Toward a Sustainable Email Marketing Infrastructure:
A System Dynamics Perspective
Oleg Pavlov1
Nigel Melville
Robert Plice
1
Corresponding author.
Oleg V. Pavlov is Assistant Professor of Economics and System Dynamics, Worcester
Polytechnic Institute, USA, opavlov@wpi.edu. Nigel Melville is Assistant Professor of Business
Information Technology, Stephen M. Ross School of Business, University of Michigan, Ann Arbor,
npmelv@umich.edu. Robert Plice is Assistant Professor of Information and Decision Systems, San Diego
State University, rplice@mail.sdsu.edu. The authors acknowledge the helpful comments and suggestions of
participants in the 13th International Conference on Computing in Economics and Finance, Montreal,
Canada, June 14-16, 2007 and 25th International Conference of the System Dynamics Society, Boston,
MA, July 29- August 2, 2007
2
Abstract
Email marketing is a legitimate, lucrative, and widely used business tool, but it is
in danger of being overrun by unwanted commercial email (also known as spam).
Conventional approaches to maintaining the robustness of legitimate email attack pieces
of the problem. In contrast, this article asserts that the email marketing infrastructure is a
complex system requiring holistic analysis. A system dynamics model is developed to
identify the underlying dynamics and enable an examination of alternative mitigation
strategies. Results reveal that the system conforms with the limits-to-growth generic
structure, filtering may have the unintended consequence of increasing the amount of
unwanted email, and the unexpected increase comes about because better filters can
actually assist spammers by mitigating an information deficit.
Keywords: Email, marketing, spam, system dynamics
3
1. Introduction
The Internet offers a new paradigm for marketing, engendering a shift from
product to customer focus that includes micro-level customization and customer
relationship management (Rust and Espinoza 2006). An example is search-based
advertising, which grew from virtually nothing in the early 1990s to a $14 billion
industry in 2006 (Elkin 2007). Another example is email marketing, which provides
twice the return on investment (ROI) relative to other forms of online marketing: $57.25
for each dollar spent versus $22.52 (Direct Marketing Association). Email marketing is
now employed by 70% of all retailers and growing 10% annually (McCloskey 2006).
Ironically, email marketing may be its own worst enemy. The Internet drastically
lowers communication costs – an underlying driver of high email marketing ROI relative
to other channels. In turn, low production costs spur greater production, inducing entry to
the industry by legitimate and not-so-legitimate marketers, further increasing the volume
of email messages sent. As a result, consumers are awash in a sea of ads and information,
some useful and some not. Knowledge workers sift through hundreds of such emails per
week, more than one-half of which are spam, leading to information overload – a
situation when more information is not better (Schwartz 2004; Simon 1971). The bottom
line is that useful email marketing messages are lost in the background noise with
negative consequences for short-term ROI and long-term industry health.
The spiraling situation of email marketing overload and its deleterious impact on
ROI has been acknowledged by the industry, for example: “I recommend we all keep our
ears to the ground on this and work through organizations like the Email Sender and
Provider Coalition to help find balanced solutions that stop spam, but allow businesses to
continue to use the email medium for legitimate, permission-based communications.”
(Nussey 2007). But, is the email-marketing industry sufficiently motivated to work
jointly to make the necessary changes to ensure industry health? And, if not, how long
will a focus just on creative content and timing of delivery continue to provide a
consistent return on investment for marketing organizations?
The fundamental thesis of this article is that the email marketing infrastructure is
a complex adaptive system whose behavior is best understood by examining the system
as a whole, rather than isolated components (Kofman and Senge 1993; Woodside 2006).
The limitations of applying reductionist thinking to social, technical, and biological
systems is well documented (Senge 1991), and exacerbated by inherent human blind
spots to the effects of dynamics and feedback (Moxnes 2000). As explicated below, such
limitations have impeded attempts to stem the tide of spam and maintain a robust email
marketing infrastructure.
Consistent with its systems nature, email marketing is analyzed in this article
using system dynamics (SD), a method that embodies the logic of feedbacks and
causality and employs mathematical equations to simulate dynamic behavior. Given their
complexity, the equations are solved numerically and results presented in graphical form,
typically of a given variable, such as volume of sent spam, versus time. The result is a
better understanding of the underlying drivers of the behavior of the email ecosystem:
“Causal processes, causal interactions, and causal laws provide the mechanisms by which
the world works; to understand why certain things happen, we need to see how they are
produced by these mechanisms.” (Salmon 1984, p. 132). The method is thus consistent
4
with the twin objectives of raising awareness of the hidden dynamics that impact industry
health and examining the effectiveness of various mitigation strategies.
The system dynamics analysis undertaken in this article reveals several
mechanisms that explain key phenomena associated with the email marketing
infrastructure and that can inform the management of that infrastructure. The first
mechanism is that limited human attention acts as a limit to growth of the email
marketing system. The second mechanism is that filters have an unintended consequence.
That is, filters lead to fewer emails in user inboxes but much greater volumes of spam in
the email system as a whole. Finally, the targeting mechanism explains why this
unintended consequence comes about: filters can be viewed as a proxy for an information
resource that spammers lack, enabling them to achieve the effect of message targeting.
The next section includes a review of prior research examining email marketing
and spam, revealing that most approaches involve a focus on one particular aspect of the
email marketing system, rather than analysis of the system as a whole. It also provides an
overview of system dynamics methodology. Then, an analysis of the behavior of
developed system dynamics models reveals causal mechanisms explaining underlying
dynamics. Concluding remarks discuss implications for marketing professionals and
policy markers.
2. Approaches to Analyzing Email Marketing
Prior Research
Prior research on email marketing and spam can broadly be classified into two
groups. The first includes studies that focus specifically at reducing spam, from a broad
range of perspectives. The second includes studies from the marketing literature that
examine more broadly relevant phenomena associated with email marketing.
Most studies of email marketing and spam reduce the problem to one or two core
components, analyze those components, and make recommendations based on that
analysis. A frequent topic has been the development of more efficient algorithms for
distinguishing regular email from spam and filtering it (Fawcett 2003; Gray and Haahr
2004). Filtering techniques are one of the most popular topics of the two major
conferences on spam, the Conference on Email and Anti-Spam (CEAS) and the MIT
Spam Conference. To enhance filtering, email providers maintain lists of computers from
which they do not accept any email, the so-called blackhole lists (Goodman and
Rounthwaite 2004). There is also a movement to augment filters by maintaining lists of
“good” senders – email from any computer that is not on the list will be rejected on the
premise that it is spam (Goodman and Rounthwaite 2004).
The implicit logic of filtering approaches is that better filters block more spam,
lower spammer profit, and lead to less spam. In other words, better filters mitigate the
problem. However, despite apparent improvement in filtering techniques, spam only
appears to be growing in volume (Garretson 2007).
Another vigorous area of research examines the regulatory environment and the
impact of legislation on email marketing and spam (Goldman 2006; Gratton 2004). Here
the thinking is that laws can constrain the type of email marketing that will be sent, and
so can be designed to allow “good” email marketing and eliminate “bad” email
marketing. But, the CAN-SPAM Act of 2003 in the U.S. does not so far appear to have
5
stopped the flood of spam (although a number of legal cases have been brought,
potentially resulting in a signaling mechanism to less-than-legitimate operators).
Beyond filtering and legal mechanisms, other approaches include reducing email
overload by aiding recipients with information processing, specifically, presorting
arriving email (Croson et al. 2005). Similarly, another proposal involves innovative use
of a filter as part of a dynamic pricing mechanism, where price is determined
proportionally to the filter score assigned to the message by a filter (Dai and Li 2004).
Increasing the cost of sending email may alter the economics of spam toward enhanced
overall efficiency (Goodman and Rounthwaite 2004). In particular, email service
providers may impose a reverse Turing test (to test that the sender is human) and require
that computers of spammers solve difficult computations in order to increase the cost of
sending spam.
Researchers have also employed welfare economics to examine the distribution of
surplus between sender and receiver based on whether a message has value to a given
receiver (Loder et al. 2006). One analysis uses game theoretic approaches to analyze the
two-agent non-cooperative game between sender and receiver, finding a unique Nash
equilibrium and enabling prediction of the optimal point in the tradeoff between Type I
and Type II filter errors (Androutsopoulos et al. 2005). Another study demonstrates the
potential of systems thinking by developing an exploratory model of email
communication, suggesting that this may be a very fruitful approach (Pavlov et al. 2005).
Several marketing research studies concentrate on predicting response rates, such
as for catalog mailing (Basu et al. 1995). Recent studies have looked specifically at email
communication. For example, a model of online clicking behavior attempts to predict and
improve response rates for email communications (Ansari and Mela 2003). In this study,
the focus is on learning how to use individual preferences to custom-design marketing
emails, including aspects of content inclusion and presentation design. The model
accounts for the heterogeneity in preferences and the existence of some unobservable
variables. Customization is found to be very desirable, but not easily achievable. There
are also companies that specialize in email customization for commercial clients,
including Yesmail and DoubleClick. Overall, marketing studies tend to be narrowly
focused on manipulating specific, short-term behaviors with the goal of improving return
on investment.
In sum, extant research has improved the effectiveness of filters, even though
their deployment has not seemed to stem the flow of spam. Prior research has also helped
inform regulatory mechanisms, given rise to proposals for new methods of spam
mitigation, and demonstrated the viability and usefulness of systems thinking applied to
the spam problem. To complement and extend existing studies, we model email
marketing as a social system and examine how the system behaves under various
conditions, in order to reveal underlying causal mechanisms. Achieving such insights
requires that we take a systems approach, viewing the phenomenon as a whole rather than
in terms of its isolated components. The analytical tools of system dynamics are
particularly appropriate for such an investigation.
System Dynamics Methodology
System dynamics provides a holistic approach to the analysis of systems (Sterman
2001). Though an event-oriented approach is common, such a view ignores feedback
6
effects, which are inevitably present. Systems thinking in the context of email marketing
motivates the development of a system dynamics model representation of the system.
Developing a system dynamics model of email marketing involves a series of five
steps (Randers 1980). First is defining the boundaries of the system by drawing on
observation, prior research, and similar sources. The email marketing system includes
senders of email (broadcasters and sponsors), receivers of email, and technology vendors
that sell products intended to automate the decision problem faced by receivers of email
(McWilliams 2004). The second step is defining key variables of the system, which
include volumes of various types of messages sent, the stock of messages received, the
processing capacity of the receiver, the revenue paid to sponsors of messages, and filter
quality (misspecification rates, such as false positives and false negatives). Third,
reference modes of the system are explicated, that is, how do individuals and
organizations think and react within the system over time. For example, what actions do
recipients take when their daily stock of attention is exhausted? The fourth step prior to
building the model is to embody relationships among variables in causal diagrams that
may involve causal links and feedback loops. Based on these preliminary steps, the
simulation model is developed and validated in the fifth step, and policy
recommendations are suggested based on dynamic behavior of the system. Given the
focus of this article on raising awareness of the hidden dynamics that impact the
robustness of the lucrative email marketing channel as well as examining the
effectiveness of various mitigation strategies, results of SD modeling are discussed in
terms of underlying causal mechanisms (Hedstrom and Swedberg 1998).
3. Causal Mechanisms in Email Marketing System
Systems thinking and the model development steps outlined above motivate the
construction of a system dynamics model of email marketing. Because the model takes a
causal approach to examining system behavior, the terminology of causal mechanisms is
employed to discuss and systematize model findings (Hedstrom and Swedberg 1998).
A causal mechanism is what lies within the black box of causality, that is, the
“Nuts and bolts, cogs and wheels – that can be used to explain quite complex social
phenomena.” (Elster 1989, p. 3). As an example, threshold-based behavior is a causal
mechanism involving individuals that decide whether to act in a certain way based on
how many others are doing likewise – such as dining in a restaurant based on the number
of others dining in the restaurant (Granovetter 1978). Another example is the selffulfilling prophecy, which describes the process underlying bank runs: a) a rumor of bank
insolvency occurs; b) some depositors withdraw assets; c) rumor is reinforced by b) via
substantive and perceptive means; d) further withdrawal; e) trust in bank and solvency
reduced; f) bankruptcy (Merton and Rossi 1968). Similar to threshold-based behavior and
bank runs, system dynamics analysis of the email marketing system reveals a causal
mechanism that lies at the heart of system behavior. To emphasize, the analysis allows
not only the identification of each mechanism (such as x causes y) but also the
identification of the process (how x causes y) (Brady 2002). We now explain the causal
mechanism, the baseline dynamic behavior it generates, and two effects that emerge from
the analysis: unintended consequence of filtering and filtering as targeting.
7
Attention as a Limiting Condition
Conventional research on email marketing and spam examines individual
variables while acknowledging some simple causality between the variables. The email
marketing context is viewed as another advertising medium with differences in key
characteristics relative to print media: it is cheaper, information intensity is higher,
personalization is easier, and so forth. The causal logic is that more and better volumes of
email marketing lead to desired consumer actions. This view of email marketing is overly
simplistic.
Email is part of a complex social system that is stimulated by economic forces
and constrained by cognitive limitations of the receiving side. A schematic view of the
situation is in Figure 1. The diagram shows some key variables, and arrows indicate the
causality between the variables. Positive arrows imply a positive relationship between
variables and negative arrows show that variables move in the opposite directions.
Thinking in terms of systems reveals two primary feedback effects: the reinforcing loop
leading to a wealth of information and the balancing loop leading to a poverty of
attention. These loops are marked by letters R (for reinforcing) and B (for balancing)
respectively. While the reinforcing loop comprises the supply of spam, the balancing loop
acts to quell the growth of messages.
In the email infrastructure, the growth process of spam volume is driven by
message broadcasters and their sponsors. Suppliers of commercial email messages –
whether legitimate newsletters from a consumer’s bank, legitimate but unsolicited
messages from online shoe stores, or illegitimate messages intended to defraud – produce
and send messages to email recipients, thus increasing the message backlog in individual
inboxes (hence a positive arrow between spam production and message backlog). The
messages can be either left in the mailbox for some duration of time, processed, or
deleted. Recipients spend some time daily reading messages, which is the daily time
endowment dedicated to processing messages. The number of valuable messages that are
processed daily depends on whether or not there are messages waiting to be processed in
the backlog, the daily time endowment and a share of valuable messages in the backlog.
Each of these three determinants has a positive effect on the number of valuable
messages processed. If a message is valuable, it generates a sale and the sponsor receives
some revenue. Assuming that spam funding is positively related to the revenue generated
by spam, we connect revenue and spam funding with a positive arrow. More spam
funding means more money on spam production, and, ultimately, the number of
messages in the backlog. We just completed the wealth of information circle, which
comprises the supply of spam.
When information load is greater than the processing capacity of an individual,
information overload occurs. In the email context, processing capacity may be viewed as
an attention endowment that is measured in units of time. Prior research suggests that
overload manifests itself in situations when the time required to complete a task is greater
than the time budget (Schultz and Vandenbosch 1998). When recipients receive excessive
amounts of email they either ignore it, adding to the backlog, or delete it. In other words,
the greater the overload of messages, the more messages will be deleted. This completes
the balancing circle called the poverty of attention.
To understand the dynamics of a system depicted in Figure 1, we build a
corresponding computational model. A detailed description of the model is given in the
8
Appendix. In the following sections we describe the baseline experiment and some
insights.
spam
production
deletion
+
spam
funding
R
+
revenue
+
message
bac klog
wealth of
information
+
+
valuable
spam
messages
proc essed
+
B
poverty of
attention
+
information
overload
-
+
+
daily time
endowment
fraction of valuable spam
messages in bac klog
Figure 1: Causal relationships in the spam system
Baseline Behavior
To emulate the growth of spam (Figure 2), we start the simulation at low levels of
spam funding (line 2) and correspondingly, spam production (line 3) – they are set to near
zero levels. Message backlog (line 4) is also relatively low. Email recipients have enough
time to process messages and hence experience relatively low levels of information load
(line 5) due to email. Correspondingly, hardly any email is deleted (line 6) without being
processed. Gradually, the revenue from spam encourages more spam production. Among
the growing inflow of spam that arrives into recipients’ inboxes, some of the messages
are deemed valuable by the recipients and are being processed (line 1). If attention of
recipients were not limited, the reinforcing feedback effect, marked with an R in Figure
1, would create an exponential growth in spam production. However, spam does not
increase indefinitely. Around the time 80, information load (line 5) starts to increase
rapidly and the forces of the balancing loop become significant. The revenue and spam
production stagnate. The initial dominance of the reinforcing feedback effect is reduced
9
by the effect of the balancing loop leading to the distinctive S-shaped variable trajectories
in Figure 2.
growth
200
500,000
40,000
10,000
1
2,000
6
6
6
4
4
4
4
6
4
100
250,000
20,000
5,000
0.5
1,000
0
0
0
0
0
0
5
5
5
5
5
3
3
3
1
1
1
1
1
6 1
3
5
2
2
2
2
3
2
1
2
3
5
4
5
4
5
4
5
4
5
5
5
6 1 2 3
10
6 1 2 3
6 1 2 3
20
valuable messages processed
2
spam funding
3
spam production
4
message backlog
5
informaiton overload
6
6
deletion
30
6
1
2
5
6
1
5
5
6
5
6
5
6
5
6
2
3
4
3
4
5
6
150
1
2
3
4
140
1
2
3
4
130
1
2
3
4
120
1
2
3
4
110
1
2
3
4
100
1
2
3
6
90
1
2
5
6
6
6
70
80
Time (Day)
4
5
6
6
60
3
4
5
2 3
2
3
4
2
1
2
3
4
6
50
1
2
3
1 2 3
40
2
3
1
1 2 3
0
4
1
4
4
4
4
5
6
Figure 2: Transitory and Steady State Behavior of Email Marketing System
The convergence of the system to a steady state may seem unreasonable in light
of the global growth of spam volume. But Melville et al. (2006) find that the spam flow
to an educational institution, which has a fixed number of email inboxes, is rather steady
during an academic year.
Unintended Consequences of Filtering
Available empirical evidence suggests two apparently contradictory facts. First,
the amount of spam arriving in email inboxes appears to be diminishing, at least as
viewed from the perspective of individuals. Second, the amount of aggregate spam as a
percentage of all email messages appears to be increasing with time. A view at email as a
system reconciles these observed outcomes.
Research on email marketing and spam has focused on the development of evermore effective spam filters. In theory, such mechanisms work well at mitigating the
information overload by accurately detecting a message with text that indicates that it is
likely to be unwanted. There is thus a logic embedded in the decision to deploy filters as
a solution, and to the expectation that better filters will lead to less spam. However,
viewed from a systems perspective, there are unintended outcomes of filtering, as we
demonstrate with the help of the following simulation. The results of the simulation are
presented in Figure 3.
To avoid the transitory dynamics, we start the simulation in the steady state. We
allow for the filtering technology to improve beginning at time 30. In the industry jargon,
6
10
the misspecification rate (Cormack, 2006) of the filter is gradually reduced (line 1).
Since the filter identifies and discards many spam messages that are clearly of no interest
to recipients, the average fraction of valuable spam messages (line 2) in the recipients’
inboxes is improving. The daily attention endowment does not change, but the share of
valuable messages in the backlog is improving (line 2), and therefore the aggregate
amount of valuable messages processed (line 3) is be increasing (confirm the causal logic
in Figure 1). More valuable messages processed leads to more revenue (line 4). Because
of the increasing revenue, spam funding is also increased, which contributes to spam
production (line 5). The filter only temporarily reduces the information overload (line 6),
which reaches the old level due to the increased spam production.
In summary, spam production is higher in the presence of better filtering, which is
an unintended consequence of filtering.
filter
0.2
0.08
180
430,000
40,000
1.1
2
2
2
2
2
4
4
5
4
3
3
4
5
4 5
4 5
5
4 5
4 5
4 5
5
2 3 4
1 23 4 5 6 1 2 3 45 6 1 2 3 4 5
3
3
3
3
3
3
2
3
0.1
0.056
112.5
225,000
20,500
0.9125
2
2
2
2
5
6
6
6
6
6
6
6
6
1
6
6
6
1
0
0.032
45
20,000
1,000
0.7250
1
0
10
20
30
1
misspecification rate
av fraction of valuable spam messages
valuable messages processed 3
4
4
sponsor revenue
5
5
spam production
6
6
information overload
40
1
50
1
2
5
6
100
6
6
2
3
4
5
150
1
3
4
5
140
2
3
4
1
1
2
3
5
130
1
2
4
1
120
1
3
6
1
110
2
5
6
1
1
4
5
6
90
3
4
5
1
2
3
4
1
1
2
3
5
6
1
2
4
5
6
1
3
4
1
70
80
Time (Day)
2
3
4
60
1
2
3
1
4
5
6
Figure 3: Unintended Consequences of Filtering
The targeting effect of filters
The outcome that filtering contributes to the growth of spam may be puzzling at first. But
this result can be better understood if one realizes that filtering mitigates the problem of
asymmetric information. An information asymmetry occurs when one party participating
in an exchange knows something that the other party does not (Varian, 1992). A classical
example of a situation with information asymmetries is a used car market (Akerlof, 1970;
Jensen, 2007). In the case of email, the information asymmetry is due to the fact that
spammers do not know the preferences of recipients.
Targeting improves response rates of a marketing campaign (Chittenden and
Rettie, 2003; Ansari and Mela, 2003). Message targeting involves choosing message
content, format, and timing so that it matches the particular needs and preferences of the
recipient to whom it is addressed. Spammers tend not to use targeting, which is not only
6
11
because targeting is costly, but also because spammers – more so than legitimate email
marketers – have a limited ability to obtain information needed to customize their
messages to the preferences of each recipient. Instead, they rely on mass mailings rather
than customized messages.
We use our model to examine the impact of targeting. The improved targeting
starting at time 30 leads to the increase in the average fraction of valuable spam messages
in backlog (line 1 in Figure 4). This increase is qualitatively the same as in the case when
the filter was improved. As can be seen in Figure 4, trajectories for some key variables,
including sponsor revenue and spam production, are similar to the trajectories in Figure 3
for the case of filtering. Filtering and targeting have the same effect on the system
because they increase the fraction of valuable spam messages in the backlog (Figure 1).
Thus, by deploying the filter, the inbox owners have inadvertently enabled the same
outcome as if asymmetry of information was reduced and spammers were able to target
their messages.
targeting
0.08
180
430,000
50,000
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
0.0638
134.5
292,500
30,250
3
3
4
4
4
4
4
3
3
3
3
3
3
2
3
3
4
4
3
4
3
4
3
4
3
4
3
4
3
4
4
4
1
4
1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4
0.0476
89
155,000
10,500
0
10
20
30
av fraction of valuable spam messages
2
valuable messages processed
3
3
3
sponsor revenue
4
4
4
spam production
40
50
1
2
1
2
3
4
1
2
3
1
2
3
4
70
80
Time (Day)
1
2
3
4
60
3
4
1
2
1
2
3
4
90
4
1
2
3
1
2
3
4
110
1
2
3
4
100
4
1
2
3
1
2
3
4
140
1
2
3
4
130
1
2
3
4
120
4
1
2
3
150
2
3
4
Figure 4: Targeting and Filtering: Two Sides of a Coin
4. Summary and Recommendations
Despite the lucrative nature of email marketing, consumers are overloaded with
information, including both wanted and unwanted communication. However, the overall
result of mitigation strategies addressing individual dimensions of the problem has led to
“record growth, and record irritation,” with spam accounting for 88 percent of all e-mail
traffic in August of 2007 (Garretson 2007). The fundamental thesis of this article is that
the email marketing infrastructure is a complex system requiring analysis that takes
account of the whole system. A system dynamics model that captures the relationships in
3
4
4
12
the complex system is shown to identify underlying dynamics and reveal unexpected
consequences.
Thus far, the primary industry response to excessive levels of spam has been to
deploy filters. As the model developed herein explains, this has had an effect opposite to
that intended, and has given spammers the equivalent of a useful information tool
equivalent to targeting their messages.
A number of other approaches to mitigating information overload have been
proposed, including variants of email postage, attention bonds, and reverse Turing tests.
Before industry and regulators choose any of these approaches for general
implementation, it would be beneficial to turn to system dynamics for an analysis of the
likely outcomes. Unlike filtering, some of these proposals would change the nature of
email as a free, anonymous communications medium. The experience with filtering
suggests that before taking such drastic measures to stem the flow of spam, there should
be confidence that the resulting system has been thoroughly evaluated, and that the
behavioral and informational consequences of the proposed changes have been analyzed.
Such an analysis must take into account the dynamic feedback and causal-loop
mechanisms that govern the system. The model presented and analyzed in this paper
exemplifies the approach, and demonstrates the benefits of SD modeling to analyze the
email-marketing industry.
Appendix: System Dynamics Model
Below we present a mathematical formulation of the model, which enables
complete reproduction and simulation of the model. The model includes recipients of
email messages, a message sponsor, two types of message broadcasters and a technology
firm that provides a filtering product. A regular broadcaster (denoted by subscript A)
sends only regular email -- each regular message is valuable. Spammer (denoted by
subscript B) is financed by the sponsor and sends spam. Only a fraction of spam is
valuable. The model has been implemented in Vensim DSS and solved using the RungeKutta method.
Model
Parameters
number of recipients
average volume of regular
communication
time a typical employee spends on
email
normal response delay
base price of sending a spam
message
time it takes to process a message
spam budget duration
100 recipients
10 messages/period/recipient

fraction of valuable spam messages
0.01 dimensionless
d
average sensitivity to overload
3 dimensionless

fraction of surplus to funding
communication
0.1 dimensionless
R
aR
TR

pB

S
20 minute/(period *recipient)
4 periods
1 $/message
1 minute/message
10 period
13
*

tolerable misspecification rate
0.1 dimensionless
filter improvement delay
10 periods
r
marginal revenue per spam
message
initial percent of false-positives
2000 $/message
initial percent of false-negatives
0.12 dimensionless
1
2
0.01 dimensionless
Variables
volumes of arriving messages of
type A
messages that are identified as
valuable are added to the backlog
blocked messages
a  aR R
mi  1 1  a  b   2 1    b
mb  a  b  mi
message backlog
(d / dt ) M  mi  mo  md
budget of time
T  RTR
mo  min(T  , M )
processed messages
the daily time required to process
messages
rate of deletion
T *   M  
number of valuable messages in the
backlog
dV AB
 viAB  voAB
dt
AB
vi  1 1  a  b 
new valuable messages added to
the backlog
valuable messages that are either
processed or deleted
average fraction of valuable
messages of all types in the backlog
number of valuable spam messages
in the backlog
new valuable spam messages added
to the backlog
valuable spam messages that are
either processed or deleted
average fraction of valuable spam
messages in the backlog
number of spam messages in the
backlog
addition of new spam messages to
the backlog
removal of spam messages from the
backlog either through processing
or deletion
average fraction of spam messages
(valuable and non-valuable) in the
backlog
communication funding
md  eod  d  M
voAB   mo  md  

V AB
M
dV B
 viB  voB
dt
B
vi  1 1  b
voB   mo  md  

VB
M
dB
 bi  bo
dt
bi  1 1  b  2 1    b
bo   mo  md 
 
d
B
M
dt  S  i  e
14
expenditure on spam by sponsors
e  S S
sponsor revenue
RS  rmo 
sponsor cost
CS  e
allocation
i  max(  RS  CS  ,0)
flow of spam messages
b  e pb
d
misspecification rate
improvements to the effectiveness
of the filtering software a
percent of false-positives
percent of false-negatives
dt    i  o
o     *  
1  1 ,    
 2   2
Steady state values for
variables
Variable
Message
backlog
M
7293 messages
2412 messages
valuable messages in backlog
406 messages
valuable spam messages in backlog
spam
B messages in backlog
communication
funding
S
aggregate misspecification rate
5287 messages
202797 $
0.1 dimensionless
References
Androutsopoulos, I., Magirou, E., and Vassilakis, D. "A Game Theoretic Model
of Spam E-mailing," The 2nd Conference on Email and Anti-Spam (CEAS 2005), Palo
Alto, CA, 2005, pp. 1-8.
Aronson, D. "Overview of Systems Thinking," pp. 1-2.
Brady, H.E. "Models of Causal Inference: Going Beyond the Neyman-RubinHolland Theory," Annual Meetings of the Political Methodology Group, University of
Washington, Seattle, WA, 2002, pp. 1-108.
Elkin, T. "Jefferies: Online Ad Spending Could Exceed $60B By 2010," in:
Online Media Daily, 2007.
Elster, J. Nuts and Bolts for the Social Sciences Cambridge University Press,
Cambridge, 1989.
Fawcett, T. "“In vivo” spam filtering: A challenge problem for data mining,"
KDD Explorations (5:2) 2003.
Garretson, C. "The Summer of Spam: Record Growth, Record Irritation," in:
Network World, 2007.
15
Goldman, E. "A Coasean Analysis of Marketing," Wisconsin Law Review
(2006:4) 2006.
Granovetter, M. "Threshold Models of Collective Behavior," American Journal of
Sociology (83) 1978, pp 1420-1443.
Gratton, E. "Dealing with Unsolicited Commercial Emails: A Global
Perspective," Journal of Internet Law (7:12) 2004, pp 3-13.
Gray, A., and Haahr, M. "Personalised, Collaborative Spam Filtering," 2004.
Hedstrom, P., and Swedberg, R. Social Mechanisms: An Analytical Approach to
Social Theory Cambridge University Press, New York, NY, 1998.
Kofman, F., and Senge, P. "Communities of Commitment: The Heart of Learning
Organizations," Organizational Dynamics) 1993, pp 5-23.
Loder, T., Van Alstyne, M., and Wash, R. "An Economic Response to Unsolicited
Communication," Advances in Economic Analysis and Policy (6:1) 2006, pp 1-38.
McCloskey, W. "2006 Retail White Paper," Email Data Source, New York, pp. 115.
McWilliams, B. Spam Kings O'Reilly, UK, 2004.
Merton, R.K., and Rossi, A.S. "Contributions to the Theory of Reference Group
Behavior," in: Social Theory and Social Structure, Free Press, New York, 1968.
Moxnes, E. "Not Only the Tragedy of the Commons: Misperceptions of Feedback
and Policies for Sustainable Development," System Dynamics Review (16:4) 2000, pp
325-348.
Nussey, B. "Spam is on the Rise Again.," silverPOP, 2007.
Pavlov, O., Melville, N., and Plice, R. "Mitigating the Tragedy of the Digital
Commons: the Problem of Unsolicited Commercial E-mail," Communications of the AIS
(16), July 2005, pp 73-90.
Randers, J. "Guidelines for Model Conceptualization," in: Elements of the System
Dynamics Method, J. Randers (ed.), MIT Press, Cambridge, MA, 1980, pp. 117-138.
Rust, R.T., and Espinoza, F. "How Technology Advances Influence Business
Research and Marketing Strategy," Journal of Business Research (59) 2006, pp 10721078.
Salmon, W. Scientific Explanation and the Causal Structure of the World
Princeton University Press, Princeton, 1984.
Schwartz, B. The Paradox of Choice: Why More is Less HarperCollins, New
York, NY, 2004.
Senge, P. The Fifth Discipline: The Art and Practice of the Learning Organization
Currency Doubleday, 1991.
Simon, H. "Designing organizations for an information-rich world," in:
Computers, communications, and the public interest, The Johns Hopkins University
Press, Baltimore, MD, 1971, pp. 38-52.
Woodside, A. "Advancing Systems Thinking and Building Microworlds in
Business and Industrial Marketing," Journal of Business and Industrial Management
(21:1) 2006, pp 24-29.
Akerlof G A (1970). The market for "lemons": Quality uncertainty and the market
mechanism. Quarterly Journal of Economics 84: 488-500.
16
Ansari A and Mela C F (2003). E-customization. Journal of Marketing Research 40 (2):
131-146.
Chittenden L and Rettie R (2003). An evaluation of email marketing and factors affecting
response. Journal of Targeting, Measurement and Analysis for Marketing 11 (3): 203 –
218.
Cormack G V (2006). The trec 2005 spam filter evaluation track. Virus Bulletin
(January): S2.
JENSEN C (2007). Their titles laundered, the cars are still lemons. The New York Times
Online (August 26).
Melville N, Stevens A, Plice R K and Pavlov O V (2006). Unsolicited commercial email: Empirical analysis of a digital commons. International Journal of Electronic
Commerce 10 (4): 143-168.
Varian H (1992). Microeconomic analysis. W. W. Norton & Company: New York.