Page 1 of 7 ANZMAC 2009
A Conceptual Framework for Viral Marketing
David B. Stewart, Monash University, [Link]@[Link]
Michael T. Ewing, Monash University, [Link]@[Link]
Dineli R. Mather, Deakin University, [Link]@[Link]
Abstract
This research seeks to temper the enthusiasm surrounding viral marketing by claiming that
there is no such thing as a pure viral marketing campaign and that the successes of many
campaigns labeled ‘viral’ are due to many factors including traditional mass communication.
The article presents a conceptual framework for viral marketing, inspired by epidemic
modeling and the theory of branching processes, that enables a campaign to be separated into
its viral and non-viral components. We then introduce a mathematical model for the viral
process and develop three campaign performance metrics: overall campaign performance
(reach), the dynamic performance of the viral component (dynamic criticality) and
performance of the viral component relative to the non-viral component of the campaign
(relative viral reach).
ANZMAC 2009 Page 2 of 7
A Conceptual Framework for Viral Marketing
Introduction
Bampo et al. (2008) focused on the effects of the digital network on the performance of viral
marketing campaigns and demonstrated that the viral propagation process can be successfully
modeled. However, researchers have yet to examine the fundamentals of the viral marketing
process nor attempted to develop a mathematical model of the viral propagation mechanism.
This study therefore seeks to develop a set of analytical tools that can be used in setting
campaign objectives and budgets, to track campaign performance, to facilitate scenario testing
(before or during a campaign), to provide diagnostic feedback in the early stages of a
campaign and to guide intervention strategies and to be used in post-campaign evaluation.
This paper begins by introducing a conceptual framework that defines viral marketing and
describes the viral and non-viral components. We then posit a mathematical model to
characterise the viral marketing propagation process. Next, we present three metrics to
measure viral marketing campaign performance. Finally, we consider both managerial and
theory-building implications of our approach, and outline directions for future research.
Literature Review
The ‘discovery’ of viral marketing in the mid-1990s (Kaikati and Kaikati, 2004) captured the
imagination of the industry and subsequently generated an array of monikers (“word-of-web”,
“word-of-mouse”, “customer-to-customer” [C2C] or “peer-to-peer” [P2P] communication and
“buzz marketing”). The term ‘viral’ arose from medical and health sciences literature (e.g.
Mather, 2000), and is an apt metaphor for a marketing activity that encourages individuals to
propagate a marketing message within their network(s), and thereby exponentially increase
the potential reach of the message. Viral marketing is generally agreed to involve spreading a
marketing message via ‘word of mouse’ and ensuring that the receivers have the interest to
pass along the message to their acquaintances (Kaikati and Kaikati, 2004, p. 4).
Viral marketing campaigns have both advantages and disadvantages, which differentiate the
viral approach from other more conventional communication strategies. Compared to mass
communication, a significant advantage for viral marketing is the ‘self selection’ process
within the message, affecting how the message spreads. As there is a greater degree of
creative license afforded through the message delivery medium that is more intimate and
personalized, the message can be more targeted and this increases the likelihood of the
message effectively reaching interested individuals (particularly “hard-to-get” audience
members) and reducing the likelihood of the message being spread to uninterested
individuals. Whereas mass communication messages need to ensure that they are interesting,
yet not offensive to the broader community, viral messages can afford to be provocative,
sharper, edgier and even perhaps “darker”.
A digital viral process also increases the speed with which the message is disseminated and
maintains the integrity of the message. Additionally, a receiver of a message is more likely to
be amenable to the message because the sender of the message has implicitly endorsed the
message by passing it on (Chiu et al., 2007). Whereas consumers are quick to delete messages
from marketers, a message from a known person is less likely to be deleted (Phelps et al.,
Page 3 of 7 ANZMAC 2009
2004), even if the sender is someone whom sociologist Mark Granovetter (1973) might call a
‘weak tie’. The level of trust in the relationship between the sender and the receiver transfers
to the ‘advice’ that is being communicated in the form of the message (Keller, 2007). All of
these factors combine to ensure that viral marketing can be highly cost effective. Quantifiably
assessing the level of campaign success, through the behavioral responses, is also potentially
easier, particularly if there is an embedded call-to-action.
Despite all the inherent advantages associated with viral marketing, we contend that viral
campaigns are not necessarily easier or more effective than traditional media and that they do
in fact require a significant level of management. For example, messages with built-in
incentives, such as coupons or competitions, can become too successful – by spreading
beyond the anticipated pool of consumers, by continuing long after the incentive competition
has ended, or in cases where consumer demand exceeds the anticipated supply of the new
products or samples that a viral campaign might be promoting. Conversely, viral campaigns
can lose momentum, and ‘die’ before reaching a significant proportion of the intended
consumer audience. These are potentially instances where managers have not exercised
control of the campaign, due, possibly, to the lack of appropriate metrics and analytical
models. In a similar vein, viral campaigns that do not grow exponentially or die prematurely
may ‘bubble along’ but have very little effect within the marketplace. Branching theorists
(Harris, 1963; Athreya and Ney, 1972; Stewart, 1976; Bampo et al., 2008) have defined such
movement patterns as listed above as super-critical (viral explosion), sub-critical (viral death)
or critical (viral campaign maintained, but with little effect). By understanding the underlying
factors that contribute to, and direct the above processes, marketing managers will be better
able to manage their viral campaigns.
It is these top-down, market-initiated viral communications which are the subject of the
present study. While there are, of course, bottom-up processes (often based on user generated
content), word-of-mouth effects, and the synthesis between the two (Kirby, 2004), these
aspects are beyond the scope of the present paper. Recent papers have also focused on the
synergistic value in integrating online and offline marketing strategies (Graham and Havlena,
2007; Ewing, 2009). Instead, the focus of this paper is on firm-initiated marketing campaigns
with an engineered viral component; that is where a marketing message, once seeded, is self-
propagated by audience members.
A Conceptual Framework for Viral Marketing
Defining the Viral Marketing Process
There are direct parallels between the customer-to-customer propagation of a marketing
message with the transmission of a contagious disease through a population. As there is an
extant rich body of literature on epidemic theory focusing on the transmission of viruses
(Becker 1989; Fulford, Forrester, and Jones, 1997; Mather and Crofts, 1999) we begin by
exploring the concepts used in epidemic theory that can be applied to a viral marketing
context. We then introduce concepts that are specific to the digital viral marketing process.
The commonly used terminology in epidemic theory to describe the transitional states of
individuals as the contagion traverses the population of interest is ‘susceptible’, ‘infective’
and ‘immune’. Individuals are susceptible before receiving the virus, infective before passing
it on, after which they are immune. This terminology carries directly over to a viral marketing
context where a member of a target audience is initially susceptible (receptive) prior to
ANZMAC 2009 Page 4 of 7
receiving the marketing message, infective (reached) upon receipt of the message and immune
(removed) once they choose not play a further role in the transmission of the message or the
campaign is terminated or expires.
Considering the transmission of the message from person to person through the life of the
campaign, Bampo et al. (2008) used the concept of generations (rather than time), which is
widely used in epidemic modeling, as a more appropriate index in a viral marketing context.
For example, a lengthy campaign could be due to audience members taking a long time to
pass the message on, rather than the message being passed on many times (generations).
Thus, campaign duration is not necessarily an indicator of success. In viral contexts, the
notion of generation better captures the transmission of the message from one individual to
the next. Bampo et al. (2008) define generation by the number of transmissions required to
reach a member along a chain of communication initiated by a single seed.
Let us now consider the parameters used to model the propagation process in epidemic
theory, namely, the contagion parameter, the epidemic intensity and the epidemic threshold
parameter. The contagion parameter (p) denotes the probability that an infective will transmit
the virus. The epidemic intensity (λ) represents the average number of people reached, but not
necessarily ‘infected’, as some may have been previously infected and are therefore immune.
The product of these two parameters, pλ denotes the epidemic threshold parameter (ETP) and
this parameter captures the progression of the epidemic by focusing on the ‘transmission’
(sending) of the contagion. Some of the epidemiology literature draws from the theory of
branching processes (Becker, 1989) to describe the behavior of epidemics within a large
population by defining it to be sub-critical (ETP < 1), super-critical (ETP > 1) and critical
(ETP = 1).
Whilst there are many parallels between the modelling of epidemics and viral marketing there
are also a number of fundamental differences: In a viral marketing context, the managerial
objective is to facilitate the spread of the marketing message (contagion) among members of
the target audience. Further, the efficacy of a viral marketing campaign can be assessed
through the cost effectiveness of the campaign design and strategy that instigated the viral
propagation thus focusing on the “contracting” (receiving) of the contagion. In order to
develop a framework that encapsulates these differences, we now consider concepts that are
specific to the viral marketing process.
Seeding
By definition, seeding is a non-viral activity that can happen through traditional or digital
communication over the life of the campaign. If the seeding is carried out through a digital
channel and the viral transmission mode adopted for the campaign is also digital, the
campaign is designed to be purely digital. Whilst a campaign is always initiated by seeding,
this does not preclude subsequent seeding. Seeding is a part of the campaign strategy and can
be approached in two ways: first, selective seeding where the marketer explicitly identifies a
select subset of the target audience (seeds), in accord with campaign strategy and
communicates directly with that group in order to initiate the campaign; second, the marketer
provides a platform for audience members to self-select as seeds and hence initiate the viral
campaign. We refer to these two forms of seeding as marketer-selected seeding and self-
selected seeding. In marketer-selected seeding, the selection is more strategic and influenced
by marketing concepts such as ‘connectors’, ‘influentials’ and ‘market mavens’ (Gladwell,
2000). Self-selected seeding, whilst having the advantage of containing the campaign to the
Page 5 of 7 ANZMAC 2009
target audience, has an extra element of uncertainty and therefore is more reliant on the
platform having greater visibility in a traditional communication context.
A Mathematical Model
We now propose a mathematical model for the viral component of a campaign which captures
the essence of the viral marketing phenomena in the case of marketer-selected seeding used to
initiate the campaign only. This model is an adaptation of a deterministic model discussed by
Frauenthal (1980, pp. 35–7). A version of the model also appears in Fulford, Forrester, and
Jones (1997, pp. 164–9) and was used by Anderson and May (1982) in modeling measles
epidemics. The model is defined by the following recurrence equation
(1) π g +1 = min{µπ g , 1}∗ φ g , (g = 0, 1, 2, ...)
where the index denotes the number of generations following seeding, πg denotes the
proportion infected at the gth generation (so π0 is the proportion seeded at the zeroth
generation), µ = pλ denotes the ETP, and φg denotes the proportion susceptible at the gth
generation. It follows immediately that
g
(2) φg = 1 − ∑ π j .
j =0
The rationale for equation 1 is as follows: At generation g there are Vg infectives, each
generating µ transmissions, resulting in a total of µVg transmissions. If this meets or exceeds
the target population size (equivalently µ πg ≥ 1) then all remaining susceptibles will become
infectives, and the process will terminate after generation g+1. Otherwise, the number of
transmissions that result in infectives after generation g+1 should be (on average) proportional
to the number of susceptibles remaining in the target population at generation g.
Conceptually, there are two competing forces operating within the model. On the one hand,
we have the number of transmissions growing at a rate of µ per generation in the early stages
of the process (when the remaining fraction of susceptibles is large). But in the “success” of
this process lies its eventual demise, since the pool of susceptible members decreases in
response to the “success”. This tension in the model is captured by the notion of dynamic
criticality, which we now explain.
When µ πg < 1 for all g (the campaign never attains 100% reach) equation (1) can be rewritten:
π g +1
(3) = µφ g , ( g = 0,1, 2,K)
πg
The behavior of the viral process is therefore dependent on the right-hand side of equation (3),
which we term as the dynamic criticality, and denote by ψg.
The proportion infective {πg} will therefore grow provided ψg>1. A necessary condition for
this to occur is for the ETP, µ to exceed 1. When this occurs, we refer to the viral campaign as
being super-viral. Eventually though, a super-viral campaign will decline, once ψg<1 (see
below). In the case where ψ0<1 (equivalently π0 > 1-1/µ ), the number of infectives will
decline at each generation. When µ ≤ 1, the percentage infective at each generation rapidly
declines. When this occurs, we refer to the viral campaign as being sub-viral.
Now consider the case where ψ0>1 and 1< µ ≤ 2. Using Microsoft Excel it can be shown that
the percentage of infectives will grow at each generation until it reaches a maximum and then
ANZMAC 2009 Page 6 of 7
it will decay (once ψg<1). The resultant reach of the target market exhibits an S-shaped curve.
Eventually, the viral process {πg} decays at a geometric rate of order ψ = µ φ, where ψ and φ
are the corresponding limiting values in the series {ψg} and {φg} as g→∞.
Metrics for Measuring Campaign Performance
Based on our conceptualization of the viral marketing process, and the mathematical model,
we present three metrics for measuring campaign performance. First, the eventual reach of the
target market (given by 1−φ) which depends on both the ETP and the number of seeds. A
highly successful viral marketing campaign would be characterised by achieving a high level
of reach within the target market (Bampo et al., 2008). Second, in assessing the effectiveness
of the strategies adopted in the campaign it is useful to measure the relative contribution of
the viral and non-viral components. The Relative Viral Reach (RVR) is defined by
( 4) Relative Viral Reach (RVR) =
(1 − φ ) − π 0 ,
π0
where the numerator represents the eventual proportion receiving the message through non-
viral means, and the denominator is the non-viral proportion seeded. We advocate the use of
the RVR in campaign objective setting. For example, an RVR objective may be that a
campaign attains an RVR of 1.5 (i.e. the viral reach is 50% greater than non-viral reach).
Third, dynamic criticality is a relative performance measure. Beyond the initial value, ψ0 it is
entirely focused on the viral component of the campaign. The initial value is itself of interest
as it indicates the efficacy of the seeding in establishing the first generation of the viral
process. By monitoring dynamic criticality, the campaign manager can assess whether the
campaign is retaining its momentum or waning, and perhaps in need of an injection of funds.
Future Research
Further research should include applying the mathematical model and metrics presented here
to empirical viral marketing campaigns to demonstrate their utility to campaign managers. In
addition, variants of the mathematical model should be developed to broaden the areas of
application to self-selected seeding, and seeding that is dependent on campaign performance.
References
Anderson, R.M. and May, R.M., 1982. The logic of vaccination. New Scientist, 410–415.
Athreya, K.B. and Ney, P.E., 1972. Branching Processes, Springer-Verlag, Berlin.
Bampo, M., Ewing, M.T., Mather, D.R., Stewart, D.B., Wallace, M., 2008. The effects of the
social structure of digital networks on viral marketing performance. Information Systems
Research 19(3), 273-90.
Becker, N.G., 1989. Analysis of Infectious Disease Data, Chapman & Hall, London.
Chiu, H.-C., Yi-Ching, H., Ya-Hui, K., Lee, M., 2007. The determinants of email receivers'
disseminating behaviors on the internet. Journal of Advertising Research 47, 524-34.
Page 7 of 7 ANZMAC 2009
Ewing, M.T., 2009. Integrated marketing communications measurement and evaluation.
Journal of Marketing Communications 15(2-3), 103-117.
Frauenthal, J.C., 1980. Mathematical Modelling in Epidemiology, Springer-Verlag, New
York.
Fulford, G., Forrester, P., Jones, A., 1997. Modelling with Differential and Difference
Equations, Cambridge University Press, Cambridge.
Gladwell, M., 2000. The Tipping Point, Little, Brown, and Company, Boston.
Graham, J., Havlena, W., 2007. Finding the "missing link": advertising's impact on word of
mouth, web searches, and site visits. Journal of Advertising Research 47: 427-35.
Granovetter, M.S., 1973. The Strength of Weak Ties. American Journal of Sociology, 78,
1360–80.
Harris, T.E., 1963. The Theory of Branching Processes, Springer-Verlag, Berlin.
Kaikati, A., Kaikati, J., 2004. Stealth marketing: how to reach consumers surreptitiously.
California Management Review 46(4), 6–22.
Keller, E., 2007. Unleashing the power of word of mouth: creating brand advocacy to drive
growth. Journal of Advertising Research 47, 448-52.
Kirby, J., 2004. Online Viral Marketing: The strategic synthesis in peer-to-peer brand
marketing. Working Paper, Digital Media Communications.
Mather, D.R., 2000. A simulation model of the spread of hepatitis C within a closed cohort.
Journal of the Operational Research Society 51, 656–65.
Mather, D.R., Crofts, N., 1999. A computer model of the spread of hepatitis C virus among
injecting drug users. European Journal of Epidemiology 15, 5–10.
Phelps, J.E., Lewis, R., Mobilio, L., Perry, D., Raman, N., 2004. Viral marketing or electronic
word-of-mouth advertising: examining consumer responses and motivations to pass along
email. Journal of Advertising Research 44, 333–48.
Stewart, D.B., 1976. Branching processes with dependent immigration. Unpublished doctoral
dissertation. Melbourne: Monash University.