http://www.400190.com/dissertation_writing/MBA/#p#分頁(yè)標(biāo)題#e#* Corresponding author. Tel.: +41 21 692 3448.
E-mail address: [email protected] (S. de Treville).
0272-6963/$ – see front matter # 2005 Elsevier B.V. All rights reserved.
doi:10.1016/j.jom.2005.03.004process yielded lead-time reduction results that
could save lives. In addition to demonstrating the
clear-cut applicability of a queuing model to an
intergovernmental organization, the more important
contribution of this case study may well be to
illustrate the benefits of combining the transfer of
straightforward operations management approaches
with simple modeling as a means for achieving buyin
and implementation in the not-for-profit context,
where managers are not accustomed to thinking of
themselves as running operations, and where the
emphasis on efficiency and cost reduction makes
lead-time reduction even more difficult than in the
competitive arena faced by for-profit organizations.
The paper is organized as follows: Section 2
provides some brief background concerning TB and
the problems caused by long lead times in the
intergovernmental operations working toward the
control and eradication of TB. In Section 3, we
present our basic process analysis of the GDF
application processing operations, followed by a
discussion of the queuing theory-based model developed
by the team. Results are presented in Section 4.
In Section 5 we suggest implications of the study for
both the WHO and not-for-profit organizations.
2. A life-saving need for lead-time reduction
It is now commonly agreed that TB control and
eventual elimination have shifted from being technical
problems to being managerial and political challenges.
The tragedy of TB is that almost 2,000,000 people die
annually (i.e., one person every 15 s) of a disease that
can usually be cured in 6months with $10worth of antituberculosis
drugs (ATDs, Stop TB website, 2003).
Several decades ago, controlling TB was a top
priority worldwide. The discovery of effective ATDs
led to a substantial reduction in TB in most developed
countries; hence, TB control became a relatively low
priority. The same decline in TB cases did not occur in
less-developed countries, however: By the late 1980s,
it became clear that TB was an urgent problem,
especially when combined with the problems of
interaction with HIV infection and increasing outbreaks
of multiple-drug-resistant strains of TB. Today,
8.5 million people develop TB every year, with 80% of
these cases occurring in 22 ‘‘high burden’’ countries
(WHO, 2003). The WHO and other partner organizations
responded to this crisis by forming the Stop TB
Partnership to mount a global attack on TB, setting
global targets of detecting 70% of people with#p#分頁(yè)標(biāo)題#e#
infectious TB and curing 85% of those detected by
the year 2005 (Raviglione and Pio, 2002).
The internationally recommended primary strategy
for controlling TB is known as DOTS (which
originally stood for ‘‘directly observed treatment,
short course’’). Between 1990 and 2001, the number
of countries that had an appropriate system for
tuberculosis control rose from less than 10 to 155
(WHO, 2003) due to implementation of the DOTS
strategy. It is generally agreed that the DOTS strategy
is the most effective strategy for controlling TB. The
World Bank referred to DOTS as ‘‘one of the most cost
effective strategies available’’ (Stop TB website,
2003). Dr. Gro Harlem Brundtland, then Director-
General of the WHO, referred to DOTS: ‘‘We have a
cure. We need to mobilize the world to use it’’ (Stop
TB website, 2003).
In spite of the effectiveness of the DOTS strategy,
however, the WHO estimated that in 2000, only 27%
of new cases were identified, implying that global
targets for TB control would not be reached until the
year 2013 at the earliest (WHO, 2002). The major
challenge in controlling TB is to ensure that patients
take their medication daily during the 6 months
required for treatment. Efforts to ensure compliance
with treatment have been hindered, however, by lack
of access to low cost ATDs of consistent quality. For
this reason, the Stop TB partners formed the GDF in
early 2001 to fund and manage procurement and
quality assurance for countries applying for assistance.
The vision of the GDF is a TB-free world. Its
mission is to: (a) ensure uninterrupted access to
quality TB drugs for DOTS implementation; (b)
catalyze rapid DOTS expansion in order to achieve
global TB targets; (c) stimulate political and popular
support in countries worldwide for public funding of
TB drug supplies; and (d) secure sustainable global TB
control and eventual elimination (Global Drug Facility
website, 2003).
The GDF began their work with the objective that
lead times from arrival of an application to delivery of
drugs to the port of the applying country would be less
than 6 months (i.e., 132 working days); lead times
398 S. de Treville et al. / Journal of Operations Management 24 (2006) 397–406would be reduced to 3 months (66 working days)
thereafter. However, in the first six countries supplied,
the actual lead times averaged 267 working days, and
the average lead time has since increased to about 400
days (181 days for processing of applications – of
which about 20 days represent actual processing time
– and 219 days from placement of order for drugs to
receipt of drugs in the applicant country). GDF#p#分頁(yè)標(biāo)題#e#
management was uncertain regarding how to meet
lead time targets: It was generally agreed within the
Stop TB organization that the GDF staff members
were working hard and displaying a high level of
motivation.
A chance discussion brought lead-time reduction
theory to the attention of the GDF manager, who then
invited a professor from the Operations Management
department of a local university to work with his
group to explore the implications of applying existing
theory concerning lead-time reduction to GDF
operations. We formed a team composed of the
professor, the GDF manager, and an assistant from the
university who became employed by the GDF to work
on the project. The objective of the GDF lead time
project was to explore the application of lead-time
reduction theory in a not-for-profit organization to a
set of activities thatwould not normally be considered
as an ‘‘operation.’’
3. Applying OM tools to the GDF operations
Reviewing the mathematics of lead time and
carrying out a basic process analysis initiated the
lead-time reduction efforts. However, a straightforward
mathematical tool was required to allow the
group to reach consensus on courses of action,
communicate the action plan to the rest of the Stop
TB unit so as to receive necessary approval and
funding, and transfer the vision for lead-time
reduction to other operations within the WHO.
3.1. Initial analysis of GDF operations
Following the principles of action research, the
project team examined the phenomenon of long lead
times while working to create an action plan to reduce
the lead times. Action research, according to Greenwood
and Levin (1998, p. 75), is ‘‘committed to the
idea that the test of any theory is its capacity to resolve
problems in real life situations.’’ Another objective of
action research is practitioner learning, which results
in an increase in the ability of practitioners to solve
problems.
A key aspect of the project entailed training the
留學(xué)生dissertation網(wǎng)GDF staff both in the mathematical principles that
drive lead times and in building and interpreting
queuing theory-based mathematical models of the
GDF operations, consistent with the objective of
action research that practitioners become more able to
resolve their own problems through learning how to
apply theory. Applying operations management theory
to a real problem also permitted testing and refinement
of the theory in the context of an office operation in a
not-for-profit organization (see Rynes et al., 2001).
As mentioned previously, the actual processing
time for applications was about 20 days, indicating#p#分頁(yè)標(biāo)題#e#
that 161 of the 181 days of lead time represented
waiting time. Our objective was to use OM tools to
discover and make a plan for the elimination of this
waiting.
We began by reviewing the relationships between
bottleneck utilization, batch (lot) sizes, and lead time
(shown in Figs. 1 and 2). These relationships –
fundamental to OM theory – were completely
unfamiliar to the group. As soon as group members
understood these relationships, they immediately
began to identify assumptions and behaviors that
had led to increased lead times. The GDF manager and
S. de Treville et al. / Journal of Operations Management 24 (2006) 397–406 399
Fig. 1. As bottleneck utilization increases, average lead time
increases at an increasing rate, becoming infinite at 100% utilization.from a hope to a goal. Intergovernmental organizations
such as the WHO operate under intense pressure
to increase resource utilization. Getting WHO and
Stop TB management to change from efforts to
maximize resource utilization to being willing to add
extra capacity (decreasing resource utilization) to
reduce lead times was going to take more than a
cursory understanding of the relationships portrayed
in Figs. 1 and 2.
Finally, moving from theory to practice was
hindered by the creativity generated by exposure to
OM concepts. The team had difficulty in knowing how
to evaluate and prioritize the large volume of ideas that
arose from applying operations management theory.
For these reasons, we decided to build a mathematical
model of the application processing operations to aid
in managing and evaluating the ideas generated, as
well as to allow a more thorough analysis of the
process than was permitted by manual process
analysis and commonsense reasoning. Application
processing was chosen as a starting point because the
key parameters were under the control of the GDF.We
chose to use modeling software based on queuing
theory rather than simulation because of its ease of
use. After this project, GDF team members should be
able to construct and interpret queuing theory-based
models of processes, but it is unlikely that they would
have been able to master simulation-based modeling
without extensive training. The software package used
for the modeling was MPX (see Network Dynamics
website, 2003).
3.2. Building the queuing model
The model building began from development of the
process flow diagram shown in Fig. 3. Processing
times were compiled by having a researcher work on
site to interview and observe each member of the GDF
involved in processing applications. As processing
times did not appear to demonstrate high variability,
we assumed a coefficient of variation of processing
times of .3 (low variability according to Hopp and#p#分頁(yè)標(biāo)題#e#
Spearman, 1996, and the default value in the modelbuilding
software).
As demonstrated in Fig. 3, the first step in the
approval process was the screening of applications.
About 70% of applications had to be returned to the
applicant for further information. Once the application
was judged complete, the information from the
application was collected into a summary sheet and
submitted to the TRC. The decisions made by the TRC
were summarized into notifications by the technical
officers. These notifications were approved by the
executive secretary of the Stop TB Partnership (who
had supervisory responsibility for the GDF). TRC
decisions resulted in an average of 7% of applications
being judged incomplete, requiring additional information
from the applicant countries and an additional
TRC round 4 months later.
Country visits were required for the first grant
awarded to each country. Once the TRC judged an
application coming from a first-time country complete,
a country visit was planned by the technical
officers, approved by the Stop TB executive secretary,
and carried out by technical officers, the GDF
manager, or a consultant appointed to the GDF.
Country visits were arranged for approximately 71%
of applications evaluated by the TRC. Applications
judged complete from countries that already had been
visited in connection with a previous application
(approximately 22% of applications evaluated by the
TRC) proceeded directly to calculation of drug needs.
After the country visit, the GDF staff member
prepared a report. In 30% of cases, the report indicated
a need for additional information and an additional
round at the TRC. Once the country visit report
indicated that the file was complete, drug needs were
calculated. Finally, the grant was prepared and the
application entered the drug-sourcing process.
Following the development of the process flow
diagram, the next challenge was to set the lot sizes and
batching policies. Given that the TRC met three times
per year, our initial inclination was to set the lot size at
one-third of the twenty applications processed each
year. It was, however, necessary to increase the lot
sizes to reflect the 1.40 TRC rounds made by an
average application; hence, we used an average lot
size of 20/3
1.40 = 9.3 applications.
The MPX software allows the user to set a transfer
batch size that is different from the production lot size,
indicating that parts in a given batch are assumed to
move to the next operation without waiting for the full
batch to be completed at that operation.2 A production
402 S. de Treville et al. / Journal of Operations Management 24 (2006) 397–406
2 Our terminology here matches that in the software: lot size for#p#分頁(yè)標(biāo)題#e#
production, batch size for transfer.lot size of ten units combined with a transfer batch of
one unit would imply that the lot of ten pieces arrives
together at the first workstation, with each piece
moving to the next workstation as soon as it is
processed, with the ten pieces reassembled into the
original lot of ten after processing at the last
workstation. The team compared this MPX logic to
the GDF application process.
In the GDF application process, applications
tended to arrive just before the TRC meetings. There
was no intentional batching, but the TRC deadlines
caused applications to be batched rather than arrive
uniformly throughout the year. Applications entered
the TRC meeting as a batch: In fact, we modeled the
TRC meeting as a ‘‘setup time’’ because the time
during which the meeting took place (5 days) was
independent of the number of applications. The batch
of applications left the TRC meeting together and
tended to be batched at several of the operations
following the TRC meeting. As an example, the
executive secretary responsible for approving notifications
and country visits usually waited to process
applications until he had the entire batch on his desk.3
The data used to construct the base case of the
model are given in Tables 1 and 2.
4. Results
The application of the model yielded several
potential improvements that could reduce the drug
approval process lead time from more than 180, to
less than 40, working days. According to the base
case of the queuing model, lead times for processing
applications averaged 183 days, compared to the
actual lead time of 181 days. The model highlighted
three main causes of long lead times: ‘‘batching’’ of
applications, utilization levels for the technical
officers handling applications, and the fact that
applications made an average of 1.4 visits through
the TRC process. Using the model, we were able to
demonstrate that batching caused by the 4-month gap
between meetings (‘‘time waiting for rest of lot’’)
was the primary cause of long lead times in
processing applications, accounting for approximately
113 of the 183 days of lead time predicted
by the model. Waiting for labor (i.e., an application
has arrived and is ready to be processed as soon as
the relevant person arrives) accounted for 43 days of
lead times. The 5-day TRC meeting itself caused an
average of 7 days lead time, given the 1.4 TRC
rounds required for an average application. Finally,
run time (i.e., the time during which the application
was actually being processed) accounted for about 20
days, as expected.
Lead times predicted by the model are given in
Table 3.We tested six scenarios in addition to the base#p#分頁(yè)標(biāo)題#e#
case:
In the first scenario, we added a technical officer on
the demand side, which reduced the time waiting for
labor from 43 to 32 days and also reduced the time
waiting for the rest of the lot by 6 days. The total
impact of adding a technical officer on the demand
side was a reduction of about 17 days.
In the second scenario, we evaluated the impact of
reducing the average number of TRC rounds per
application. As can be seen from Fig. 3, applications
were returned to the applicant countries at two points
in the process: immediately after the TRC meeting
(7%) and after the country visit (30%). These returns
were due to incomplete information, implying that the
S. de Treville et al. / Journal of Operations Management 24 (2006) 397–406 403
Table 1
Model data (MPX format)
Time units
Operations Hours
Flow time Days
Production period Year
Hours per day 8
Days per year 210
Labor
Overtime (%) Unavailable (%)
1 Applicant 0 0
1 Consultant 0 0
1 Manager, GDF 50 55
1 Executive secretary, STB 0 70
2 Technical officers, demand 15 50
1 Technical officer, supply 15 50
Equipment for all employees modeled as having unlimited capacity
Demand/year Lot size
Product
Approved application 20 9.3
3 The instinctive tendency of people throughout the organization
to batch applications supports our claim that instruction in lead time
reduction principles should not be limited to those officially working
in operations.Combining the above two actions, we modeled
a scenario in which an extra technical officer was
added on the demand side and repeat visits to the TRC
after the country visit were reduced from 30% to 10%
of applications. In this scenario, lead times were
reduced to 135 days, representing a 30-day improvement
compared to only adding a technical officer, and
a 12-day improvement compared to reducing the
repeat visits to the TRC without the addition of the
technical officer.
The next scenario considered a more extensive
process redesign, in which the TRC was reconfigured
to meet ‘‘virtually’’ to process each application, with
each virtual meeting taking 16 h of technical officer
time. Under this scenario, the lead times for
application processing dropped from 183 to about
64 days, that is, a reduction of 119 days. Run time was
predicted to increase to 22 days, as the TRC evaluation
was now modeled as a per unit processing time rather
than as a setup time for a batch of applications.
Waiting for labor was predicted to decline slightly
from the base case, dropping from 43 to 41 days. The
113 days of time spent waiting for the rest of the lot
and the 7 days spent in the TRC meeting were#p#分頁(yè)標(biāo)題#e#
eliminated.
Adding an extra technical officer to the virtual TRC
meeting scenario reduced lead times still farther, to 45
days. Finally, combining virtual TRC meetings, an
extra technical officer, and reduced repeat TRC visits
resulted in a lead time of 39 working days.
The team was surprised by how much was learned
from the modeling exercise. Our initial expectation
was that modeling would provide limited value over
manual process analysis. Not only were we able to
quantify the impact of rework and high utilizations
easily, but also the impact of the TRC batching was far
beyond the 79 days of waiting time that we had
expected. Our commonsense reasoning substantially
underestimated the amount of waiting time caused by
the applications exiting the TRC meeting as a batch. It
was only in interpreting the model results that we
realized the extent of the batching-related waiting that
occurred at operations such as preparation and
approval of country visits. Finally, the modeling
allowed us to capture the interaction effects between
scenarios. We found it much simpler to discuss,
evaluate, and communicate action plans. What is
more, the level of confidence in the team in OM
analysis increased tremendously once we could
explain exactly why lead times were so long.
5. Implications for the WHO and other not-forprofit
organizations
The next step in the project was to present the
model to the entire Stop TB department to illustrate
the potential for lead-time reduction. In presenting the
model, we explained that achieving model results
consistent with observed lead times required the
addition of 15% overtime for the technical officers and
50% overtime for the director of the GDF, which we
then observed to be consistent with practice. A minor –
but most appreciated – outcome of the modeling
exercise was the opportunity it provided for Stop TB
and GDF management to explicitly acknowledge the
regular overtime worked by GDF staff members.
Based on the model results, Stop TB management
made the decisions to: (a) add a technical officer,
especially since the workload of the technical officers
was expected to increase over time; (b) formulate a
plan to increase the frequency of Technical Review
Committee meetings—eventually to virtual meetings;
and (c) increase commitment to reduce process
variability through standardization and documentation,
including a decision to seek ISO 9000 series
certification. Based on these changes, the lead time for
application processing is in the process of being
reduced from 6 months to the target of less than 2
months.
The doctors and global health care specialists
involved in this project were unaccustomed to#p#分頁(yè)標(biāo)題#e#
considering themselves as ‘‘operations managers.’’
Consistent with Suri’s (1994; 1998) empirical results,
these managers were unaware of the mathematical
principles that drive lead times and were therefore
taking actions and making decisions that increased
lead times. It was essential during the project to keep
the message and the underlying OM logic as simple
and straightforward as possible to ensure maximum
buy-in and impact. When these doctors and specialists
understood the mathematics of lead time, their
behavior began to change.
We expect that similar behavioral changes would
result from this transfer of knowledge in other not-forprofit
organizations. To the extent that our results
S. de Treville et al. / Journal of Operations Management 24 (2006) 397–406 405generalize to other not-for-profit organizations, it may
be worthwhile exploring ways to transfer knowledge
concerning these OMand basic modeling principles to
all managers, whether or not they intend to ‘‘manage
operations.’’ Finally, the positive response of these
doctors and specialists – as well as other doctors who
saw the results – to mathematical modeling of
operations indicates a need for research into expanding
the use of easily accessible modeling software
packages such as MPX.
In the for-profit arena, access to the knowledge and
tools required for managing lead times can easily
determine who wins and who loses competitively, with
market share, profit, and investments in capacity and
inventory at stake. In organizations such as the GDF,
however, such knowledge and tools translate into lives
saved. How many of those people currently dying
needlessly of TB every 15 s will stay alive given
timely access to ATDs?
Acknowledgements
The authors would like to gratefully acknowledge
the insightful comments of Bob Hayes, Roger
Schmenner, Greg Diehl, and Guest Editor Bill
Youngdahl.
References
Global Drug Facility website, 2003. Accessed December 21, 2003
from http://www.stoptb.org/GDF/whatis/whatis.html.
Greenwood, D.J., Levin, M., 1998. Introduction to Action Research:
Social Research for Social Change. Sage, Thousand Oaks, CA.
Hopp, W.J., Spearman, M.L., 1996. Factory Physics, first ed. Irwin
McGraw-Hill, Boston.
Network Dynamics website, 2003. Accessed March 6, 2003 from
http://www.networkdyn.com/RPMframe.html.
Raviglione, M.C., Pio, A., 2002. Evolution of WHO policies for
tuberculosis control, 1948–2001. The Lancet 359, 775–
780.
Rynes, S.L., Bartunek, J.M., Daft, R.L., 2001. Across the great
divide: Knowledge creation and transfer between practitioners
and academics. Academy of Management Journal 44 (2), 340–#p#分頁(yè)標(biāo)題#e#
355.
Stop TB website, 2003. Accessed December 21 from http://
www.stoptb.org/tuberculosis/default.asp.
Suri, R., 1994. Common misconceptions and blunders in
implementing quick response manufacturing. In: Proceedings
of the SME AUTOFACT ’94 Conference, Detroit, MI,
November.
Suri, R., 1998. Quick Response Manufacturing. Productivity Press,
Portland, OR.
World Health Organization, 2002. Global Tuberculosis Control:
留學(xué)生MBAdissertationSurveillance, Planning, Financing. WHO Report 2002. Geneva,
Switzerland, WHO/CDS/TB/2002.295.
World Health Organization, 2003. Global Tuberculosis Control:
Surveillance, Planning, Financing. WHO Report 2003. Geneva,
Switzerland, WHO/CDS/TB/2003.295.
406 S. de Treville et al. / Journal of Operations Management 24 (2006) 397–406