the final instrument and the eventual score. Even when we have limited the scope of our survey to the ‘Social’ aspect of an individual’s web presence (although implications may be ‘business’ related too), within this, reasons could be different. For example, one person may just want to stay in touch while another may consider this as a form of entertainment which then is his primary involuntary motive to network socially.

Network Potential

Through research in sociology and social analytics, this has been established as one of the basic feature of a ‘target’ with regards to our motive of identifying information propagators. Network Potential can be loosely defined as the ability of a person to connect with individuals. For example, Network Potential will not simply be determined by how many people one can befriend online, but also with how varied the friends are in terms of personality. This becomes an important antecedent because as much as it depends on one’s own personality, it also depends on the surroundings and environment around a person.

Objectives

The primary objective of our research project is to develop a scale to measure “Network Usage and Potential”. “Network” here means various social networking tools such as social networking websites (Facebook, Orkut, Twitter, etc.), weblogs, or blogs as they are popularly known, various kinds of forums (educational, professional, social, etc.) and the like.

From our literature review, we find that although tools may be available to measure Internet Usage, and these may thus be adapted for measuring Network Usage, there is no instrument which jointly measures the usage of a person (of social networking tools as a whole) and also maps his potential as a network user. The scale which we have developed aims to jointly measure a person’s usage of social networking tools along with his networking potential. The scale could be translated thus:

A person, who spends more time with social networking tools, will have higher usage.

The “network potential” indicates how well connected a person is with the help of his social networking tools, or how wide is his reach due to his use of social networking tools. For example, a person who has 500 contacts on Facebook will be far better connected than one who has 100, and thus has higher potential as far as his networking ability is concerned.

The two concepts combined together will indicate how useful a person is with respect to his networking abilities. The usefulness could be for marketers. It is a well-known fact that internet marketing is on the rise and is one of the most important avenues to be covered by any marketer, especially those targeting the youth. By administering this scale, marketers could determine who has a relatively higher Network Usage and Potential within his/her peer group. This user could then be targeted for disseminating information or as an advertising medium.

A person must not only use networking tools more often, but also have high potential, in order to be truly useful. For example, a user who spends a lot of time with social networking tools, but does not have the capacity or ability to reach out to people, will be of little use to any marketer. In the same way, in spite of having an intricate network, if the user spends little or no time leveraging this resource; it would be futile for a marketer to target him. Thus, it is essential to measure both usage and potential together, which is the objective of this scale.

Methodology and Findings

Developing a scale involves a lengthy and detailed procedure right from developing an exhaustive list of items to narrowing down upon the final instrument. Outlined below is the step-wise procedure followed consisting of 6 stages:

Stage 1: Developing the Items

Once the various categories/variables are decided (Usage, Reasons/Usefulness and Network Potential in the present case), an exhaustive list of items is prepared. The group sat together and listed down any/all questions that could be related to the topic under discussion. Although the scale was for measuring usage and potential with respect to social networking tools, questions related to internet usage and tech-savviness were also included at this stage. This is because these factors are indirectly related to the variables that were to be measured, i.e. it may be argued that a person who has high usage of social networking tools would have developed a high level of tech-savviness along the way, or spending more time on the internet would expose an individual to more of these tools and perhaps enhance his potential. It was ensured that all questions that could be thought of were included in this exhaustive master list – no member rejected another’s suggestions at this stage. After this exercise was completed within the group, a list of approximately 80 items was generated for the next step.

Stage 2: Brainstorming

This is the stage where questions are examined properly, especially with respect to their logic and relevance.

Face validity is conducted as a part of brainstorming. Face validity is a test to check whether the list of items “seems” to meet the requirements of measuring network usage and potential. It is not really “validity” in the technical sense of the term, as it only aims to check whether the items “appear to” (hence the term “face”) measure what they are supposed to. It is a very rudimentary test, based mainly in intuition and considering its nature and objective, is not performed by experts. A few of our own peers were explained the objective of our research, as also the concept of the Network Usage and Potential Scale. They were then asked to browse through the questions and comment on which ones would not be relevant to our objective. Answers were not accepted at face value, but several arguments took place back and forth in order to arrive at a consensus.

At this stage, we also examined the feasibility of the items with respect to respondents’ ability to answer them. For example, we wanted to measure the daily internet usage of an individual, but we realised that since this would fluctuate wildly for many users, we would be able to get only an approximate value of the average. Thus the range which we offered in the options was a wide one. In a similar way, the options for several items were changed after considering whether it was feasible for respondents to answer them.

On the basis of the above procedures, certain items were dropped. The remaining items (72 in number) were then categorised into the three variables they purported to measure, i.e. Usage, Reasons/Usefulness and Network Potential (complete list in Exhibit 1).This was done to facilitate the next step.

Brainstorming is an accepted way of achieving an initial set of items. Our scale development methodology incorporates further expert review of such an initial list of items and hence is expected to provide us with an exhaustive and representative list of items.

Literature also suggests that scales like Likert Scale are unidimensional in nature. They try to assign score to items along a uniform measurable parameter from higher to lower or greater to lesser and the like. However, there are evidences in use where multidimensional scales have been developed owing to the special nature of the model under study. For example, if we measure achievement (academic) on a single dimension, one would place each and every person on a line that ranges from low achievers to high achievers. However, it will be difficult to get a score for someone who is a high science achiever and terrible in maths, or vice versa. A unidimensional scale is incapable of capturing that type of achievement.

We feel that since our desired scale tries to categorically use items to map each of three factors mentioned above, we expect the items to fall on separate dimensions and hence vary in their nature.

Stage 3: Content Validity

This is the stage where the items actually undergo a thorough scrutiny. 8 experts were asked to evaluate the items (i.e. the categorised 72 items remaining from the previous stage), on the basis of their importance and relevance to our research objective. 8 experts from diverse fields were chosen by us for this exercise. They are:

Prof. Soumyakanti Chakraborty – Having done an FPM in Management Information Systems, he is an Assistant Professor at XLRI School of Business & Human Resources.

Gayatri Makhijani – After doing her B.M.M. (Advertising) from Jai Hind College, Mumbai, she worked in the field of Internet Advertising and is currently employed at Sapient Corporation.

Aditi Shukla – Having done her B.E. in Information Technology, she worked with Yahoo Software Development India as a Software Engineer for almost 2 years.

Kabardhi Pashuparthi – Has worked as a Senior Applications Engineer at Oracle India Private Limited and is presently pursuing his MBA from ISB, Hyderabad.

Saradha Govindarajan – She holds a B.E. in Geoinformatics and has worked in the field of Market Research at AC Nielsen.

Rohit Pande – An Electronics & Communication Engineer, he is currently working as an analyst at Mu Sigma based in Richmond, Virginia at the Microsoft Headquarters.

Riddhi Deb – She holds a M.Sc. in Physics and worked in the field of Market Research for over 2 years at International Marketing Research Bureau, India.

Sharad Pandey – An engineer, he was worked for over 5 years in the field of Application Development and was last working with Oracle India Private Limited.

We sent the experts an excel sheet with the categorised questions. Against each question they were asked to mark their opinion on whether the question is:

• Essential
• Relevant but not Essential
• Irrelevant

Their responses were subjected to a test developed by C. H. Lawshe. The method developed by Lawshe tries to determine the level of agreement between the experts. A formula is used to generate the Content Validity Ratio, which is:

Content Validity Ratio = (ne – N/2) / (N/2)

Where, ne = Number of panellists indicating that item as essential

     
       N = Total Number of panellists

Based on the number of panellists, this ratio must have a minimum threshold level for each item. Since number of panellists in the present case was 8, for each item the Content Validity Ratio had to be at least 0.75 for it to be retained in the instrument. This is considered a reasonable measure to ensure that agreement between experts is not merely by chance. Those items falling short of this requirement were dropped at this stage itself. Based on this test (results outlined in Exhibit 2), 24 items were dropped and remaining 48 items (complete list in Exhibit 3) were used during the next stage.

Stage 4: Phase 1 Survey

The Phase 1 Survey was conducted using the resulting questionnaire from the earlier stage comprising of 48 items. Since the very objective of the report is related to online social networking tools, it was completely feasible to conduct the survey online as all our respondents would have access to internet facilities. Other than the 48 questions, we also asked the participants’ Name, Age and Gender, to keep track of the variety of responses. A snapshot of the survey page is available in Exhibit 4.

We floated the survey to all the XLRI batch of 2009-2011 students and received 98 responses and a few of our personal contacts. The demographics of the survey respondents included about 90% of the respondents from the age group 20-30 years with the male to female gender ratio of 4:1. Since the survey measured a variety of factors, there was no single scale used. For example, there were certain questions which required only a yes/no response, while there were others that aimed to measure the average time spent on a particular activity. A single scale, like the Likert scale could not appropriately measure responses on both questions. Thus for each question, responses were designed specific to that item. However, the final score which was obtained from each item was uniformly in the range of 1 to 5. A higher score always corresponded to a higher network usage/network potential. Each of the items constituted a variable in the SPSS data set and they were all from the “scale” category.

Based on the responses (snapshot given in Exhibit 5), a number of tests were carried out to measure the reliability of the scale. The various tests and their results are summarised below:

Item-Total Correlation

Since the entire instrument aims to measure the same dimensions, it logically follows that a person who scores high on individual items must have an overall high score on the entire scale. Thus we calculate the correlation between the score of the respondents on each individual item and their total score. This is a test of consistency and thus only items which have an item-total correlation greater than 0.4 were retained. As a result of this analysis, 17 questions were dropped from the instrument.

Inter-Item Correlation

Next, we check for the consistency between questions. The correlation of each item with every other item is calculated. Each item must correlate with at least 50% of the remaining items to the extent of 0.35. If it does not, then we may infer that it does not measure what the rest of the instrument is measuring, and thus we drop it. As a result of this analysis, 8 questions were dropped from the instrument.

Exploratory Factor Analysis

We conducted an exploratory factor analysis on our questionnaire using the SPSS software. From the EFA, we got the following results

• Kaiser-Meyer-Olkin measure of sampling adequacy = 0.824 which is significantly greater than the prescribed 0.6.
• The Bartlett’s test of Sphericity has a significance level of 0.00 which is significantly less than the prescribed limit of 0.05.
• From the rotated component matrix, 7 factors were extracted.

On examining the group of questions within each factor, it was observed that the questions were not homogeneous among themselves as a group. It seemed more relevant to use the pre-decided categories determined by us based on brainstorming and logical reasoning which were: Usage, Reasons/Usefulness and Network Potential. Thus the EFA test was conducted again, after setting 3 factors as output to avoid spurious behavior of the test. This time from the rotated component matrix, we got 3 factors containing 8, 7 and 3 factors respectively. Keeping a threshold value of 0.6, we eliminated 6 items. So, our survey for phase II consisted of 18 items in total.

Split-Half Reliability

This test aims to measure the internal consistency between the items in two halves of the instrument. The questionnaire is divided into two halves (SPSS divides it into a lower half and upper half) and the correlation between the total scores of the items in each of the halves is calculated. Based on the responses to our survey, the Guttman Split-Half coefficient worked out to 0.64. Since the minimum required is around 0.6, the items in our instrument may be said to have sufficient level of internal consistency.

Cronbach’s Alpha

Another measure of internal consistency, it uses the following formula to determine the level of reliability:

a = KK-11- i=1Ksi2st2

Where: K = number of items in the scale

     
      &nbsp si2= The variance of scores on item I across respondents

st2 = the variance of total scores across subjects where the total for each respondent represents the sum of the k item scores

Based on the responses obtained from the Phase 1 Survey, the Cronbach’s Alpha works out to 0.874, which is a very good indicator of internal consistency.

Stage 5: Phase 2 Survey

The Phase 2 Survey was conducted using the resulting questionnaire from the earlier stage. Once again, this was an online survey consisting of 18 items. A snapshot of the survey page is available in Exhibit 6.

We floated this survey to all of our online contacts (almost 300 respondents) and received 118 responses (from different individuals as compared to the phase 1 survey). The demographics of the survey respondents included about 70% of the respondents from the age group 20-30 years, 20% respondents from the age group 31-40 years and with the male to female gender ratio of 3:1. These responses were again measured on a number from 1 to 5 (the questions were the same and hence this aspect did not change). Based on these responses (Snapshot given in Exhibit 7), certain tests were once again conducted.

Confirmatory Factor Analysis

We conducted a confirmatory factor analysis on our questionnaire using the AMOS software. From the CFA, we confirmed that our results from EFA stand correct and our 3 factors are apt. We kept a threshold value on regression factors of 0.7. The following are the path diagrams for each of the 3 factors:

Nomological Validity

In order to check for Nomological (Construct) Validity, we develop a consequent variable, i.e. we try and find a variable which would measure the effects of having a high network usage and potential. On brainstorming, the group narrowed down upon “Network Influence” as a consequent variable.

“Network Influence” basically measures how much a person can influence a network or be influenced by it. For example, we often come across people who are always approached by those in their peer group for any information, even if it is other than his/her area of expertise. This is mainly because he/she has an extended network which can be tapped for this information. Alternatively, a person who has an intricate network will always try and look within his/her network as a primary source of information on unknown topics. This is the meaning of our concept of Network Influence. We hypothesise that a person with high network usage and potential will also have high network influence. This should follow logically – an individual who is well-connected, i.e. has high network potential, and also uses his networking tools often, i.e. high network usage, would be widely known to have a vast resource pool to tap, and hence be approached often. Alternatively, because of his affinity to his networks, that would instinctively be the first resource he would think of when faced with an unknown.

The questions for Nomological Validity have been included in the Phase 2 Survey (Question Nos. 19 to 21 of Exhibit 8).

In order to check for Nomological Validity, the correlation between each of the three questions and the remaining questions in the instrument is checked. Based on the responses to the Phase 2 Survey, the correlation for each of the three questions works out to 0.81, 0.73 and 0.71. Thus the Nomological Validity is reasonably established.

Stage	Number of Questions Initially	Number of Questions Dropped	Questions Dropped	Number of Questions Retained
Content Validity	72	24	5, 6, 7,8, 9, 11, 12, 13,17, 19, 22, 25, 31, 32, 34, 35, 37, 43, 51, 53, 55, 56, 58, 63	48
Item Total Co-relation	48	17	1,2,3,4,6, 7, 9, 11, 14, 15, 16, 17,18, 19 20, 25, 41	31
Inter Item Correlation	31	7	5, 8, 12, 13, 33, 34, 46	24
Exploratory Factor Analysis	24	6	10, 22, 30, 31, 37, 45	18

Stage 6: Final Scale

So, after all the tests were conducted, we were left with 18 items/questions in our survey.

Having performed all types of validity tests, we arrived at the final scale, which is reproduced below. The score for each item is given against the option in the response. The final score obtained from the entire scale will be a sum total of the scores from each of the responses.

Usage of Scale

The scale should ideally be administered to a group. This is because currently there is no absolute measure in place to benchmark a person. Once the scale is administered to a sufficiently large number of people and the results are recorded, we may isolate the top 1% of respondents’ scores and develop a benchmark figure from there. When the scale is administered to a group of respondents, those with the highest scores (say top 1% to 5% depending upon total number and requirement) should be the ones shortlisted for any of the uses mentioned earlier in the report.

Limitations of Research

• Scale may be useful only for a group of people; measuring an individual’s Network Usage and Potential would yield no results as there would be no benchmark to derive any conclusions.

Usage measures their past behaviour over a period of time. There is no guarantee that past behaviour (extent of usage) would be replicated in the future. Also, it is possible that usage may have fluctuated wildly – measurement would take place only at a certain point of time and thus may not be an indication of actual usage.

If there are any anomalies at an individual level, these would not be detected as data has been studied only at an aggregate level.

The questions on Nomological Validity only take perception of the respondent into account and may not be entirely reliable in all cases.

A uniform Likert Scale has not been used on all items. Since there were certain questions which required only a yes/no response, while there were others that aimed to measure the average time spent on a particular activity, it was not possible to use the same scale to measure all items.

Future Research Potential

• Factors to be studied may be extended from the present 3 to a more comprehensive list
• A test re-test reliability may be conducted to fine-tune the scale further
• A uniform scale may be used for all questions

Implications

Implications of our scale may be considered on three fronts:

Academic

Based on the literature review covered earlier in the report, we notice that there is no measure to examine both Network Potential and Usage together. Thus our scale aims to extend the existing literature to a certain extent. There is also scope for further research on this topic, which would have significant academic implications.

Analytical

The data collected by us for the purpose of scale development may be used for other avenues as well. For example, the responses of people on the question for Internet Usage may be used to analyse the patterns in the use of people irrespective of their netwrok usage. A question-wise analysis would yield valuable data.

Practical

The primary use of the scale is for marketers. They may target individuals with high scores on the scale for two reasons: a) Advertising – The interface used by the high scorers in their respective social networking tools may be targeted for internet advertising because of the high reach and frequency which would be typical in their case. b) Viral Marketing – By targeting and marketing to a high scorer, you may convince him of your offering and this would ensure wider and faster dissemination of information through the word-of-mouth method.

References

• Jepson, Kevin., Credit Union Journal, 1/21/2008, Vol. 12 Issue 3, p16
• http://www.socialresearchmethods.net/kb/scalgen.php
• http://www.socialresearchmethods.net/kb/scallik.php
• Zhang, Jie and Daugherty, Terry; ‘Third-Person Effect and Social Networking: Implications for Online Marketing and Word-of-Mouth Communication.’ American Journal of Business; Sep2009, Vol. 24 Issue 2, p53-63
• Klaassen, Abbey, Bulik, Beth Snyder, ‘ Isn’t the entire web social these days?’ Advertising Age, 00018899, 3/30/2009, Vol. 80, Issue 11
• Nair, Neh Vivek, ‘ “Social Networking” on the Generation Next.’ , SCMS Journal of Indian Management; Jul2009, Vol. 6 Issue 3, p43-52
• Calin and Carmen, SOCIAL NETWORKING: REASONS TO JOIN AND THINGS DONE BY THE ROMANIAN CONSUMERS – AN EXPLORATORY ASSESSMENT.’ Annals of the University of Oradea, Economic Science Series; 2009, Vol. 18 Issue 4, p869-873
• HOOD, Tom — Interviews, ‘ How to Leverage Social Networking.’ ‘ Journal of Accountancy; Aug2009, Vol. 208 Issue 2, p26-29
• http://en.wikipedia.org/wiki/Social_Intelligence_Architect
• http://en.wikipedia.org/wiki/Social_network
• http://kdl.cs.umass.edu/software/documentation/tutorial/ch06s07.html
• http://www.ncbi.nlm.nih.gov/pmc/articles/PMC1466914/
• Agarwal, Shalija and Mital, Monika; ‘ An Exploratory Study of Indian University Students’ Use of Social Networking Websites: Implications for the Workplace
• Trusov, Bucklin and Pauwels; ‘ Effects of Word-of-Mouth versus Traditional Marketing: Finding from an Internet Social Networking Site, Journal Of Marketing, 2009
• Steffes and Burgee, ‘ Social Ties and Online Word of Mouth’ Internet Research, Volume 19, No. 1, 2009