Inf Syst Front (2011) 13:637–653
DOI 10.1007/s10796-010-9242-4
You’ve got email! Does it really matter to process emails
now or later?
Ashish Gupta & Ramesh Sharda & Robert A. Greve
Published online: 3 June 2010
# Springer Science+Business Media, LLC 2010
Abstract Email consumes as much as a quarter of
knowledge workers’ time in organizations today. Almost a
necessity for communication, email does interrupt a worker’s
other main tasks and ultimately leads to information
overload. Though issues such as spam, email filtering and
archiving have received much attention from industry and
academia, the critical problem of the timing of email
processing has not been studied much. It is common for
many knowledge workers to check and respond to their
email almost continuously. Though some emails may require
very quick responses, checking emails almost continuously
may lead to interruptions in regular knowledge work.
Managing email processing can make a significant difference in an organization’s productivity. Previous research on
this topic suggests that perhaps the best way to minimize the
effect of interruptions is to process email frequently for
example, every 45 min. In this study, we focus on studying
email response timing approaches to optimize the communication times and yet reduce the interruptive effects. We
investigate previous recommendations by performing a twophase study involving rigorous simulation experiments.
A. Gupta (*)
School of Business, Minnesota State University Moorhead,
1104 7th Ave South,
Moorhead, MN 56563, USA
e-mail: gupta@mnstate.edu
R. Sharda
Department of Management Science and Information Systems,
Spears School of Business, Oklahoma State University,
Stillwater, OK 74078, USA
e-mail: ramesh.sharda@okstate.edu
R. A. Greve
Meinders School of Business, Oklahoma City University,
Oklahoma City, OK 73106, USA
e-mail: rgreve@okcu.edu
Models were developed for identifying efficient and
effective email processing policies by comparing various
ways to reduce interruptions for different types of knowledge workers. In contrast to earlier research findings, results
indicate that significant productivity improvements could
be achieved through the use of some email processing
policies while helping attain a balance between email
response time and task completion time. Findings also
suggest that the best policy may be to respond to email two
to four times a day instead of every 45 min or continuously,
as is common with many knowledge workers. We conclude
by presenting many research opportunities for analytical
and organizational IS researchers.
Keywords Email management . Interruption .
Performance . Simulation modeling
1 Introduction
Emails have become increasingly necessary for communicating and exchanging information as we have migrating
towards always-on and geographically dispersed but digitally connected workplaces. The benefits of using emails
and the associated productivity gains are well documented
in the literature. However, managers and researchers are
now beginning to see the flip side of excessive reliance on
emails. For example, Weber (2004) in his MIS Quarterly
editorial recognizes the need for better understanding of
problems associated with email. Still another editorial note
(Whittaker et al. 2005) calls for more research on preparing
email for new business realities.
One challenge, among several others, that knowledge
workers are facing is managing the high volume of emails
on a timely basis, in the most efficient and effective
638
manner. Due to increased volumes, workers are spending
more time on emails than they did in the past. A survey of
840 organizations reports that 47% of their workers spend
1–2 h, and 34% spend more than 2 h on any given workday
processing email (American Management Association
2004). This is causing several problems. First, it leads to
a perception of a shortage of time thereby resulting in
information overload (Dennings 1982; Markus 1994;
Berghal 1997; Jackson et al. 2003). Second, it leads to
bounded rationality as knowledge workers have limited
time and resources available to make decisions and
complete their tasks, since too many emails are vying for
the knowledge workers’ attention. Many policies aimed at
reducing email overload have been suggested. These
policies usually involve eliminating any unnecessary email
transmissions. This is typically accomplished with filtering
software or could take the form of policies limiting the use
of the carbon copy field. Implementing prioritization
schemes so that time sensitive messages receive their
proper attention is yet another strategy employed.
In this study we focus on policies that look at the timing
of email processing. Knowledge workers often use audible
and visual notifications such as messengers and have the
tendency to respond to messages as soon as emails arrive
(Jackson et al. 2003). This often results in the interruption
of ongoing tasks. Jackson’s study suggests that although the
time lost due to each email interruption may be small, the
cumulative effect can become sufficiently large given that
an organization is comprised of several knowledge workers,
each receiving dozens of emails in need of processing each
day. To cope up with this, Jackson et al. (2003) suggested
that knowledge workers limit the checking of email
messages to once every 45 min. The impact of email
interruptions is more likely to be felt by the receivers of
email, rather than the senders, because senders usually
originate new emails during their naturally occurring breaks
or at times that suit them.
One frequently touted benefit of email is its asynchronous nature or the ability to process messages at a
convenient time. We are often instead using it much like
any synchronous communication tool such as the telephone
or chat. By using email like a telephone, knowledge
workers lose the key benefit of email (its asynchronous
nature) and accept additional interruptions to other important activities. That being said, a tradeoff clearly exists
between interruptions and potentially slow responses.
The prime issue that we consider in this study is what
timing-based email processing policies enable knowledge
workers to effectively allocate their attention. We hope to
bring to light policies that mitigate the negative impact of
email interruptions (non-value added time resulting from
interruptions to workflow) without significantly compromising the time required to resolve email messages.
Inf Syst Front (2011) 13:637–653
Ideally we may think of knowledge workers as selflearning agents who will determine the ideal email
processing policy from their past experience but, as is
evident from several recently published reports and studies,
this does not seem to be the case. Perhaps most knowledge
workers act as irrational agents by processing emails as
soon as they arrive, in spite of being aware of the
interruptions! This is similar to the example of driving too
fast. Even though people are aware that driving fast is
irrational and dangerous. However, they still drive fast.
Controlled driving can help reach the destination safely but
with a tradeoff in terms of time. By selecting a specific time
frame or frames dedicated to email, and not answering
email outside of this time frame, can we control email’s
interruptive nature, without compromising on our ability to
resolve email messages in a timely manner?
The argument here is not that knowledge workers should
refrain from processing emails as they lead to interruptions.
Another argument is that knowledge workers do not feel
they are being interrupted by emails, since emails are very
much a part of their regular tasks and that much of the work
is facilitated through emails. The issue that we are trying to
address in this study is the following. Irrespective of how
well emails get woven into the fabric of our always-on,
geo-dispersed work environment, knowledge workers still
have to switch between tasks and emails and therefore
spend unproductive time resuming interrupted tasks once
email processing is complete. We are trying to consider
possible solutions to reduce this time loss, given that
knowledge workers do not have much control over the
number of incoming emails.
An important caveat here is that we are not looking into
different polychronic behaviors exhibited by knowledge
worker. Highly polychronic knowledge workers see interruptions as positive influence and their performance improves
as a result of interruptions in contrast with workers exhibiting
low polychronicity (Hall and Hall 1990). The non-value
added time that we aim to minimize usually remains
constant for each email interruption across all knowledge
workers irrespective of whether they are highly polychronic
or not. Though it would be interesting to investigate the
performance differences demonstrated by knowledge workers having different polychronic inclinations, it is not the
goal of this study. Also, we restrict the scope of this study to
those knowledge workers who spend significant amounts of
time processing their emails. Specifically, we explore the
following research questions in this study:
Research question 1: Are there email processing
policies that will enable a balance between email
response time and task completion time?
Research question 2: Will fewer interruptions result in
significantly more efficient work completion?
Inf Syst Front (2011) 13:637–653
Research question 3: Will fewer interruptions significantly lower the numbers of hours worked daily?
Research Question 4: To what extent will achieving a
fit between email arrival patterns and email processing
patterns influence the success of given email processing policies?
We adopt a computational modeling approach and
simulate the work environment of a knowledge worker
and compare different timing-based processing policies.
The following section provides a literature review. The 3rd
section describes the research questions and hypotheses.
The 4th section provides the model details with an
accompanying appendix for technical details. The 5th and
6th sections summarize the results. Finally, the 7th section
provides discussion, limitations and future work.
2 Literature review
Interruptions have been defined in several ways. Jett and
George (2003), in their literature review on interruptions
provide a generic definition. Interruptions are defined as
incidents or occurrences that impede or delay organizational members’ progress on work tasks. They propose four
major types of interruptions: (1) intrusion, (2) break, (3)
distraction and (4) discrepancy. An intrusion is normally
viewed from a time management perspective and is defined
as an unexpected and unscheduled encounter that interrupts
the flow and continuity of an individual’s work, thus
bringing that work to a temporary halt (Jett and George
2003). It appears that the above description could describe
email’s intrusive effect on knowledge workers. Another
definition from the theory of distraction describes an
interruption as “an externally generated, randomly occurring, discrete event that breaks the continuity of cognitive
focus on a primary task” (Corragio 1990).
Although some research on interruptions has been done
within disciplines including IS, human-computer interaction (HCI), management, and cognitive psychology, we still
have more to learn. Much of the related research has
focused on either interface design or the impact of
interruptions on task performance. For example, research
within the field of human-computer interaction (HCI) has
been mainly focusing on developing interfaces to reduce
interruptions and cognitive overload and (e.g., McFarlane
2002). Authors who have studied the interruptive nature of
technologies such as email and instant messaging on
primary task performance suggest that the intensity of the
interruption effect depends upon the point at which the
primary task gets interrupted (e.g., Cutrell et al. 2000;
Czerwinski et al. 2000; Speier et al. 1999, 2003).
Interruptions had less of an impact when a task was
639
interrupted during an earlier processing stage. For example,
a task that is interrupted during its planning phase will have
a smaller penalty attached to it than a task that is interrupted
during later stages, typically called the execution and
evaluation phases (Czerwinski et al. 2000). Recently,
research in HCI has also studied the problems of email.
An entire special issue of HCI Journal focuses on
redesigning and reinventing email (Whittaker et al. 2005).
Duchenaut and Watts (2005) propose taxonomy for future
work on email systems. Other efforts are focusing on
developing and designing emails for the 21st century (Kerr
and Wilcox 2004). None of these studies have looked at the
time based policies that could be used to mitigate the
impact of email interruptions.
Knowledge workers live in an environment that is
constantly interrupted by email (Ducheneaut and Bellotti
2001). When an email arrives randomly, additional time is
needed to switch from a current work medium to the email
medium. This time is referred to as switching time (Cutrell
et al. 2000; Czerwinski et al. 2000) or more commonly as
interruption lag (Trafton et al. 2003). Jackson and
colleagues (2001 and 2003) found that a knowledge worker
takes an average of 1 min and 44 s to react to a new email
by activating the email application. After processing the
email, the knowledge worker has to spend a small amount
of time before fully resuming the interrupted task. This time
is primarily spent on recollection and reengaging in the task
that was interrupted. This recovery time is also referred to
as resumption lag (Trafton et al. 2003) and has been
reported to be around 64 s per email interruption (Jackson
et al. 2003). Although this time component may appear to be
small, because of the large number of messages arriving every
day, the cumulative interruption and resumption lags become
large, and hence increase the knowledge worker’s non-valueadded time of a knowledge worker (Jackson et al. 2003).
Jackson and colleagues (2001 and 2003) performed
several studies to understand the role of email as an
interrupter in organizations. They suggest that the overall
interruption effect of email is more than that caused by
phone calls. Ironically, the frequency of interruptions can be
controlled by controlling the time frame(s) during which
interruptions are allowed to occur. Thus, it is possible to
reduce the effect of interruptions by scheduling the hours
during which email is processed. Jackson et al. (2003)
suggests that knowledge workers should check email every
45 min. However, the Jackson studies do not consider
several work environment characteristics such as different
content complexities of emails or different arrival patterns.
These factors may moderate the influence of the timing of
email processing on knowledge worker performance. There
is a need to study the effect of interruptions caused by emails
in a more detailed and elaborate manner. The following
section describes the first phase of our study’s experiment.
640
Inf Syst Front (2011) 13:637–653
3 Research framework
3.1 Phase I experiment
The research model for this phase of the study is illustrated
in Fig. 1. Three performance measures that are evaluated in
the study include (1) percent increase in knowledge worker
utilization, (2) average email response time, and (3) average
task completion time. Utilization is used as a measure of a
knowledge worker’s information overload in this study. It is
defined as the probability of a knowledge worker being in a
busy state (Her and Hwang 1989). The percent increase in
utilization reflects the non-value added time spent by
knowledge worker on a given day due to interruptions.
Work environment can be influence by several characteristics such as primary and secondary task characteristics,
sender receiver distance, task complexity, etc. (Te’eni
2001). We chose two work environment characteristics,
namely dependency on email communication, and email
arrival pattern. Based on the survey conducted by American
Management Association (2004), we categorize knowledge
workers, on the basis of their dependency on email
communication, into four different types: very high users
of email, high users, low users, and very low users. “Very
high” users spend an average of 4 h per workday
processing email, “high” users 3 h, “low” users 2 h and
“very low” users 1 h. “Very high” and “high” users of email
typically represent workers with a higher need for communicating at work, e.g. executives, CEOs, distribution and
marketing managers, sales personnel, marketing managers,
workers at geographically dispersed organizations, etc.
“Low” and “very low” users of email are knowledge
workers with less communication requirements, e.g. office
assistants, analysts, programmers, etc.
Emails often follow different arrival patterns in different
environments. Figure 1 shows two different email arrival
patterns. A time stationary exponential distribution is
representative of work environments where emails arrive
at a rate that remains roughly constant throughout the
workday, whereas a non-stationary arrival pattern is found
Work Environment Context
Dependency on Email Communication
(Very Low, Low, High, Very High)
Email Arrival Pattern
(Stationary, Non-Stationary)
Performance variables
Email processing policies
(a) % increase in utilization
(b) Average email response time
(c) Average primary task
completion time.
(d) Efficiency
Fig. 1 Research model for the phase I and phase II experiment
in those environments where the arrival pattern varies with
the time period.
There is a lack of consensus among research studies that
have been conducted to identify the best email processing
policy. Jackson et al. (2001 and 2003) have proposed that
email should be processed no more than every 45 min to
increase employee productivity at the workplace. Another
study (Venolia et al. 2001) argues that processing email
once per workday is a better policy than continuous
processing. Besides the lack of general agreement among
prior research findings, the three policies mentioned above
do not represent the broad spectrum of policies that a
knowledge worker might be able to use in efficiently
managing email and primary tasks.
According to the Single-Resource theory (Kahneman
1973), frequently diverting resources such as the attention
of a knowledge worker to a secondary task (email)
decreases the performance on the primary task. This theory
suggests segregating the time during which emails and
other tasks are given higher priority for processing, thereby
reducing the interaction between the two, can potentially
reduce the number of interruptions. Interruption-related
literature also confirms that whenever an interruption
occurs, switching time (interruption lags) and recall time
(resumption lag) is spent before the interrupted task is
resumed. As explained in the previous section, when the
frequency of interruptions increases, the cumulative resumption and interruption lags increases as well thereby
delaying the completion of the primary tasks. This further
justifies that controlling the time-frame within which email
is allowed to interrupt can reduce the number of interruptions, thereby reducing the cumulative switching and
recall time. Such controls also allow for better attention
allocation, which is a scarce resource in modern organizations (Davenport and Beck 2000).
To establish such a time-frame, we introduce the notion
of “email hour” slots. The total knowledge work hours in a
particular workday can be split into two categories: one,
during which email is given the highest priority, termed as
“email-hour” slots and the other, during which primary
tasks are given the highest priority, termed “non-email
hour” slots. By adjusting the length of each email-hour slot
and varying the number of such email-hour slots in a
particular work day, we may be able to reduce the number
of interruptions without adversely affecting the primary
task completion times as well as email response times.
Perlow (1999) qualitatively studied the effects of the
frequency and timing of interruptions on individual and
group productivity of knowledge workers. This study found
that neither continuous interruptions nor perfect synchronization between interrupting and interrupted tasks is good
for effective time management. This leads us to believe that
both extremes are not good for knowledge workers’
Inf Syst Front (2011) 13:637–653
performance and that the optimal policy is somewhere in
the middle. Zijlstra et al. (1999) also found that interruptions could cause people to perform a primary task more
quickly, but postulated that the relationship between
interruptions and task performance would be an inverted
U-shape, indicating that the cumulative effect of interruptions at some point does have a negative effect on primary
tasks. A controlled interruption policy such as processing
emails twice or four times a day will likely keep a better
balance between email response time and task completion
time than a continuous email processing policy and once-aday email processing policy. Hence, the research question
is:
Research question 1: Are there email processing
policies that will keep a balance between email
response time and task completion time?
Hypothesis 1(a) For different email arrival patterns (stationary and non-stationary), processing emails two or four
times a day will result in significantly higher email
response time and low primary task completion time than
with continuous email processing.
Hypothesis 1(b) For different levels of dependency on email
communication (Very High, High, Low, Very Low) processing emails two or four times a day will result in significantly
higher email response time and low primary task completion
time than with continuous email processing.
Hypothesis 2 Processing emails two or four times a day will
result in significantly less increase in worker utilization (due
to interruptions) than with continuous email processing.
3.2 Phase II experiment
The second phase of experiment contributes differently
from first experiment phase in that it models a different
type of knowledge worker, different performance measures, and it takes different approaches to modeling the
knowledge work environment, including modeling attention as an entity. Phase I experiment studies email
policies’ effects on utilization, email response time, and
task completion time. Phase II considers a different type
of knowledge worker—project managers who primarily
handle complex tasks and are rarely, if ever “caught up.”
Instead of utilization, which implies that the knowledge
worker experiences some “caught up” time, this study
considers knowledge worker efficiency and the total
amount of time needed by the knowledge worker to
complete a daily threshold of work. Efficiency is defined
as the knowledge worker’s productive time at work
divided by the knowledge worker’s total time at work.
641
Second phase also allows for one type of email message to
always take priority over other email messages, so that all
email need not interrupt the knowledge worker—just
those in need of urgent resolution. The modeling approach
is different as well. Rather than modeling the knowledge
worker as a server, this phase models a knowledge
worker’s attention as an entity that flows from one area
of focus to another. Modeling knowledge worker attention
as an entity is described in the following section.
Additionally, emails with different priorities were also
modeled during this phase. Email that require urgent
attention are of highest priority (priority-1) whereas emails
not defined as urgent must be processed within either one
business day (priority-2) or within 1 week (priority-3).
We learn from Speier et al. (1999, 2003) that interruptions can adversely affect complex tasks. We learn from
Trafton et al. (2003) that interruptions can cause waste in
worker productivity in the form of both interruption and
resumption lags. Jackson et al. (2003) gives further
evidence of the existence of these lags and approximations
for the durations of these lags. By eliminating email
interruptions, knowledge worker efficiency should improve.
Research question 3: Will fewer interruptions result in
more efficient work completion? Productive time
includes working on both primary work and email.
Total time includes primary work and email work as
well, but also includes time wasted in interruption and
resumption lags. Specifically, will the proposed email
processing policy significantly improve knowledge
worker efficiency?
Hypothesis 3 Dividing non-priority email work into two
specific time frames (Scheduled Attention-2) will result in
significantly greater efficiency when compared to processing email continuously.
Research question 4: Will fewer interruptions
lower the numbers of hours worked daily?
Hypothesis 4 Holding email hours twice daily (Scheduled
Attention-2), will result in significantly fewer total hours
worked daily when compared to processing email continuously (Continuous Attention).
Research Question 5: To what extent will email
arrival patterns influence the success of given email
processing policies?
Just as we would schedule employees during the busiest
times of day, it is intuitive that scheduling email hours
during periods of rapid email arrival rates should allow for
prompt resolutions.
642
Inf Syst Front (2011) 13:637–653
Hypothesis 5 Email hours scheduled during peaks in arrival
patterns (Scheduled Attention-2P and Scheduled Attention4P) will have significantly shorter resolution times when
compared to email hours not scheduled during peaks in
arrival patters (Scheduled Attention-2 and Scheduled
Attention-4).
4 Model formulation
In this section, we will briefly describe the conceptual
development of the model, the stages of interruptions
within knowledge work environment, and different email
processing policies. The policies that were compared in
phase I and II are described in Table 1.
Although the timely processing of all primary tasks is of
importance to a knowledge worker, email processing
cannot be indefinitely ignored as email could facilitate the
sharing of information that is necessary for completion of
primary tasks. We implement email processing policies by
establishing email hours. The continuous email processing
policy implies that every working hour is an email hour. An
email arriving during email-hours will interrupt other
ongoing primary tasks, because the highest priority is given
to email.
During non-email hours, primary tasks have the highest
priority and email cannot preempt a primary task. The
manner in which any new knowledge object will be
handled is contingent upon three things: when it arrived
(during email hours or non-email hours), whether it is email
or a primary task, and what is the length of the (email or
primary task) queue? An email that arrives during email
hours will take priority over any other task.
Whenever an interruption occurs, additional time is spent
on switching from one task to another (for example,
moving from one medium to another and activating the
email application). This nonproductive time is referred to as
an interruption lag. Processing of an interrupted primary
task is resumed once the processing of the interrupt (email)
is complete, some time is needed to recall the work done on
previously interrupted work. This nonproductive time is
referred to as a resumption lag. An email arriving at a time
during which the knowledge worker is idle does not cause
any interruption. Similarly, if a task arrives during an emailhour slot, then it waits in the queue till the remaining emails
have been processed. During non-email hours, an email is
processed only if there are no primary tasks in need of
processing, because primary tasks are given the highest
priority during non-email hours.
Various email types that differ from each other on the
basis of content, complexity, and arrival rate were modeled,
bringing greater realism to the model. For example, some
email messages require longer processing time than others
due to their complexity. Phase I model also considered
various email characteristics and types. Five different email
types were modeled (see Appendix I, Table A.1). These
include spam (type 1), priority email (type 2), and
informative email (type 3). Emails that require response
were further categorized into two additional types: those
with service times that do not change based on the age of
the email (type 4), and those with service times dependent
on the age of email (type 5). Type 5 email service times
change as a function of the time for which they remain
unanswered. This type of email, if it waits a while, requires
no action. This could result from someone else on the
message recipient list responding or other circumstances
resulting in resolution of the email.
Phase II experiments modeled two separate entities.
First, a knowledge worker’s flow of attention is modeled.
“Attention” represents the focus of the knowledge worker’s
mental efforts. “Flow of attention” implies that the
knowledge worker’s attention shifts between different areas
of focus. Second, the flow of email messages is modeled
separately from the flow of knowledge worker attention.
Upon arrival in the knowledge worker’s inbox, the email
message must wait for the knowledge worker’s attention.
The delay incurred by the email message is dependent upon
the priority of the email message, and the knowledge
worker’s email processing strategy. Email messages are
prioritized according to urgency, and queued accordingly.
Table 1 The email processing policies
Processing strategies
Descriptions
Continuous attention (C)
Scheduled attention-1 (C1)
Scheduled attention-2 (C2)
Scheduled attention-2P
Scheduled attention-4 (C4)
Scheduled attention-4P
Scheduled attention-6 (C6)
Jackson attention (C8)
This
This
This
This
This
This
This
This
processing
processing
processing
processing
processing
processing
processing
processing
strategy
strategy
strategy
strategy
strategy
strategy
strategy
strategy
requires
requires
requires
requires
requires
requires
requires
requires
processing email as they arrive (giving first priority to email).
holding email hours once daily, every morning.
holding email hours twice daily.
holding email hours twice daily, during two peak email arrival periods.
holding email hours four times daily.
holding email hours four times daily, during four peak arrival periods.
holding email hours six times daily.
holding email hours every 45 min.
Inf Syst Front (2011) 13:637–653
Priority-1 (urgent) email messages immediately gain the
attention of the knowledge worker provided that the
knowledge worker is not idle (at lunch or gone for the day),
and all priority-1 email messages having arrived earlier have
been processed. Non-urgent email messages gain the
attention of the knowledge worker under differing circumstances depending on the knowledge worker’s email processing strategy. If the knowledge worker employs a
“continuous” email processing strategy, then an email is
processed after all email of higher priority have been
processed and after all email of equal priority having arrived
earlier have been processed. If the knowledge worker
employs a “scheduled attention” email processing strategy,
then non-priority email messages must wait for a specific
time or times during the day during which the knowledge
worker processes non-priority email messages. During these
time periods an email is processed after all email of higher
priority have been processed and after all email of equal
priority having arrived earlier have been processed.
Upon starting the workday, the knowledge worker will
begin with his or her primary work, unless the specific
email processing strategy calls for processing email at this
time. For example, email hours could be scheduled for first
thing in the morning. Before the primary work takes place,
the knowledge worker must take time to again familiarize
with the work that is to be performed (the resumption lag).
The primary work continues until one of four things
happen. First, an urgent email message could interrupt the
knowledge worker. Second, the knowledge worker may
break for lunch; third, the knowledge worker’s email
processing strategy may dictate that it is time to process
non-urgent email messages. And fourth, the knowledge
worker may have completed a given level of work and
leave for the day.
Appendix II describes the mathematical model supporting the phase I and II simulation models.
643
Numerous simulation models representing different
types of work environments were developed and run for
phase I and II in order to compare the performance of
various types of knowledge workers under different email
processing policies. Our approach of building a series of
simulation models in this study is based upon the guidelines
postulated by Chwif et al. (2000): “divide your model into
parts and model each part separately creating a series of
simpler models instead of one ‘huge’ one” and “only after
you validate, analyze and have the results, add more
complexity if you feel it is really necessary.”
Phase I experiments were based on the parameters listed in
appendix I (Table A.2). All tasks in Phase I followed an
exponential inter-arrival time distribution. Simulation models
representing each scenario described in appendix I
(Table A.2) were run for all five policies. Thus, in Phase I,
90 simulations were run for the duration of 500 days with a
warm-up time of 50 days. The warm-up time was determined
externally by analyzing the data using Welch’s method
(Welch 1983). Sixteen different work scenarios were implemented and compared in this phase. Thus, 80 different
simulation models were replicated 20 times and run for
500 days with a warm-up time of 50 days, leading to
generation of 1,600 data points. Collected data were analyzed
using Multivariate Analysis of Variance (MANOVA) in order
to control the experiment-wide error rate. The different email
processing policies comprise the independent variable.
Similarly, within phase II experiments, simulations of each
processing policy were performed and performance measures
were collected. The collected data included six performance
measures: Efficiency, Hours-Worked, and the EmailResolution-Time for each of the four priorities of email
message. Each model was run for 90 days with a warm-up
time of 30 days and for 20 replications. The warm-up period
was obtained using Welch’s method.
6 Results
5 Research method
6.1 Results of phase I experiment
Due to the time-length and the nature of the policies being
compared, it is extremely difficult to conduct this study as
an experimental or field study with enough control. Hence,
a simulation-based computer experiment was chosen to
study this problem. Axelrod (2003) describes simulation as
a way of doing ‘thought experiments’ and as a technique
that can give surprising ‘emerging’ results due to due to
presence of various interactions among entities that are
often difficult to anticipate. Computer simulation has been
used as a method for theory development and investigation
(Hans-Joachim et al. 2001; Peschl and Scheutz 2001;
Di Paolo et al. 2000) and can be used to conduct virtual
experiments (Winsberg 2003).
Figure 2 (a) shows that the percent increase in utilization of
the knowledge worker using the once-a-day (C1) policy is
2% for very low email dependency, whereas it is 5.7% for
very high email dependency. For four-times-a-day (C4)
policy, the percent increase in utilization varies from 3% to
5% for very low to high dependency on email. With
Continuous policy (C), the percent utilization increased the
most for all levels of dependency (8–15%). Figure 2 (b)
shows that percent utilization increased from 4% to 14% as
we moved from C1 to C for time-stationary arrivals and
increased from 3.7% to 11.5% on moving from C1 to C for
non-time stationary arrivals. C1, C2 and C4 performed
644
Inf Syst Front (2011) 13:637–653
a
a
% increase in utilization (base value=0.9)
9
Avg Primary Task Completion time (in min)
2000
7
6
Email dependency
5
4
very high
3
high
2
low
1
very low
C1
C2
C4
C8
Estimated Marginal Means
Estimated Marginal Means
8
1000
Email Dependency
very high
high
low
C
very low
0
POLICY
C1
C2
C4
C8
C
POLICY
b
% increase in Utilization (base value 0.9)
b
16
Avg Primary Task Completion time (in min)
2000
12
10
8
6
Email Arriv. Pattern
4
Exponential (Expo)
Estimated Marginal Means
Estimated Marginal Means
14
1000
Email Dependency
very high
high
Non Stationary Expo.
2
C1
C2
C4
C8
low
C
POLICY
very low
0
C1
better than C8 and C. Figure 3(a) and (b) describe the effect
of policy on additional time spent per day due to
interruptions across various levels of email dependency
and arrival patterns.
For the C1 policy, the cost of interruptions in terms of
time varies from 10 to 20 min. For C4, it varies from 12 to
22 min, and for the C policy, it varies from 35 min to an
hour. Figure 4 shows the impact of various policies on
email response time across various levels of email
dependency. The C1 policy showed the longest average
wait time (250–370 min) whereas the C policy showed the
smallest wait time (15–20 min) for all levels. Similar
patterns appeared across various levels of content complexity and email arrival patterns and hence those figures are
not shown here. The average primary task completion time
increased substantially during the use of C (200 min) and
C8 (700 min), whereas for C1, C2 and C4 policies, it was
C4
C8
C
POLICY
Fig. 3 (a): Effect of policy x email dependency on additional time spent.
(b): Effect of policy x email arrival pattern on additional time spent
Additional time spent (in min)/day
40
Estimated Marginal Means
Fig. 2 (a): Effect of policy x email dependency on percent increase in
utilization. (b): Effect of policy x email arrival pattern on percent
increase in utilization
C2
30
Email Dependency
20
very high
high
10
low
very low
0
C1
C2
C4
C8
C
POLICY
Fig. 4 Effect of policy x email dependency on email response time
Inf Syst Front (2011) 13:637–653
645
Additional time spent (in min)/ day
70
Estimated Marginal Means
60
50
40
30
Email Arriv. Pattern
20
Exponential (Expo)
10
Non Stationary Expo
C1
C2
C4
C8
C
POLICY
Fig. 5 Effect of policy x email dependency on primary task
completion time
between 30 min and 100 min (Fig. 5). The same behavior
was noted across different email dependency levels and
email arrival patterns. Hypotheses 1(a, b) and 2 were found
to be significant at the significance level of 0.05.
6.2 Results of phase II experiment
Table 2 summarizes the results of the MANOVA model.
The email processing policy was found to have a
significant effect (α=0.001). All of the email processing
policies were compared across multiple performance
measures. Tests for differences between groups (email
processing policies) were performed using the Bonferroni
approach for adjusting alpha to account for inflation of the
overall type I error rate resulting from multiple performance
measures.
Hypothesis 3 was supported. The expected gains in
efficiency were found to be statistically significant (α=0.001).
Hypothesis 3 Dividing non-priority email work into two
specific time frames will result in significantly greater
efficiency when compared to processing email continuously.
Table 2 Multivariate tests
Effect
Intercept
Strategy
a Exact statistic
b The statistic is an upper bound
on F that yields a lower bound
on the significance level
Pillai’s trace
Wilks’ lambda
Hotelling’s trace
Roy’s largest root
Pillai’s trace
Wilk’s lambda
Hotelling’s trace
Roy’s largest root
The efficiency that resulted from the Scheduled Attention-2
processing policy was 97.35%, indicating that less than 3% of
the knowledge worker’s work day was wasted on interruption
and resumption lags. The Continuous email processing policy
resulted in efficiency of 94.34%, a mean difference of roughly
3%. What can a knowledge worker do with 3% of their time
back? 3% of a 9 h day corresponds to roughly 16 min per day.
16 min per day corresponds to an hour and 20 min per week,
or around 69 h per year. Consider the knowledge workers used
in this case study who are billed out at $300–$400 per hour.
Considering an organization with dozens or hundreds of
knowledge workers, the cost of email interruptions adds up!
But it need not. Choosing the Scheduled Attention-2 processing policy achieved efficiency without adversely affecting the
successful resolution of email messages.
Hypothesis 4 was not supported. The Scheduled Attention2 email processing policy resulted in an average daily total
hours worked of 10.5524, while the Continuous email
processing policy resulted in an average daily total hours
worked of 10.0095. The difference was statistically significant (α=0.001), however the direction of the difference was
not as expected. The explanation is of interest, however.
Hypothesis 4 Holding email hours twice daily will result in
significantly fewer total hours worked daily when compared to processing email continuously.
In light of efficiency, considerably more work is being
accomplished with the Scheduled Attention-2 email processing policy. The Scheduled Attention-2 email processing
policy results in an average of 10.2728 productive hours daily
(10.5524 * 97.35%). The Continuous email processing policy
results in an average of 9.4430 productive hours daily
(10.0095 * 94.34%). Consider a project requiring 160 h of
work. Using the Continuous email processing policy, the
project could be completed in approximately 17 work days. If
the knowledge worker instead employed the Scheduled
Attention-2 processing policy, the project could be completed
within roughly 15.5 work days. With the Scheduled Attention2 processing policy, email has a tendency to hold the
Value
F
Hypothesis df
Error df
Sig.
1.000
.000
1048683.824
1048683.824
4.591
.000
507.821
426.365
87739879.955(a)
87739879.953(a)
87739879.953(a)
87739879.953(a)
137.677
744.819
3528.889
18013.910(b)
6.000
6.000
6.000
6.000
72.000
72.000
72.000
12.000
502.000
502.000
502.000
502.000
3042.000
2736.980
3002.000
507.000
.000
.000
.000
.000
.000
.000
.000
.000
646
Table 3 Scheduled attention-2
(pattern and no pattern), email
resolution times for both
priority-2 and priority-3
email messages
Inf Syst Front (2011) 13:637–653
Priority 2 email
Priority 3 email
Scheduled attention 2 pattern
Scheduled attention 2 no pattern
Mean
Median
Min
Max
Mean
Median
Min
Max
Mean
Median
Min
Max
Mean
Median
Min
Max
3.5226
2.2690
.0012
22.3474
2.3879
1.6391
.0010
23.2022
3.6419
3.1322
.0007
22.5887
4.0288
3.2893
.0086
23.8542
knowledge worker at work for longer hours. With Continuous
processing, you are more or less always caught up. With the
Scheduled Attention-2 email processing policy, the “email
hours” are scheduled in the morning and at the end of the day.
The knowledge worker will often stay late catching up on the
day’s email processing needs.
Scheduled Attention-2 might be especially effective for
the knowledge worker who is facing a deadline or who is
more concerned with getting things done than going home
at a particular time. Alternatively, Scheduled Attention-2
could be tweaked to process email a bit earlier in the day in
hopes of always being caught up by the end of the day. As
indicted in Table 3 below, the mean resolution times for
priority-2 email messages do not support hypothesis 5. The
difference between 3.5226 h (Scheduled Attention-2—with
pattern) and 3.6419 h (Scheduled Attention-2—no pattern)
is not statistically significant.
those email that do not require a particularly prompt
response (within 1 week), so the result may be seen as less
important. However, all things being equal, faster resolutions are desirable.
Hypothesis 5 also considers the mean resolution times of
priority-2 and priority-3 email messages resulting from the
Scheduled Attention-4 (no pattern) and the Scheduled
Attention-4 (with pattern) email processing policies described in Table 4.
The mean resolution times for priority-2 email do not
support this hypothesis (2.7862 and 2.5445 h are not
different statistically). The mean resolution times for
priority-3 email do not support the hypothesis either
(1.9202 h is not statistically different from 1.9236 h). An
analysis of frequency distributions, Figs. 7 and 8 below
reveal little difference in the two alternative email processing policies.
Hypothesis 5 Email hours scheduled during peaks in arrival
patterns will have significantly shorter resolution times
when compared to email hours not scheduled during peaks
in arrival patterns.
7 Discussion, limitations, and implications
for future research
However, the mean resolution times for priority-3 email
do support this hypothesis (2.3879 h is a statistically
significant shorter resolution time than 4.028 h (α=
0.001)). The schedule for processing email according to
arrival patterns is based on the total arrivals of email of
both priority-2 and priority-3.
Priority-3 email messages more closely matches the
scheduled email hours, causing the mean resolution time to
be shorter with priority-3 emails than the priority-2 email
messages. An analysis of frequency distributions, Fig. 6,
indicates that with respect to priority-3 email, scheduling
email hours during peaks in arrival patterns prevents a
bimodal distribution of resolution times. If the knowledge
worker were to adopt two specific “email hours” during his
or her day, selecting email hours that coincide with email
arrivals can improve the resolution time of a good many of
the priority-3 email that are processed. Priority-3 emails are
Figure 2 (a, b) illustrate the variation of percent increase in
utilization of the knowledge worker due to interruptions
with respect to the policy used. The graphs show that
percent increase in utilization first decreases and then
increases after reaching a certain minimum value. The
percent increase in utilization reaches lowest values at C4
and C2, thus providing the answer to RQ1.
Length and the number of the email-hour slots, and the
time gap between email-hour slots in a particular policy
impacts the number of interruptions and the resulting
increase in overload. The C1 policy comprises a single
email-hour slot of 3 h. Since the length of the single
email-hour slot is relatively long in C1, the probability of
any new email arriving and leading to an interruption is
rather high during the three hour duration. The length of
each email-hour slot in C4 is relatively small (approx.
45 min). Due to the shorter email-hour slots in the C4
policy, the email queue builds up. The primary tasks are,
Inf Syst Front (2011) 13:637–653
a
Scheduled Attention-2 (No Pattern) Email Processing Strategy,
Priority 3 Email Resolution Times
6000
120.00%
5000
100.00%
4000
80.00%
3000
60.00%
2000
40.00%
1000
20.00%
25
24
23
21
20
19
18
17
16
15
14
13
12
9
10
8
7
6
5
4
3
2
1
0
0
11
Frequency
Cumulative %
22
Frequency
Fig. 6 Scheduled attention-2
(pattern and no pattern), email
resolution times for priority-3
email messages
647
0.00%
Hours
Frequency
b
Scheduled Attention-2 (Pattern) Email Processing Strategy,
Priority-3 Email Resolution Times
6000
120.00%
5000
100.00%
4000
80.00%
3000
60.00%
2000
40.00%
Frequency
1000
20.00%
Cumulative %
e
24
or
M
.5
22
21
.5
19
18
.5
16
15
.5
13
12
.5
9
10
7.
5
6
5
4.
3
5
0.00%
1.
0
0
Hours
thus, less likely to be processed during these email-hours.
Assuming the same rate as other cases, fewer emails arrive
during this time, leading to reduced interruptions. Thus,
the probability of interruption due to a newly arrived email
is also small, implying smaller cumulative resumption and
interruption lags. On the other hand, as the number of
email-hour slots increases such as in C8, the frequency of
email processing increases. The likelihood of primary task
Table 4 Scheduled attention-4
(pattern and no pattern), email
resolution times for priority-2
and priority-3 email messages
Priority 2 email
Priority 3 email
processing also increases leading to an increase in
interruptions. In the continuous policy, the length of each
email-hour slot approaches zero whereas the number of
email-hour slots approaches infinity. The priority is always
rendered to email. Hence, the number of interruptions
increases, leading to an increase in the cumulative sum of
interruption lag and resumption lag in the continuous
policy.
Scheduled attention-4 pattern
Scheduled attention-4 no pattern
Mean
Median
Min
Max
Mean
Median
Min
Max
Mean
Median
Min
Max
Mean
Median
Min
Max
2.7862
0.9759
0.0009
21.1193
1.9202
1.3045
0.0014
21.1648
2.5445
1.4184
0.0010
17.7263
1.9236
1.3236
0.0023
17.9526
648
Fig. 7 Scheduled attention-4
(pattern and no pattern), email
resolution times for priority-2
Inf Syst Front (2011) 13:637–653
a
Scheduled Attention 4 (Pattern) Email Processing Strategy,
Priority-2 Email Resolution Times
120.00%
25000
100.00%
20000
Frequency
80.00%
15000
60.00%
10000
40.00%
5000
20.00%
Frequency
Cumulative %
25
24
23
22
21
20
19
18
17
16
15
14
13
12
11
9
10
8
7
6
5
4
3
2
1
0.00%
0
0
Hours
b
Scheduled Attention 4 (No Pattern) Email Processing Strategy,
Priority-2 Email Resolution Times
25000
120.00%
100.00%
20000
Frequency
80.00%
15000
60.00%
10000
40.00%
5000
20.00%
25
24
23
22
21
20
19
18
17
16
15
14
13
12
11
9
8
7
6
5
4
3
2
1
0
0
10
Frequency
Cumulative %
0.00%
Hours
The experiments also modeled different complexity
levels of email. The variation in primary task completion
times across policies is a result of the interaction between
three different factors. The most important factor is the
number of interruptions that occur during the use of a
particular policy. Second is the number of email-hour slots
that a particular policy comprises, and third, to a lesser
extent, is the time interval between consecutive email-hour
slots. In phase I and II, Scheduled attention policies such as
two times a day, four times a day performed better than
continuous email processing for primary tasks. The difference can be attributed to a large number of interruptions
that occur while using the C policy in comparison to C4
and C8. The increase in the number of interruptions clearly
affected the cumulative sum of the interruption lag and the
resumption lag, leading to a rise in the overall task
completion time. The gap between the completion times
of primary tasks further widens from C4 to C due to an
increase in the number of email-hour slots and also due to
the reduction in the time interval between email-hour slots
as we move from C4 to C. Thus C2 and C4 performed
better than all other policies for all types of primary tasks
and email because the number of interruptions was least in
these policies.
The mean response time for email decreases with the
increase in the number of email-hour slots. The logical
explanation is that as the gap between the consecutive
email-hour slots becomes smaller from C1 to C8 and
completely vanishes in C, email gets processed more
quickly, resulting in the reduction in average email response
time. What is less obvious, however, is that the marginal
reduction in email response times between C4 and C8 or C
come at a high cost in terms of increases in other task
completion times. The same results are seen in the second
experiment. These experiments also revealed that scheduled
attention policies (C2 andC4) outperformed C8 and C in
balancing the primary task completion time and email
response time for all work environments and for all levels
Inf Syst Front (2011) 13:637–653
Fig. 8 Scheduled attention-4
(pattern and no pattern), email
resolution times for priority-3
649
a
Scheduled Attention-4 (Pattern) Email Processing Strategy,
Priority 3 Email Resolution Times
120.00%
8000
7000
100.00%
6000
80.00%
Frequency
5000
60.00%
4000
3000
40.00%
2000
20.00%
Frequency
1000
Cumulative %
25
24
23
22
21
20
19
18
17
16
15
14
13
12
11
9
10
8
7
6
5
4
3
2
1
0.00%
0
0
Bin
b
Scheduled Attention-4 (No Pattern) Email Processing Strategy,
Priority 3 Email Resolution Times
120.00%
8000
7000
100.00%
6000
Frequency
80.00%
5000
60.00%
4000
3000
40.00%
2000
20.00%
Frequency
1000
Cumulative %
0.00%
25
24
23
22
21
20
19
18
17
16
15
14
13
12
11
10
9
8
7
6
5
4
3
2
1
0
0
Hours
of email dependencies, email arrival patterns, email
complexities, and email arrival rate.
This study provides numerous insights into the impact of
interruptions looks into the problems of email overload and
interruptions, simultaneously. The approach undertaken by
the authors required the development of complex simulation models to represent human information processing and
validating them using the procedure suggested by Sargent
(2003). However, known validity and generalizibility issues
are associated with any simulation-based study. Another
limitation of this study is related to the issue of sample size
and power. With simulation, sample size is generally
inflated due to the need for running a large number of
replications so that variability of stochastic output is
reduced. This gives more artificial power to the sampling
experiment. Hence, statistical significance of results should
not be overemphasized in a simulation based study. Rather,
researchers should pay careful attention to the practical
significance of the results.
Some assumptions have to be made when developing
simulation models. However, these assumptions provide
opportunities for future work as well. For example, the
impact of different time zones on email arrival pattern was
not modeled in this study. This becomes an important topic
for study given the recent proliferation of virtual communities and the heavy outsourcing of knowledge work to
foreign offices that operate under different time zones.
In the current study, we assumed that all the interruptions
caused by email are harmful and delay the processing of
primary tasks. In an information-sharing context, not all
email can be associated with a negative cost. Some email
may actually speed up completion of other tasks at hand.
Some internally generated email (a message from a project
partner) may have a reward associated with it. Thus, more
650
comprehensive simulation scenarios should be designed in
future studies. One of the underlying assumptions of this
study is that recall time increases with the increase in the
time that has been spent on processing before an interruption occurs. Further research needs to be done where the
resumption lag varies non-linearly with the fraction of task
completed when the interruption occurs.
The current study has focused on the individual
knowledge worker. Future research may study the problem
at the network level. It would be interesting to see how the
performance of policies changes when a network of
knowledge workers is studied. Studies should be conducted
to account for a greater number of discrete policies that are
not just time or frequency based. Future replications based
on independently developed models, and experimental or
field studies should be conducted to further validate our
results. Perhaps not impossible, but it is very difficult to get
human subjects for studies where such policies can be
dictated to be used by participants all day for several days.
Several economic and accounting or cost based approaches
can also be taken to study the same problems, since the
extra time spent in task processing can be treated as cost in
terms of time.
Future research needs to be pursued that will help
establish a research framework for studying email processing policies. For example, factor such as individual
characteristics cannot be easily modeled using simulation
and alternate approaches could be used to study it. One of
the limitations of this study is that it does not consider the
behavioral differences between knowledge workers. Our
findings suggest what a knowledge worker should ideally
do; but these policies need not always behave in the same
manner if the behavioral aspects are to be considered. For
instance, some workers like variety and work more
efficiently with interruptions than others do. So instead of
C4, C policy may turn out to be optimal in this case.
Various personality and cognitive styles should be considered in any future work to study this problem. As pointed
out by Weber (2004) in his recent MIS Quarterly editorial
note, tremendous research opportunities exist in email
research.
The results of this study contribute to the understanding
of email overload and interruptions in an informationprocessing environment and contradict some of the suggestions made in earlier research. A concrete recommendation
from earlier research is that continual or high frequency
email processing is not the best policy to reduce the
interruption effect of email. This study shows that checking
email two to four times a day is a better policy in the work
environments studied. These policies tend to reduce the
overload due to interruptions and at the same time attempt
to achieve an optimum balance between primary task
Inf Syst Front (2011) 13:637–653
completion time and email response time. We found that a
good policy is not to have too few (C1) or too many (C8
and C) email priority hours. The optimal number is
somewhere in the middle i.e. C2 to C4.
Through this study, we illustrate how simulation can
serve as a very useful tool for analyzing the phenomenon of
email overload and interruptions. This study also shows
that simulation can provide enormous advantages in
studying a problem for which data collection becomes a
major challenge due to the site unavailability, where field or
experimental studies are difficult to conduct, and where
human subjects cannot be utilized easily.
In future studies, simulation may be combined with
qualitative data analysis for initial theory development
purposes (Peschl and Scheutz 2001). We encourage IS
researchers to use simulation along with qualitative and
quantitative methodologies as their vehicle for conducting
research, for initial theory development, extension, and
verification. Finally, this study has important practical
significance for those geographically dispersed and service
organizations where knowledge workers are spending large
amounts of time on processing email such as in contact
centers, virtual teams, and marketing firms.
Developing organizational wide policies to encourage
users to check their emails on a scheduled basis rather than
continuously could save an organizations thousands of
hours each year. Such schedules can also be implemented
by scheduling deliveries of emails to the users’ email boxes
periodically rather than continuously. It is also conceivable
to develop policies that are appropriate for different classes
of users. Further work is necessary to validate the results of
this study in industry and to develop implementation
mechanisms.
Appendix I
Table A.1 Email types, processing time of email and primary task
Notation
Email/task type
Discrete arrival
percentage
Processing
time (min)
1
2
3
4
Priority email
Spam
Informative email
Email with
non-diminishing
service time
Email with
diminishing
service time
Primary task
5%
5%
20%
55%
Expo(10)
Expo(0.5)
Expo(5)
Ref. Table A.2
15%
Ref. Table A.2
5
Expo (6)
Inf Syst Front (2011) 13:637–653
651
Table A.2 Parameters used in experiment phase I
S#
Type 4 email (E)
Processing Time (PT)
Type 5 E
PT (min)
Total email
PT per day
Avg. email
arrival rate
Primary task (P)
arrival rate /day
E utilization
(Util)
P util
Min
(E+P)Util
1
2
3
4
5
6
7
8
9
10
11
12
13
5
15
5
15
5
15
5
15
5
15
5
15
5
5
15
5
15
5
15
5
15
5
15
5
15
5
1
1
1
1
2
2
2
2
3
3
3
3
4
12
5
12
5
24
10
24
10
36
15
36
15
48
62
62
62
62
52
52
52
52
42
42
42
42
32
0.125
0.125
0.125
0.125
0.25
0.25
0.25
0.25
0.375
0.375
0.375
0.375
0.5
0.775
0.775
0.775
0.775
0.65
0.65
0.65
0.65
0.525
0.525
0.525
0.525
0.4
0.9
0.9
0.9
0.9
0.9
0.9
0.9
0.9
0.9
0.9
0.9
0.9
0.9
14
15
16
15
5
15
15
5
15
4
4
4
20
48
20
32
32
32
0.5
0.5
0.5
0.4
0.4
0.4
0.9
0.9
0.9
Appendix II. The mathematical model
Notation
i
j
k
d
s
t
X
λ
email processing strategy employed
ijkt
Pkds
type of email
message
urgency (priority)
of message
category of
processing need
day
sequence number
time period of day
i=1,2..I
{i=1 for SPAM,
i=2 for Irrelevant,
i=3 for Read only,
i=4 for Reply}
j=1,2..J
{j=1 for Urgent (Priority-1),
j=2 for within
Business Day (Priority-2),
j=3 for within 1 week
(Priority-3),
j=4 for Irrelevant}
k=1,2..K
{k=1 for <1 min,
k=2 for 1–10 min,
k=3 for >10 min}
d=1,2..D
s=1,2..S
t=1,2..T
{t=1 for 8:00 a.m. until 10:00 a.m.,
t=2 for 10:00 a.m. until 12:00 p.m.,
t=3 for 12:00 p.m. until 2:00 p.m.,
t=4 for 2:00 p.m. until 4:00 p.m.,
t=5 for 4:00 p.m. until 6:00 p.m.,
t=6 for 6:00 p.m. until 8:00 a.m.}
pkds
fkds P ðxÞ
Rds
rds
fds R ðxÞ
Lds
lds
fds L ðxÞ
Qd
qd
f d Q ð xÞ
Wqjs
Wsjs
arrival rate of email messages of type i, urgency j,
processing need k, occurring during time period t
random variable that represents the processing time
required for email of type k, occurring on day d,
having sequence number s
E(Pkds)
probability density function (pdf) of Pkds
random variable that represents the resumption lag
occurring on day d, sequence number s
E(Rds)
pdf of Rds
random variable that represents the interruption lag
occurring on day d, sequence number s
E(Lds)
pdf of Lds
random variable that represents the threshold of
productive work (email processing and primary
work) to be completed on day d
E(Qd)
pdf of Qd
email’s wait in the queue (time spent waiting for the
knowledge worker’s attention) for email of urgency
j having sequence number s
email’s wait in the system (email resolution time)
for email of urgency j having sequence number s
Wsjs ¼ Wqjs þ Pkds
W sjs
mean email resolution time for email of urgency j
W sj ¼
Yd
P
s Wsjs =S
total email processing occurring on day d
Yd ¼
PP
k
s Pkds
652
Zd
Gd
Inf Syst Front (2011) 13:637–653
total amount of primary work completed on day d
total lag time occurring on day d
Gd ¼
Hd
P
s Lds
þ
P
s Rds
total hours worked by the knowledge worker on day d
Hd ¼ Yd þ Zd þ Gd
H
Ed
mean hours worked by the knowledge worker
P
d Hd =D
knowledge worker efficiency occurring on day d
Ed ¼ ðYd þ Zd Þ=Hd
E
Qd
mean knowledge worker efficiency
P
d Ed =D
threshold of productive work (email processing
and primary work) to be completed on day d
Qd <¼ Yd þ Zd
References
American Management Association. 2004. Workplace email and
instant messaging survey. AMA Research. URL: http://www.
epolicyinstitute.com/survey/survey04.pdf
Axelrod, R. (2003). Advancing the art of simulation in the social
sciences. Japanese Journal for Management Information System,
12(3), 1–19. Special Issue: Agent-Based Modeling.
Berghal, H. (1997). Email—the good, the bad and the ugly.
Communications of the ACM, 40(4), 11–15.
Chwif, L., Barretto, M., & Paul, R. (2000). On simulation model
complexity. In J. A. Jones, R. R. Barton, K. Kang & P. A. Fishwick
(eds.), Proceedings of 32nd Winter Simulation Conference,
Orlando, Florida (pp. 449–454).
Corragio, L. (1990). Deleterious effects of intermittent interruptions
on the task performance of knowledge workers: A laboratory
investigation. Unpublished Ph. D. thesis, U. of Arizona.
Cutrell, E., Czerwinski, M., & Horvitz, E. (2000). Effects of instant
messaging interruptions on computing tasks. In Extended
Abstracts of CHI ’2000, Human Factors in Computing Systems,
(The Hague, April 1–6, 2000), ACM press, 99–100.
Czerwinski, M., Cutrell, E., & Horvitz, E. (2000). Instant messaging
and interruption: Influence of task type on performance. In Paris,
C., Ozkan, N., Howard, S. and Lu, S. (eds.), OZCHI 2000
Conference Proceedings, Sydney, Australia, Dec. 4–8, pp. 356–
361.
Davenport, T. H., & Beck, J. C. (2000). Getting the attention you
need. Harvard Business Review, 78(5), 119–126.
Denning, P. (1982). Electronic junk. Communications of the ACM, 25
(3), 163–165.
Di Paolo, E. A., Noble, J., & Bullock, S. (2000). Simulation models as
opaque thought experiments. In: M. A. Bedau, J. S. McCaskill,
N. H. Packard & S. Rasmussen (eds.), Artificial Life VII:
Proceedings of 7th International Conference on Artificial Life
(pp. 497–506). Cambridge, MA: MIT Press.
Ducheneaut, N., & Bellotti, V. (2001). E-mail as habitat. Interactions,
8(5), 30–38.
Duchenaut, N., & Watts, L. (2005). In search of coherence: a review
of email research. HCI Journal, 20(1, 2) (forthcoming).
Hall, E. T., & Hall, M. R. (1990). Understanding cultural differences:
Keys to success in West Germany, France and the United States.
Yarmouth: Intercultural.
Hans-Joachim, M., Karsten, S., Florin, A., & Heinz, H. (2001).
Computer simulation as a method of further developing a theory:
simulating the elaboration likelihood model. Personality &
Social Psychology Review, 5(3), 201–215.
Her, C., & Hwang, S. (1989). Application of queuing theory to
quantify information workload in supervisory control systems.
International Journal of Industrial Ergonomics, 4, 51–60.
Jackson, T., Dawson, R., & Wilson, D. (2001). The cost of email
interruption. Journal of Systems and Information Technology, 5
(1), 81–92.
Jackson, T., Dawson, R., & Wilson, D. (2003). Understanding email
interaction increases organizational productivity. Communications of the ACM, 46(8), 80–84.
Jett, Q. R., & George, J. (2003). Work interrupted: a closer look at the
role of interruptions in organizational life. Academy of Management, 28(3), 494–507.
Kahneman, D. (1973). Attention and effort. Englewood Cliffs:
Prentice-Hall.
Kerr, B., & Wilcox, E. M. (2004). Designing remail: reinventing the
email client through innovation and integration. CHI 2004, 24–29.
Markus, M. L. (1994). Finding a happy medium: explaining the
negative effects of electronic communication on social life at
work. ACM Transactions on Information Systems, 12(2), 119–
149.
McFarlane, D. C. (2002). Comparison of four primary methods for
coordinating the interruption of people in human-computer
interaction. Human-Computer Interaction, 17(1), 63–139.
Perlow, L. (1999). The time famine: towards a sociology of work time.
Administrative Science Quarterly, 44(1), 57–81.
Peschl, M. E., & Scheutz, M. (2001). Explicating the epistemological
role of simulation in the development of theories of cognition.
Proceedings of the seventh colloquium on Cognitive Science
ICCS-01, 274–280.
Sargent, R. G. (2003). Verification and validation of simulation models.
In: S. Chick, P. J. Sánchez, D. Ferrin & D. J. Morrice (eds.),
Proceedings of 2003 Winter Simulation Conference (pp. 37–48).
Speier, C., Valacich, J., & Vessey, I. (1999). The influence of task
interruption on individual decision-making: an information
overload perspective. Decision Sciences, 30(2), 337–360.
Speier, C., Vessey, I., & Valacich, J. (2003). The effects of
interruptions, task complexity, and information presentation on
computer-supported decision-making performance. Decision Sciences, 34(4), 623–812.
Te’eni, D. (2001). Review: a cognitive-affective model of organizational communication for designing IT. MIS Quarterly, 25(2),
251–312.
Trafton, J. G., Altmann, E. M., Brock, D. P., & Mintz, F. E. (2003).
Preparing to resume an interrupted task: effects of prospective
goal encoding and retrospective rehearsal. International Journal
of Human-Computer Studies, 58, 583–603.
Venolia, G., Dabbish, L., Cadiz, J. J., & Gupta, A. (2001). Supporting email
workflow. Microsoft Research Tech Report MSR-TR-2001-88.
Weber, R. (2004). A grim reaper: the curse of email. MIS Quarterly,
28(3), iii–xiii.
Welch, P. D. (1983). The statistical analysis of simulation results.
In S. S. Lavenberg (Ed.), The computer performance modeling
handbook. NY: Academic.
Whittaker, S., Bellotti, V., & Moody, P. (2005). Introduction to the
special issue on revisiting and reinventing email. HCI Journal,
20(1–2) (forthcoming).
Winsberg, E. (2003). Simulated experiments: methodology for a
virtual world. Philosophy of Science, 70(1), 105–121.
Zijlstra, F. R. H., Roe, R. A., Leonova, A. B., & Krediet, I. (1999).
Temporal factors in mental work: effects of interrupted activities.
Journal of Occupational and Organizational Psychology, 72(2),
163–185.
Inf Syst Front (2011) 13:637–653
Ashish Gupta is Associate Professor in School of Business at
Minnesota State University Moorhead. Ashish has a PhD in
Management Science and Information Systems from Oklahoma State
University. His research interests are in the areas of information
overload, email management, instant messaging, interruptions, Healthcare, simulation modeling. His recent articles appeared in journals
such as Communications of AIS, Information Systems Frontiers,
Annals of Information Systems, etc. He is serving as guest editor of
the Special issue of Decision Support Systems on ‘Modeling for Better
Healthcare’. He serves on the editorial board of IJDSST and IJITSA.
Ramesh Sharda is Director of the Institute for Research in
Information Systems (IRIS) , ConocoPhillips Chair of Management
of Technology, and a Regents Professor of Management Science and
Information Systems in the Spears School of Business at Oklahoma
State University. His research has been published in major journals in
management science and information systems including Management
Science, Information Systems Research, Decision Support Systems,
653
Interfaces, INFORMS Journal on Computing, Computers and Operations Research, and many others. His coauthored text book (Decision
Support and Business Intelligence Systems by Turban/Sharda/Delen,
9th edition, Prentice Hall) has just been released. He serves on the
editorial boards of journals such as the INFORMS Journal on
Computing, Decision Support Systems, Information Systems Frontiers,
and OR/MS Today. Ramesh is also a cofounder of a company that
produces virtual trade fairs, iTradeFair.com.
Robert Greve is an Associate Professor of Information Technology at
the Meinders School of Business at Oklahoma City University. Prior
to his current position, Robert served as a Visiting Professor at
Oklahoma State University-Tulsa. Robert’s research interests include:
email management, information overload, simulation, and the use of
technology in the classroom. Dr. Greve has presented at numerous
conferences and published research concerning how best for organizations and individuals to process email. Dr. Greve is also the
cofounder of LiveClassTech.com.