Social Media Metrics -A Framework and Guidelines for Managing Social Media

Journal of Interactive Marketing

Special Issue: Social Media Think:Lab Thought Leaders‘ Summit

Social Media Metrics -

A Framework and Guidelines for Managing Social Media

Kay Peters*

Yubo Chen

Andreas M. Kaplan

Björn Ognibeni

Koen Pauwels

* Corresponding Author

1

Professor of Marketing, SVI-Endowed Chair, University of Hamburg, Esplanade 36, 20354 Hamburg, Germany, email: kay.peters@uni-hamburg.de, Tel. +49 (40) 42838-9510; Visiting Professor at the Graduate School of Management, University of California Davis, USA

2

Professor of Marketing, School of Economics and Management, Tsinghua University, Beijing, China 100084, email: chenyubo@sem.tsinghua.edu.cn.

3

Professor of Marketing, Director of Brand and Communication, ESCP Europe, 79 avenue de la République, 75543 Paris cedex 11, France, email: Kaplan@escpeurope.eu

4

Managing Director, BuzzRank, Rosenstrasse 3, 20097 Hamburg, Germany, email:

bjoern@ognibeni.de.

5

Professor of Marketing, Graduate School of Business, Oezyegin University, Nişantepe District,

Orman Street, 34794 Çekmeköy–Istanbul, Turkey, email: koen.pauwels@ozyegin.edu.tr

Social Media Metrics -

A Framework and Guidelines for Managing Social Media

Abstract:

Social media are becoming ubiquitous and need to be managed like all other media that organizations employ towards their goals. However, social media are fundamentally different from any traditional or online media because of their social network structure and egalitarian nature.

These differences require a distinct measurement approach as a prerequisite for their proper analysis and subsequent management. To guide organizations and researchers in developing the right social media metrics for their dashboard, we are providing them with a tool kit consisting of three novel components: First, we theoretically derive and propose a holistic framework that covers the major elements of social media, drawing on theories from sociology, marketing, and psychology. We continue to support and detail these elements, namely motives, content, network structure, and social roles & interactions, with recent research studies. Second, based on our theoretical framework, the literature review, and our practical experiences, we suggest 9 guidelines that may prove valuable for designing appropriate social media metrics and constructing sensible social media dashboards. Third, we derive managerial implications and suggest an agenda for future research.

Keywords: Social Media; Key Performance Indicators; Dashboard; Return on Invest; Learning

Theory; Interactionist Social Theory; Network Theory; Attribution Theory; M-O-A

Paradigm

Social Media Metrics -

A Framework and Guidelines for Managing Social Media

Introduction

Social media are becoming an ever more important part of an organization’s media mix.

Accordingly, organizations are starting to manage them like traditional offline and online media (e.g., Albuquerque et al. 2012, Hartmann 2010; Zhang et al. 2012). To this end, many

organizations subsume social media metrics into their marketing dashboards as a reduced

collection of key performance metrics (Pauwels et al. 2008). In a first approach, managers may be tempted to apply the concepts of traditional media metrics for the measurement, analysis, and management of social media.

However, social media are substantially different from the other media (e.g., Godes et al.

2005, Hoffman and Novak 2012). In contrast to other media, they rather resemble living, interconnected, egalitarian and interactive organisms beyond the control of any organization.

Thus, they require a distinct approach to measurement, analysis, and subsequently their management. But what are these fundamental differences of social media, what are their main interacting elements, how should organizations and researchers capture them in metrics for their analysis, what can they learn from the extant marketing literature, and how should they integrate such metrics into their social media dashboards?

To guide organizations and researchers in developing the right social media metrics for

their dashboard, we are providing them with a tool kit consisting of three novel components: First,

we theoretically derive and propose a holistic framework that covers the major elements of social

media, drawing on theories from sociology, marketing, and psychology. We continue to support

and detail these elements, namely motives, content, network structure, and social roles &

interactions, with recent research studies. Second, based on our theoretical framework, the literature review, and our practical experiences, we suggest 9 guidelines that may prove valuable for designing appropriate social media metrics and constructing sensible dashboards. Third, we derive managerial implications and suggest an agenda for future research.

The Framework: Theoretical Foundation

For the derivation of guidelines on how design appropriate metrics and construct a sensible dashboard for social media we require a proper framework. We develop this framework by first defining what constitutes a metric, a dashboard, and social medium. In a second step, we derive a framework from theoretical considerations and support it through the recent literature on social media.

Definitions

Metric. Farris et al. (2006) define a metric as a measuring system that quantifies a trend, dynamic, or characteristic. More generally, one could argue that metrics either describe or

quantify a state, i.e., characteristic, or a process, i.e., a dynamic, trend, or evolution. Additionally, states or processes may be stochastic and thus require additional information on the level of certainty, i.e., likelihoods or variance. In research as well as business, metrics are employed to define goals, measure their degree of completion or the deviation from them, and subsequently implement measures to improve on these metrics. Farris et al. (2006) categorize different types of metrics: Amounts (e.g., volumes, sales), percentages (e.g., fractions or decimals), counts (e.g., unit sales or number of competitors), ratings (e.g., scales), and indices (e.g., price index).

Ailawadi et al. (2003) summarize 7 requirements from a 1999 MSI workshop on brand

equity metrics. The ideal metric should be (1) grounded in theory, (2) complete (encompassing all

facets), (3) diagnostic (able to flag downturns or improvements and provide insights into the reasons for the change), (4) able to capture future potential, (5) objective, (6) based on readily available data, (7) a single number, (8) intuitive and credible to senior management, (9) robust, reliable and stable over time, yet able to reflect changes in brand health and (10) validated against other equity measures. For social media metrics, we believe relevant challenges include theory grounding, completeness and diagnosticity (1-3), being intuitive and credible to senior

management, and reliable over time (8-9). Due to the distinct nature of social media, we will argue that objectivity may be rather replaced by inter-subjectivity and pragmatic corridors of comfort (5) and add that convenience of available data or metrics should not preclude the construction of theoretically indicated and more important metrics (6). In contrast to brand equity measurement, we suggest the need for balancing metrics to sufficiently describe phenomena in social media (7) – we agree with Ambler and Roberts (2006) that a pursuit of a single silver-bullet metric in this context is ill-advised and will subsequently illustrate this challenge with examples. In this paper, we offer a holistic framework for a reasonably complete and diagnostic compilation of social media metrics and pay special attention on their over-time dynamics and reliability issues. As for credibility to senior management (8), social media metrics need to be connected to marketing actions and other dashboard metrics, and related to financial consequences, i.e., relevant

outcomes. As the latter is not the focus of our paper, we refer to De Haan, Pauwels, and Wiesel (2013) and Sonnier, McAlister, and Rutz (2011) for a demonstration.

Dashboards. As we already indicate above, a single metric can rarely sufficiently describe

all relevant aspects to a goal, let alone social media. Hence, a sensible collection of metrics is

usually required to guide managers towards the completion of their goals. Pauwels et al. (2008)

define a dashboard as “a relatively small collection of interconnected key performance metrics and

underlying performance drivers that reflects both short- and long-term interests to be viewed in

common throughout the organization.” They identify the challenges related to constructing reasonable marketing dashboards as (1) the integration of various qualitative and quantitative business activities to performance outcomes, (2) measuring short- and long-term marketing results and assessing the evolution of marketing assets, and (3) isolating marketing effects from a plethora of other influences in the market place (compare also Ambler 2003). An effective dashboard reflects a shared definition and understanding of key drivers and outcomes within the firm, diagnoses poor or excellent performance, allows for evaluating actions, enables organizational learning, and supports decision making to improve performance (Pauwels et al. 2008). Several steps are proposed on how to construct a dashboard: (1) select key metrics, (2) populate with data, and (3) connect the metrics with each other as well as to financial outcomes (e.g., Reibstein et al.

2005). Connecting the metrics is important to make the dashboard insights actionable, the latter is needed make it accountable.

However, the recent fragmentation of (social) media, proliferation of additional sales

channels, and the advent of “big data” manifested in the collection of user generated content

(UGC) on the web and in social media presents considerable challenges to the design of traditional

marketing dashboards, or subsumed social media dashboards. Accordingly, Fader and Winer

(2012) recommend a rather cautious approach to rich UGC data: it requires processing of vast

amounts of data of which most is qualitative in nature and may take a life of its own. New metrics

from social media such as likes, followers, and views might be simple, compare to traditional

media and are therefore tempting for firms to focus on. But such metrics may not reflect the

important aspects of social media and therefore mislead marketing efforts in a way that may even

harm an organization’s prospects. Hence, any new metric that will be suggested for such new

types of data or media requires a theoretical foundation and subsequent research to explore the

relationship between input, metric, and financial outcomes. Only then such metrics can be

integrated into appropriate general marketing or subsumed social media dashboards. A prerequisite for doing all this is to define and understand social media.

Social Media. The term ‘Social Media’ is a construct from two areas of research,

communication science and sociology. A medium, in the context of communication, simply is a storage or transmission channel to store or deliver information or data. In the realm of sociology, and in particular social (network) theory and analysis, social networks are social structures made up of a set of social actors (i.e., individuals, groups or organizations) with a complex set of the dyadic ties among them (Wasserman and Faust, 1994, pp. 1-27). Combined, social media are communication systems that allow their social actors to communicate, i.e., exchange information, along dyadic ties. As a consequence, and in stark contrast to traditional and other online media, social media are egalitarian in nature. A brand is only another node in the network, not an authority that can impose an exposure to commercial messages as in other media. Of course, we see attempts to use banners or “sponsored stories” in such networks that mimic classic display advertising. But those messages are diametrically opposed to the dialogic nature of social

networks built on individual relationships, as they often rudely interfere in other users (intimate) conversations on (unrelated) issues.

Across social media, Alba et al. (1997) describe this dyadic relational interactivity as the main differentiating characteristic of social media compared to other traditional offline and online media: a medium needs to be multi-way, immediate, and contingent to qualify as a social one.

Stewart and Pavlou (2002) add that social media may have different degrees of interactivity, and

that for understanding contingency, context and structure, goals, sequences of actions and

reactions, as well as the characteristics of the respective medium need to be known. In addition,

Kaplan and Haenlein (2010) assign social media via social presence and self-presentation into six

different groups: (1) collective projects (e.g., Wikipedia), (2) blogs and microblogs (e.g., Twitter), (3) content communities (e.g., YouTube or brand sites), (4) social networks (e.g., facebook, myspace, linkedin), (5) MMORPGs (e.g., World of Warcraft), and (6) social virtual worlds (e.g., SecondLife). Taken together, these definitions already suggest that social media may require distinct metrics compared to traditional media, capturing in particular

 their network characteristics, i.e., actors and dyadic ties,

 the dynamics that reflect their immediate and multi-way nature,

 the contingency aspects of information exchanged, and

 the specifics of the respective social medium.

The distinct nature of social media thus prohibits the simple transfer of metrics from traditional media. In an initial step, the metric development for social media requires a new theory-driven framework that accounts for these fundamental differences and provides an appropriate foundation.

A Theoretical Framework for Understanding Social Media

From a marketing perspective, ‘understanding’ translates into relating marketing input via social media with desired marketing outcomes. Only if this connection is sufficiently understood, learning from observed outcomes may improve future decision making on marketing inputs. This logic relates to the Stimulus (S) → Organism (O) → Response (R) paradigm with its feedback loop from Social Learning Theory as an overarching frame as depicted in Figure 1 (e.g., Bandura 1971; see Belk (1975) for an early marketing application).

[INSERT FIGURE 1 ABOUT HERE]

Initially, marketing inputs (Stimuli) and outcomes (Response) are assumed to compare to frequently used marketing instruments (e.g., information, advertising, pricing) and (intermediate) success metrics (e.g., awareness, brand liking, market share, sales, profit; Farris et al. 2006). As social media, according to the definition above, constitute a different kind of organism compared to traditional media, they require a closer investigation.

As described above, ‘social media’ are media that mimic or reflect social systems, which are networks of actors that are linked through relational patterns. As social network theory and subsequent network analysis focus on ties, both take predominantly a relational perspective, i.e., observed effects are primarily investigated through the properties of relations between actors, instead of the actors’ properties (e.g., Burt 1980). In a first extension, Blau (1974, 1977) pushes this relational view further, taking a rather macro-level perspective and describing the network through a set of multidimensional parameters, i.e., nominal parameters (e.g., age) or gradual parameters that rank order members (e.g., income). A striking feature of these parameters is that they pertain to distributions and dynamics rather than states, i.e., when describing a social network income distribution (heterogeneity) is more important than income itself, or the evolution of income (distribution) is more important than the underlying states. Blau (1977) suggests that inequality in distributions impedes and heterogeneity promotes intergroup relations. This hypothesis has been extended by Granovetter (1973, 1983) who explores the “strength of weak ties” in social structures with respect to word-of-mouth or innovations, linking the insights from sociology to the marketing domain. In a second extension, Burt (1980) describes certain models of network structures. He distinguishes network model approaches along two dimensions, the

aggregation level of actors and the reference frame within an actor is analyzed. The aggregation

level reaches from micro-level (i.e., actor related analysis of ties) via intermediate or meso-level

(i.e., multiple actors as subgroups) to macro-level models (i.e., actors or groups as a structured

system). In the second dimension, he categorizes network analysis approaches as “relational”

when the intensity of actor pairs is the focus of analysis, and as “positional” when all defined relations between actors need consideration to evaluate a relative position in a given network.

Accordingly, and lending from sociology its theoretical foundations, we add the network structure (actors and ties) to the organism description in our framework (see Figure 1).

A common criticism of network theory refers to its rather technical and relational view of a social network, thus ignoring the specific content that is exchanged along the ties as well as the actors of the social network. Social network analysis mostly infers the importance of a dyadic link from the intensity of its use, i.e., how often information is exchanged and in which direction it travels. The content of communication is rarely paid attention to. However, in marketing research the content of communication is at least as important as its frequency. In particular, the impact of a piece of communication depends on many situational factors, not least the actor and her

motivations. Hence, to augment the actors’ as well as the content perspective, we draw on the Motivation, Opportunity, and Ability (M-O-A) paradigm as elaborated by MacInnis, Moorman, and Jaworski (1991): Motivation is defined as goal-directed arousal (e.g., Park and Mittal 1985), i.e., the desire or readiness to process information, Opportunity as the extent to which distractions or limited exposure time affect actors’ attention to information (e.g., Batra and Ray 1986), and Ability as the actors’ skills or proficiencies in interpreting information, e.g., given prior

knowledge (e.g., Alba and Hutchinson 1987). In our context, we suggest motivation to be the most important aspect describing actors’ aspirations best. We refer to the opportunity and ability of actors to digest information later when we derive our rules guiding the development of metrics.

Accordingly, we add Content and Motivation as further elements to our framework in Figure 1.

Given the three poles of actors’ Motivation, the Content that travels along the dyadic ties, and the Network Structure which describes the underlying infrastructure of nodes and connections, we observe social interactions taking place. At each node or actor, information is not only received and simply forwarded, but it may also be perceived, evaluated, and subsequently altered in many ways. Consistent and sustained actions may earn actors certain social roles within their network.

Interactionist social theory defines social roles as neither given nor permanent. A social role is continuously mediated between actors in a social network, especially by observing and copying the behavior of others. That happens in an interactive way, i.e., any role is contingent on the other actors in oscillation between cooperation and competition. As social rules are dynamic concepts, they are constantly shaped through the process of social interactions, which sociology defines as a dynamic, changing sequence of social actions between individuals or groups. However, all actors constantly try to define their current situation, strive for a superior social role and try to sign up other actors in support of it (Mead 1934).

To better understand why and how social roles are assumed or assigned, and why and how content is altered through social interactions in certain ways, one may draw on the insights of attribution theory. In essence, attribution theory posits that actors in social networks strive to assess the true properties of objects of interest. To ascertain the external validity of their perceptions, actors usually employ the co-variation principle across multiple observations to attribute it toward (1) a distinct stimulus object in the entity dimension (i.e., content), (2) herself or an observer in the person’s dimension (i.e., other actors and their particular motivations), or (3) the context described by time and modality (e.g., here the network perspective) (e.g., Mizerski,

Golden, and Kernan 1979, p. 126). As such, attribution processes are assumed to instigate such

activities as information-seeking, communication and persuasion (Kelley 1967, p. 193). Any

attributions are made on a background of antecedents, are governed by informational dependence

(in particular with respect to ties in the social network) and result in social influence (given the position or role of an actor in the network; Kelley 1967). Important antecedents of attributions are the motivation of the actor herself, her prior beliefs, and prior information received (e.g., Folkes 1988). Hence, the particular impact, subsequent modification, and further sharing of a piece of content received by an actor may depend on her own as well as the sender’s suspected

motivations, the type of content and way it is framed, the (social role or) position of the sending person in the network as well as who else received it at the same time. Hence, at the intersection of the three poles we observe social interactions and modifications which lead to the assumption of social roles over time, and which feedback into motivations, content, and network structures, Accordingly, we add a complementing fourth element to the framework in Figure 1. Collectively, these four elements describe the Organism “Social Media”.

Given the theoretical derivation of our holistic framework, we further support and detail the description of its elements by extracting the current insights from the extant literature on social media.

Motivation. Edvardsson et al. (2011) suggest a social construction approach to social systems, where social positions and roles in a network are the result of three interacting forces:

Signification or meaning draws upon interpretive schemes and semantic rules (e.g., based on

values and motivations), domination or control is exercised by drawing on unequal distribution of

tangible and intangible resources (e.g., abilities and network links), and legitimation or morality

refers to social norms and values that individuals use to evaluate behavior. Their finding contends

that value arising from social networks is asymmetric for the people involved and can only be

understood in the context of the network, and that the co-creation of this value is shaped by social

forces within social structures. Seraj (2012) analyzes characteristics that result in perceived online

community value. She identifies three value components: intellectual value, representing co- creation and content quality, social value, which originates from platform activity through social ties, and cultural value, which represents the self-governed community culture. Participants in social networks may take up to 7 different roles which can be attributed to combinations of these value components. Adjei, Noble, and Noble (2010) add that uncertainty reduction may be another motivation for consumer information exchange within brand communities, which is driven by the quality of consumer conversation. Eisenbeiss et al. (2012) relate individual motives of socializing, creativity, and escape via group norms, social identity, desires, and “we-intentions” to

participation behavior. They show that four segments are prevalent in their social network sample.

Three segments join because of a single motivation and only the smallest segment has multiple motives. Their finding underlines the case for heterogeneity when measuring inputs and outcomes on social networks.

Taken together, all suggested and empirically investigated motives are tied to the value created for the participating individuals. Consolidating the collective insights, we subsume them into the structure suggested by Seraj (2012): (1) intellectual value stemming from co-creation and content quality (Seraj 2012), which may subsume signification as described by Edvardsson et al.

(2011), creativity (Eisenbeiss et al. 2012), and uncertainty reduction (Adjei, Noble and Noble 2010), (2) social value from platform activities and social ties (Seraj 2012) which also entails domination (Edvardsson et al. 2011) as well as socializing, escape, and social identity (Eisenbeiss et al. 2012), (3) cultural value which represents the self-governed community culture (Seraj 2012) and subsumes legitimation (Edvardsson et al. 2011) and “we-intentions” (Eisenbeiss et al. 2012).

We add these three dimensions to the Motivation element in our framework (see Figure 1).

Content. De Vries, Gensler, and Leeflang (2012) take a more detailed view on how created content drives social media action. They first characterize the content along the dimensions of vividness, interactivity, information, entertainment, position, and valence. They continue to show that these characteristics influence the number of likes and comments differently. Van Noort, Voorfeld, and Reijmersdal (2012) also highlight the importance of interactive content on diverse cognitive, affective, and behavioral outcomes. Liu-Thompkins and Rogerson (2012) extend these findings to Youtube videos. Again, entertainment and educational value drive the popularity and ratings of videos, and the effect of network structure on popularity is nonlinear. Berger and Milkman (2012) investigate which characteristics make online content go viral. They find that activating content reflecting anxiety, anger, or awe, and content that is practically useful or surprising is more likely to go viral. Accordingly, valence of content alone is not sufficient to explain virality. More important, understanding collective outcomes requires investigating the quality of individual-level psychological processes actually driving social transmission, i.e., sharing. Kozinets et al. (2010) categorize such individual-level processes in the context of online word-of-mouth. They identify four different approaches to message conveyance in blogs mirroring different narrative styles in a coproduction environment, resulting in different quality aspects of content: evaluation, explanation, endorsement, and embracing. Each of these qualitative styles alters original marketing messages in very distinct but systematic way, depending on the forum, the communal norms and the nature of the marketing message.

Recent papers have also focused on the ability of firm-initiated marketing to generate

word-of-mouth (e.g., Trusov et al. 2009) of a specific content (DeHaan et al. 2013). The

distinction between firm-initiated and customer-initiated content also applies within the same

platform (DeHaan et al. 2013), even within a ‘social media’ platform. For instance, a sponsored

story on Facebook –or firm employees’ posting on the fan page– are firm-initiated actions that

may eventually trigger customers to forward such messages or to produce their own content.

Indeed, a key goal and benefit of firm-initiated marketing is its power to stimulate conversations around a brand or product, which then causes a social media ripple effect that ultimately increases business performance.

Taken together, it emerges that content may have three sufficiently distinct aspects, which of course may overlap to some extent. These aspects are content quality (subsuming

characteristics (e.g., interactivity, vividness), domain (e.g., education, entertainment, information) and narrative styles), content valence (subsuming emotions (e.g., anger, anxiety, joy), and tonality (e.g., positive, negative)), and content volume (subsuming counts and volumes).

Network Structure. Recently, Katona, Zubscek, and Sarvary (2011) provide a good

summary of the interdependencies between network structure and social relationships. First, the

embeddedness of individuals in their social networks is important when judging the influence of

relationships on individuals (i.e., the positional view according to network theory). Accordingly,

merely counting relationships or ties within the network is not sufficient (Krackhardt 1998; i.e., a

simple relational view according to network theory). Second, there are two synergetic effects for

triangles of relationships. Individuals that share the same contacts are better informed due to

redundant paths and those common ties are reinforced according to increased trust (e.g., Coleman

1988, Granovetter 1973). Third, individuals at the intersection of clusters may have greater

leverage on information flow (Burt 2005). In sum, several similar analyses in marketing suggest

accounting for both, the relational and positional view taken by network theory (see Appendix for

a detailed and extensive literature review on related network metrics).

These effects translate into certain network dimensions that describe social media in the context of network structure. These usually dimensions usually are (e.g., Freeman 2006,

Granovetter 1973, Hanneman and Riddle 2011, Kadushin 2012, McPherson et al. 2001, Moody and White 2003, Scott 2012):

 Size (e.g., the size of the total (number of actors) or local network (degree))

 Connections (e.g., homophily , multiplexity, mutuality, network closure)

 Distributions (e.g., centrality, density, distance, tie strength)

 Segmentation (e.g., clustering coefficient, betweenness)

Some of these measures are relatively new and partially account for either actor or content characteristics (e.g., homophily, multiplexity), while others still focus on technical relational or positional perspectives (e.g, degree, centrality). Following network theory and its model

structures, these network dimensions may help describing relational or positional perspectives at all network levels, i.e., the macro-, meso-, or micro-level of the network (e.g., Burt 1980).

Social Roles and Interactions. Above, we motivate the intersectional element of our social media organism by interactionist social theory and attribution theory. This can be underlined by drawing on selected studies already cited above. With respect to motivations, Edvardsson et al.

(2011) suggest that social positions and roles are the distinct result of interacting forces that draw on motives and network structure. Even theoretical models on network structure show that the structure of social networks can depend on the distribution of motivations, so the resulting structure is in effect endogenous (e.g., Galeotti and Goyal 2010, Ballester, Calvó-Armengol, and Zenou 2006). Within the content domain, Kozinets et al. (2010) suggest that altering of

(marketing) messages depends on the co-producing actors, social norms (i.e., their motivation) and

the original content itself. Hence, content is an input to the social interaction which is modified in

its course. From the network perspective, several sources (see above) suggest that trust and in turn

the social role of an actor within her social network is not only driven by network structure, but may also be driven by the repeated reception of consistent content from different actors as attribution theory posits. Zhang and Zhu (2011) show that a decreasing network size and content volume will have a negative impact on users’ incentive to contribute in social media, again underlining interaction effects between all elements of the framework. Other studies show the positive effect of network structure, i.e., higher closure coefficients and more redundant ties, and homophily (i.e., similar neighbors in the network) on behavior diffusion (i.e., in our framework social interactions; Centola 2010, 2011). Taken together, it supports the distinction of the intersectional element as a separate entity in our framework.

The literature on social roles and social interactions is rich on different classifications.

Social theories suggest family, tribal or functional roles as social roles that actors can take,

whereas the cited interactionist theory argues that social roles may be in flow and contingent. In

the context of social media we are currently aware of promising (e.g., Edvardsson et al. 2011,

Seraj 2012), but not yet consistent research results on social roles that actors assign or assume,

which constitutes a significant research gap. With respect to social interactions, social theories

suggest characteristics of social interactions (e.g., solidary, antagonistic, mixed, intensity,

extension, duration, organization), but do not provide explicit classifications. After an extensive

literature review and searching the web for various usage classifications on social media activities,

we consolidate several practitioner analyses on social media and arrive at “Sharing, Gaming,

Expressing, and Networking” as the currently dominating social interactions taking place on social

media. Again, we suggest this as an emerging area for further research to better understand what

actors of social media are actually spending their time on.

Combining all four elements with their different aspects results in our suggested framework depicted in Figure 1. Within any social medium, all four components interact

continuously, altering and reinforcing each other as in a living organism. Participating individuals are heterogeneous, and dynamics are inherent in all components as network theory and

interactionist social theory suggest. As individuals may participate in several social media like Facebook and Twitter, any social network may not be fully understood in isolation. Due to the egalitarian and networked character, firms automatically lose control of what happens with their marketing input when they enter this heterogeneous and dynamic territory. Hennig-Thurau et al.

(2010) recently describe this effect as marketing “pinball”. The framework also underlines the definition of social media based on social interactions and interactivity as distinguishing features:

Any reaction to marketing input will be immediate, multiway, and contingent. Hence, any dashboard first requires metrics that sufficiently capture the four elements of our suggested framework, before these metrics themselves can be related to marketing input and outcomes.

Guidelines for Designing Social Media Dashboard Metrics

Constructing sensible social media metrics for a company’s dashboard requires a holistic approach. Our theoretically derived framework guides companies and agencies to understand and capture the relevant phenomena in appropriate metrics. We refrain from reviewing marketing input and outcome measures as these are commonly known. A dashboard, however, requires linking marketing inputs via social media metrics to outcomes that correspond to the goals of an organization. Given the variety of organizations and social media, there is no such thing as “the”

dashboard or metric for social media. Every organization needs to construct chose the appropriate

metrics for its specific dashboard tied to its organizational goals, structure, social media selection,

etc.

However, the framework and its theoretical foundation yield some fundamental guidelines that organizations should observe when designing metrics and dashboards. As these guidelines are flowing from the underlying nature of social media, they should present some general insights that carry validity for any kind of existing social media as well as those yet to emerge. They should help organizations to avoid some often observed pitfalls and result in finely balanced dashboards enabling managers to consistently navigate their social media space (see Figure 2).

[INSERT FIGURE 2 ABOUT HERE]

Guideline #1: From Control to Influence

For brands, social media work differently compared to traditional media. In the traditional media setting, managers and agencies create and distribute advertising to consumers. They communicate indirectly via uni-directional media. All consumers that would like to watch TV or read a certain magazine, are to some extent exposed to this communication. Hence, managers have control and authority over brand communication. They also have a simple S-O-R scheme to test, where they can track the effectiveness of their input to higher awareness, and where they measure their success rate without anyone taking note. In sum, traditional media compare to control, inside- out talking, and disguised measurement.

In social media, brands and their managers are just another actor in the network. A

frequent analogy used is the transformation of brand managers from a lab scientist into just

another lab mouse. For instance, managers can still post content, e.g., advertorial videos or

comments, but whether someone dares to notice is decoupled from the consumption of the

medium. When the piece of content is not of interest to the initially linked actors in the personal

network of the brand, i.e., does not fit the motivations of directly linked consumers, the content

will neither be read nor, and that is even worse, be shared with third parties. In essence, sustained reach cannot be bought like in traditional media. To generate relevant content that fits the

motivational structure of the brand’s audience, managers need to know about the motivations of actors in the first place. Hence, any dashboard on social media requires the continuous assessment of motivational profiles via adequate metrics.

Assessing these motivations usually requires a dialog between the brand and actors in the network, and as a dialog is bi-directional, above all an organization needs the ability to listen.

Listening, understanding, and responding to individual actors changes the concept of traditional media in another meaningful way: previously pure inside-out communication turns into balanced outside-in communication. Besides building up the capabilities to listen and respond at the individual level, firms need metrics that monitor the listening and responding performance.

Traditional media metrics are mostly made for inside-out communication. Hence, we require new listening and response metrics for social media dashboards.

Overall, the dialog nature of social media propels brand communication from a patriarchal

to a participatory paradigm. Congruent with this change, brand managers move from a position of

control to one of influence: Only if their content hits the motivations of directly linked actors, they

will share it with others and subsequently build the reach the brand pursues. For building this

reach, it requires messages to ripple through the network. Accordingly, the initial personal

network of a brand may be less important than the combined networks of its followers, because

that is its potential second step reach. So when measuring influence in networks, it goes beyond

the number of likes or followers – rather it is the organic reach that counts. Such kind of metric is

not known in the traditional media environment, but essential in social media.

Another consequence of this participatory nature is that discourse may take place without the brand as an actor. So called “shitstorms”, i.e., bad communication spiraling out of control of a focal brand, are the extreme form of it. Hence, social media dashboards need not only metrics to listen to personal networks, but to the noise across social systems. The different levels that need to be monitored should be inspired by network theory as well as the requirements of the focal brand, i.e., the micro-level of direct followers, the wider community or relevant audience, as well as the overall system at the macro-level.

Finally, managers should be aware that everything they do in terms of listening, measuring and responding is often transparent in the social network, i.e., from likes to levels of activity, everything is not only known to the managers, but all “mice in the lab”. We add this here for closure but pick this issue up in guideline #5 in detail.

Guideline #2: From States to Processes & Means to Distributions

For traditional media, decision makers focus on metrics that express media performance in states rather than processes or dynamics. For instance, they measure the state of awareness, purchase intent, buys, etc. However, social media are based on networks, and network theory predicts that distributions, i.e., heterogeneity, and dynamics are more important than states when describing social systems (see theoretical discussion above).

The importance of dynamics has two important facets. First, it is the growth or decline in

numbers that is a relevant signal. For instance, a brand page can be very popular in terms of total

likes (a state), but if growth is slowing over a certain time –and this is transparent to all users in

the network– the relevance to other users is also declining. Hence, often the 1

or 2

derivation of

a state may be a more important metric to track than the actual state (e.g., Tirunillai and Tellis

2012). Additionally, these processes may exhibit certain path dependencies. In our extensive

literature review on empirical social media research (see Appendix), Moe and Schweidel (2012) highlight the path dependency of online reviews: If the initial review is a 5-star, subsequent users may only differentiate themselves by adding worse reviews, so that the average evaluation

deteriorates. Given a median of users, the evolution also crucially depends on the heterogeneity of the user base, e.g., when a core group of activists emphasizes negative ratings. Sun (2012) shows empirically with data from Amazon that a higher standard deviation of ratings lifts a product’s relative sales rank only when the average rating is below 4.1. These examples show that the unfolding dynamics depend as much on the motives and social roles of the activists as on the prevalent distributions with respect to content as well as to the network structure.

Second, also derived from network theory in conjunction with our intersectional element, the activity on ties or nodes is an important metric, not just the established link itself. Thus, it is the dynamics or intensity that qualifies a link or actor more than its mere presence (again, a state).

Traffic generated by actors as well as traffic on links can be easily measured online. The challenge for a brand is to filter out the right nodes and links to watch from the plethora in the network.

However, current social media tracking systems often already allow for the real-time tracking of prominent posts or tweets. And some new scores like the EdgeRank in facebook or the KloutScore across media assess the time-dependent importance of such network actors.

Accounting for both types of dynamics highlights the dilemma of static numbers: If the

number of total “likes” is high but static, and the activity of these “likes” is very low, this suggests

a previously interesting, but nowadays “dead” site. In sum, metrics that capture such network

dynamics and the underlying heterogeneity are crucial ingredients for social media dashboards.

Guideline #3: From Convergence to Divergence

For traditional media, organizations thrive on convergence towards better states reflected in metrics. For instance, the higher “brand sympathy” across the population the better.

In social media, however, divergence is not always bad. To the contrary, certain brands may thrive on adversity in social media as differentiation increases and reinforces the

identification of its core users. One example may be Abercrombie & Fitch who were the target of a social media campaign by users based on a seven year old quote of its CEO stating that the brand is indeed “exclusionary”. The resulting adverse reaction towards the brand in social media by users may nevertheless improve the identification of its core brand users. Additionally, whereas divergence and subsequently lower product evaluations may trigger substantial marketing audits offline, it may be a naturally evolving phenomenon in social media online (e.g., Chen, Fay, and Wang 2011, Godes and Silva 2012, Moe and Trusov 2011).

Another aspect associated with convergence and divergence is that within social networks it matters who says something to whom in what context, e.g., the objectivity of offline “means”

(where managers evaluate answers to specific questions) is replaced by a qualified inter-

subjectivity. This can be explained by looking at hotel ratings: If a young person was looking for a party hotel and had fun during holidays, it may give a great review. However, an older couple looking for rest will evaluate very negative for the noise it had to endure. The mean or converged assessment will be senseless without accounting for the actor’s motives, the content perspective of evaluation, and the targeted network population. Hence, metrics not only need to account for heterogeneity, but especially with respect to content also need to assess contingency aspects.

These contingency “key-words” for new metrics have to be defined for each brand in its individual

context.

Guideline #4: From Quantity to Quality

As we stated above, in traditional media usually states and quantities are key. We also highlighted previously, that dynamics in the form of intensity on nodes and links are key rather than the mere existence of nodes and links. At this point, we drive this insight even further and qualify intensity in more detail. In essence, we refer to the engagement levels of actors that are tied to motives, content, and social roles. Above we mention that a high number of “dead likes” is counterproductive when building a loyal base of followers. Hence, beyond simple “talk abouts”

(mentioning of keywords or brands), many social media dashboards measure the different types of interactions and categorize them by the associated level of engagement, e.g., a “like” has less value than a “comment” compared to “sharing” content and derive a respective score (see buzzrank interaction rate in Figure 3 for an example). EdgeRank and Kloutscore are similarly constructed, but at the individual level (see Figure 4).

[INSERT FIGURE 3 ABOUT HERE]

It is intriguing that such “engagement” levels are similar to what marketing research

predicts for theoretically derived involvement components that drive consumer actions. For

instance, Arora (1982) distinguishes three different levels of involvement –situational

involvement, enduring involvement, and response involvement– and analyzes their internal

structure. Situational involvement is casual and pertains to time and situation, whereas enduring

involvement depends on experience with a matter and its relationship with the actor’s value

system (or motives in our framework). Response involvement finally arises from enduring

involvement in conjunction with complex cognitive and behavioral processes. In social media,

higher engagement –or respective involvement– levels are crucial for generating sustained traffic

and dialog. In contrast, for many brands we currently still observe sweepstakes for generating

followers or posting unrelated questions to generate traffic in an attempt to boost static numbers.

Developing and employing adequate metrics to measure engagement levels of consumers –as well as their evolution and heterogeneity– will drive brand managers to more sincere and sustained modes of interaction, i.e., higher quality contacts. Such highly engaged fans, and not necessarily high numbers of them, are crucial in building sustained and authentic reach in social media. And in contrast to awareness levels in traditional media, these engagements cannot be “bought” in instances but need consistent nurturing over time. If these engaged actors additionally play relevant social roles in the network, they will also be the best defense in case of “shitstorms” (see above) when a brand itself can hardly do the right thing, but needs advocates to speak on its behalf. Accordingly, quality based metrics should be preferred over sheer volume numbers when constructing social media dashboard metrics.

Guideline #5: “Heisenberg”-Rule

Lending insights from physics, the Heisenberg rule suggests that once you try to measure something, you may alter its state and/or dynamics. We observe such phenomena in social media when social roles and related motives, i.e., profits, social, intellectual, or cultural value, are tied to certain metrics. Just think of KloutScore, EdgeRank, Google- or Youtube-Rankings which are measuring a user’s influence within or across social media. Figure 4 illustrates how these metrics are constructed, and as the underlying rules are again transparent, users start to play it when it is important to them. And again, motives, content, roles and network structure from the framework (see Figure 1) interact to produce such results.

[INSERT FIGURE 4 ABOUT HERE]

Another consequence is that as any such numbers will be gamed, they will be distorted

from the start. In contrast to traditional media, where managers can control the hidden

measurement of success, social media metrics may often require working at a certain level of fuzziness or corridor. We pick this issue up again later in our guideline # 9.

Guideline #6: “Ying & Yang”-Rule

The previous guideline suggests that important metrics will be gamed by users, i.e., in transparent and participatory environments like social networks there will hardly be a number that suffices alone. Additionally, in guideline #4 we suggest that metrics on states are not informative when not accompanied by metrics that capture dynamics and heterogeneity. The “Ying & Yang”- rule posits that for social media environments, dashboards need complementing metrics that balance each other. For instance, sustainable growth of a fan base can only be assessed by qualified growth combining metrics on growth and the quality of engagement levels, and a KloutScore can only work over time when balanced by a metric that punishes “fake” engagement and impact.

Additional aspects of this rule refer to the consistency and reliability of metrics. For example, networks are always in flow and change their size, composition, usage levels and structure like a living organism as they evolve. Hence, over time any metric that is employed in dashboards may deteriorate in consistency over time as its base shifts, adaptations are made (e.g., inclusion of a new social medium in KloutScore or changes to EdgeRank-calculations). Especially for new social media, early stages of diffusion inherently bias comparisons with later stages. For example, heavy users of social media tend to adopt earlier than people with lower usage.

Accordingly, average usage time may eventually go down over time as more people join these

media. These dynamics should hold at all levels of analysis, from the total network down to brand

hubs within those media. In comparison, if one assigns a sub-section of the dashboard to the

brand’s activists, then these metrics should be relatively comparable over time even as the number

of low involvement users keeps growing. Hence, we suggest constructing metrics that account for underlying dynamics and heterogeneity through base shifts or correct them for later changes when long-term evaluations are made.

Guideline #7: From General to Specific

Social media dashboard metrics need to cover all relevant social media for an organization.

On the one hand, consumers may use specific platforms for specific actions, e.g. Twitter to complain (as they desire a fast company reaction), Facebook to boast about successful purchases (as it only goes out to close friends) and Instagram to combine brand visuals. These social media may also have different characteristics that require different metrics, as Twitter is an asymmetric 1:n social medium compared to Facebook which is symmetric (n:n). On the other hand, users are active across social media and subsequently one can regularly observe spillovers, e.g., from Twitter to Facebook activity and vice versa. Both specific and spill-over effects encourage us to account for both in (social media) dashboards, general metrics at the meta-level across social media, and specific metrics that reflect the particular nature of any social medium.

Additionally, the guideline refers to the level of measurement within each social medium.

As network theory suggests, there are no metrics that cover all levels of a network, i.e., from the specific micro- via the meso- to the general macro-level. The same holds for the analysis at different levels of aggregation on content, motives, or their interactions in terms of social roles. As dynamics and heterogeneity are usually relatively high in social networks, any dashboard needs several layers of metrics than can be combined for specific analysis at specific aggregation levels.

Hence, many specific questions may require tailored approaches to measurement.

Guideline #8: From Urgency to Importance

Social media are living organisms. Accordingly, dashboards will always keep on blinking in real-time. Deviations, even substantial ones, are the rule rather than the exception.

Organizations that are used to traditional media are often overwhelmed by the pulse of social media, and knowing about path dependencies and rather quick reinforcement loops may jitter their nerves, tempting them to interfere rather sooner than later in conversations. But as we know from past experiences, interference may be just the wing of the butterfly that was required to send developments spiraling. Hence, when designing dashboards, organizations need to extract the essence of conversations, sentiments, and moods in the audience, but may also define a corridor of comfort which is defined via heterogeneity and dynamics around crucial metrics. Within this comfort zone, organizations need to let go.

Guideline #9: Balancing Theory & Pragmatism

Finally, we suggest balancing theoretical considerations with pragmatism when designing and implementing metrics for social media dashboards. As our framework underlines, there is a lot that brand managers can take-away from existing theories. The diversity of origins of these

theories, e.g., sociology, network analysis, marketing and psychology offers a rich pool of insights that may guide them towards sensible metrics. As social media mimic our social systems, the sheer complexity stemming from dynamics and heterogeneity, paired with their egalitarian nature, suggests more than a dose of pragmatism. However, this pragmatism should not lead to

complacency in the sense that someone measures what is convenient to measure or what seems

handily available. To some extent, this is natural when a field is still young and emerging – as it

appears when assessing the current state of the extant literature (see Appendix for recurring

available network measures compared to the holistic framework in Figure 1). Accordingly, as we

encourage as much theoretical consistency or rigor as possible when designing social media

metrics, we at the same time acknowledge that relevance needs to come first. To us, it is more important that the effort associated with implementing metrics is balanced by their relevance for the organization, and that metrics are actually tied to managerial implications.

Implications for Practice and Research

Based on our framework and the elaboration of guidelines we continue to derive implications for managers and provide guidance for future research.

Managerial Implications

Our theoretically driven framework and the generalizing guidelines should enable managers to take a better top-down approach to social media metrics and dashboards. Today, many organizations and agencies rather use metrics provided by the social network operators or other handily available ones. However, these may not be the most relevant to inform their marketing decisions. Our framework enables them to first assess what is important to know, and then look for the best proxies available. Even if proxies may not be available, frequent ad-hoc research, e.g. on user motives, may suffice for the time being.

In extension, the framework and its theoretical foundation will also help them modifying their marketing input. In contrast to classic advertising, which is usually not meant for

participation (i.e., mostly sharing of videos or simply collecting likes), they need develop new

forms of advertorial content that ignites users to engage, modify and then share it: they need to

learn to feed and nurture their base – a living organism. Another implication is that compared to

classic advertising media which can be off and on at the disposal of the brand managers, this

living organism needs constant feeding to survive. Or else, if your brand does not feed it, it turns

elsewhere for “food” or produces food on its own, whether you like it or not.

Another striking implication and major challenge for organizations is that user

participation will and should not stop with your brand communication. For social media and its egalitarian dialogs, organizations need the capability to listen, digest the information, and respond sensibly. As user dialogs also include logistics, product features and innovation, quality issues and the like, organizations need to reorganize around in-bound and out-bound interfaces (i.e.,

integrating all in- and out-bound communication channels in service hubs) with almost all internal functions over time. And as organizations integrate their communication interfaces, they will also feel the need for quick and consistent communication response to the plethora of users across all interfaces – in essence, they will sooner or later feel the need for a central content hub that serves all channels on all relevant topics in almost real-time.

Future Research

For academic researchers, our framework, literature review and guidelines set up several

areas for future research. First, the conceptual framework we offer is just a start; although it draws

on prominent theories from several research domains, however, we should use the opportunity of

social networks to search for an overarching theoretical foundation. Also, we need to explore what

other theories could add to our understanding of the phenomena in social networks. In particular,

we encourage further research on the social roles users assume and the types of social interactions

we observe, and how both of these link to actors’ motivations, the content, and the network

structures. With respect to the network structure, we find the theoretical models predicting the

resulting network structure as endogenous (based on prevalent motive structures) appealing. Most

research treats it as exogenous (see literature review in the Appendix). If network structure is in

constant flow and endogenous, that may open up territory for completely new dynamic approaches

to network modeling?

Second, we offer a general framework and generalizing guidelines on how to construct sensible metrics and subsequently dashboards. As we do not examine the practical usefulness of specific social media metrics, experience tells us that many companies have a bottom-up, data- driven process to collecting and employing such metrics. For instance, small and medium sized enterprises often take at face value the metrics offered for free in, e.g., Facebook Insights or Google Analytics (Wiesel et al. 2011). We encourage research on the effectiveness and efficiency of this convenience or availability driven bottom-up approach. The results should be compared to a strategy based on our theoretically inspired top-down approach of first deciding what should be measured and next looking for the best empirical proxies (e.g., DeHaan et al. 2013).

Third, our literature review reveals many disjoint studies on selected and specific social media topics. We feel that we are barely scraping at the surface of potential knowledge on social media, which may also reveal a lot more insights for managing other media better. We also suggest that more holistic research covering multiple elements of the suggested framework is necessary to answer the tough question on social media in a few years: what have we really learned from all these studies and how many decision makers would have to change their minds or actions based on the generated substantial insights?

Finally, we would like to suggest further research on adequate organizational structures

and processes that guide organizations in their change process towards seamless dialog interfaces

with social media. Metrics and dashboards are a start, but how can they successfully implement

the organizational changes affecting all other aspects of marketing beyond brand communications?

Summary

Social media are becoming ever more ubiquitous and important for marketing purposes.

However, social media are substantially different from traditional or other online media due to the network structure and their egalitarian nature. As such, they require a distinct approach to

management. A prerequisite for managing social media is their effective measurement. Marketing or subsumed social media dashboards, a sensible collection of key performance metrics linking marketing input via metrics to (financial) outcomes, are the tool of choice – but how should organizations design their dashboard metrics for social media?

Due to the huge variety of (and still emerging new) social media and the specific needs of brands, there is no silver-bullet kind of metric or metric compilation that addresses all

requirements for all brands alike. However, due to the shared fundamentals of social media there are common threads that allow at least a unified approach to the construction of appropriate metrics and subsequently dashboards. To help organizations developing and employing such an appropriate compilation of metrics, we provide them with a tool kit consisting of three novel components: First, we theoretically derive and propose a holistic framework that covers the major elements of social media, drawing on theories from sociology, marketing, and psychology. We continue to support and detail these elements, namely motives, content, network structure, and social roles & interactions, with recent research studies. Second, based on our theoretical framework, the literature review, and our practical experiences, we suggest 9 generalizing guidelines that may prove valuable for designing appropriate social media metrics and

constructing sensible dashboards. Third, we derive managerial implications and suggest an agenda for future research. We hope that these contributions may provide a reasonable tool kit for

research and practice when analyzing, understanding, and managing social media.

Figures & Tables

Figure 1: A S-O-R Framework for Social Media Metrics

Figure 2: Theories, Framework, & Guidelines

Figure 3: Excerpt of KPIs from a Social Media Dashboard (Buzzrank 2012)

Figure 4: KloutScore and EdgeRank Metrics

References

Adjei, Mavis T., Stephanie M. Noble, and Charles H. Noble (2010), “The influence of C2C communications in online brand communities on customer purchase behavior,” Journal of the Academy of Marketing Science, 38, 5, 634–53.

Ailawadi, Kusum L., Donald R. Lehmann, and Scott A. Neslin (2003), “Revenue Premium as an Outcome Measure of Brand Equity,” Journal of Marketing, 67, 5, 1-17.

Alba, Joseph W., and J. Wesley Hutchinson (1987), “Dimensions of Consumer Expertise,” Journal of Consumer Research, 13, 2, 411–54.

Alba, Joseph, John Lynch, Bart Weitz, Chris Janiszewski, Richard Lutz, Alan Sawyer, and Stacy Wood (1997), “Interactive home shopping: consumer, retailer, and manufacturer incentives to participate in electronic marketplaces,” Journal of Marketing, 61, 3, 38–53.

Albuquerque, Paulo, Polykarpos Pavlidis, Udi Chatow, Kay-Yut Chen, and Zainab Jamal (2012),

“Evaluating Promotional Activities in an Online Two-Sided Market of User-Generated Content,”

Marketing Science, 31, 3, 406–32.

Algesheimer, René, Sharad Borle, Utpal M. Dholakia, and Siddarth S. Singh (2010), “The Impact of Customer Community Participation on Customer Behavior: An Empirical Investigation,”

Marketing Science, 29, 4, 756–69.

Ambler, Tim (2003), Marketing and the Bottom Line, London: FT Press.

Ambler, Tim, and John Roberts (2006), Beware the Silver Metric: Marketing Performance Measurement Has to Be Multidimensional, Marketing Science Institute , Report #06-113.

Ansari, Ansim, Oded Koenigsberg, and Florian Stahl (2011), “Modeling Multiple Relationships in Social Networks,” Journal of Marketing Research, 48, 4, 713–28.

Aral, Sinan, and Dylan Walker (2011), “Creating Social Contagion Through Viral Product Design:

A Randomized Trial of Peer Influence in Networks,” Management Science, 57, 9, 1623–39.

Arora, Raj (1982), “Validation of an S-O-R Model for Situation, Enduring, and Response Components of Involvement,” Journal of Marketing Research, 19, 6, 505–16.

Ballester, Coralio, Antoni Calvó-Armengol, and Yves Zenou (2006), “Who's who in networks.

Wanted: The key player,” Econometrica, 74, 5, 1403–17.

Bandura, Albert (1971), Social Learning Theory, New York: General Learning Press.

Batra, Rajeev, and Michael Ray (1985), “How Advertising Works at Contact,” in: Psychological Processes and Advertising Effects, Linda F. Alwitt and Andrew A. Mitchell, eds., Hillsdale, NJ:

Lawrence Erlbaum Associates, 13–44.

Belk, Russell W. (1975), “Situational Variables and Consumer Behavior,” Journal of Consumer Research, 2, 3, 157–64.

Berger, Jonah, and Katherine L. Milkman (2012), “What Makes Online Content Viral?” Journal of Marketing Research, 49, 2, 192–205.

Berger, Jonah, Alan T. Sorensen, and Scott J. Rasmussen (2010), “Positive Effects of Negative Publicity: When Negative Reviews Increase Sales,” Marketing Science, 29, 5, 815–27.

Blau, Peter M. (1974), “Presidential Address: Parameters of Social Structure,” American Sociological Review, 39, 5, 615–35.

Blau, Peter M. (1977), “A Macrosociological Theory of Social Structure,” American Journal of Sociology, 83, 1, 26–54.

Burt, Ronald S. (1980), “Models of Network Structure,” Annual Review of Sociology, 6, 79–141.

Burt, Ronald S. (2005), Brokerage and Closure: An Introduction to Social Capital, Oxford, New York: Oxford University Press.

Buzzrank (2012), Facebook-Report, Hamburg (www.buzzrank.de).

Campbell, Colin, Leyland F. Pitt, Michael Parent, and Pierre Berthon (2011), “Tracking Back- Talk in Consumer-Generated Advertising: An Analysis of Two Interpretative Approaches,”

Journal of Advertising Research, 51, 1, 224–38.

Celen, Bogachan, Shachar Kariv, and Andrew Schotter (2010), “An Experimental Test of Advice and Social Learning,” Management Science, 56, 10, 1687–1701.

Centola, Damon (2010), “he Spread of Behavior in an Online Social Network Experiment,”

Science, 329, 5996, 1194–7.

Centola, Damon (2011), “An Experimental Study of Homophily in the Adoption of Health Behavior,” Science, 334, 6060, 1269–72.

Chen, Yubo, Scott Fay, and Qi Wang (2011), “The Role of Marketing in Social Media: How Online Consumer Reviews Evolve,” Journal of Interactive Marketing, 25, 2, 85–94.

Chen,Yubo, Qi Wang, and Jinhong Xie (2011), “Online Social Interactions: A Natural Experiment on Word of Mouth Versus Observational Learning,” Journal of Marketing Research, 48, 2, 238–

54.

Chintagunta, Pradeep K., Shyam Gopinath, and Sriram Venkataraman (2010), “The Effects of

Online User Reviews on Movie Box Office Performance: Accounting for Sequential Rollout and

Aggregation Across Local Markets,” Marketing Science, 29, 5, 944–57.