Somtimes Less is More - Using DPL in Check Risk Analysis for ACE
Michael Monticino used DPL to perform risk analysis on the different types
of checks that ACE cashes. A typical summary of a DPL-based analysis project
describes the intense brainstorming sessions involved in developing the decision
model, the in-depth statistical analysis performed to estimate uncertainties,
the extensive sensitivity analysis, and the ultimate cost savings obtained through
the enlightened decision analysis process. This project summary is a little
different. Cost savings resulted, but they were savings from not doing
an in-depth analysis. One of the often-overlooked uses of the decision analysis
process is that it can help determine the value of performing costly in-depth
analysis. In particular, DPL can be used to determine under what ranges of conditions
would changing a company's current procedures be cost effective. If these conditions
turn out to be unlikely, then spending the time and money on further analysis
would not be worthwhile.
In a recent consulting project for America's Cash Express Inc. (ACE), one of
the largest check cashing companies in the U.S., I was asked to perform risk
analysis on the different types of checks that ACE cashes. The idea was to correlate
check characteristics with frequency of returns and identify those check types
for which ACE tellers should spend the most effort verifying authenticity. Check
characteristics include dollar amount, whether the check is a company, personal
or government check, whether it is hand-written or typed, the date the check
was written, whether the person cashing the check is a regular or new customer,
and whether a regular customer is trying to cash a check on a day that they
do not regularly visit the store. There are several actions that an ACE teller
can take to verify that a check is "good." They can just ask for a picture ID
and verify that the person on the ID, the person the check is made out to, and
the person cashing the check are the same. Besides verifying ID, they can also
call the bank to make sure that the check is not stolen and that there are sufficient
funds to cover the check. Several more verification policies are available,
each more stringent than the last and each costing more in teller time. The
time it takes to cash a check translates into money in a number of ways, including
teller pay, the number of checks which can be cashed in a day, and lost business
from frustrated customers waiting in line.
At first glance, the risk analysis portion of the work appeared to be a classic
multivariable statistics problem. In order to get an idea of the data requirements,
I decided to use DPL to develop a decision model of the major components in
the problem. After discussions with Donald Neustadt, ACE's CEO, we decided that
the overall objective was to maximise profit from cashing checks. The most important
factors affecting this objective are shown in the influence diagram. The alternatives
available were the different check verification policies. Working with Mr. Neustadt
and Wanda Strong, ACE's vice President for Cash Management, I was able to specify
nominal values for most of the model components -- one notable exception was
the cost of trying to collect on returned checks. My intention was to see how
sensitive the optimal check verification policy was to the return rate for the
various check types. This would give an indication of how accurate I would need
to be in estimating return rates (how small the confidence intervals would need
to be) and hence how much data I would need for the analysis.
Influence Diagram for Check Risk -- Verification Policy
Model
Then came the surprising part, for every check type considered, either the
return rate would have to be far higher than possible (Don Neustadt commented
"...if our return rates were that high we would be out of business.") for a
change from the current verification level to be optimal, or unknown values
(like collections costs) prevented a complete analysis.
So DPL allowed us to show that in the majority of the cases, ACE is employing
the correct verification policy and in a small number of other cases an extensive
cost centre analysis would have to be performed to make an accurate assessment.
DPL allowed us to do this relatively quickly and inexpensively, and it saved
ACE from performing a potentially expensive risk analysis project.
Top of page
R&D Funds Allocation
John Palmer used a DPL model to help General Motors allocate funds to R&D
projects. GM was evaluating two competing technologies in their engine development
program. One technology utilised a new type of coating (Project A), while the
second technology incorporated a new material (Project B). Project A had been
going on for some time when Project B was identified. While they could be researched
simultaneously, the two technologies offered the same benefits and could not
be implemented together. The decision alternatives were therefore:
1. Continue with Project A only
2. Stop Project A and start Project B
3. Do both
4. Cease Project A and not do Project B
GM was pleased with the results of the DPL analysis, and the resulting decision
has worked out well.
Top of page
Paediatric Decision Making
A group at the Simon Graduate School of Business Administration
under the direction of Dr. Leslie Marx recently used DPL to evaluate the cost-effective
use of muscle relaxants in paediatric care units. Children with severe respiratory
disease are unable to breathe without a tube in the trachea enabling the use
of a respirator. Children may interfere with the respirator, thus putting their
lives in further danger. In certain cases, muscle relaxants must be administered
to temporarily paralyse the child, allowing the doctors to administer treatment.
The goal of the project was to find the most cost effective drug and
delivery mode for insuring that patients do not interfere with their treatments.
The study evaluated five muscle relaxant drugs and two delivery modes currently
in use. Each of these five drugs and two delivery modes are equally efficacious
in their ability to stop muscle movement. Each of the drugs and delivery modes
were evaluated across five cost criteria: direct drug, nursing, equipment, residual
care, and pain and suffering (personal cost).
The DPL analysis revealed that intermittently delivered doxacurium is the most
cost effective drug and delivery mode for muscle relaxation. A sensitivity analysis
revealed that number of days paralysed and patient weight were the only two
factors that would lead to a change in treatment. This recommendation represents
a major change from current practice.
Top of page
Selecting a Site Remediation Strategy
Anthony Apatyi, of the Air Force Institute of Technology, recently
worked on a project involving the selection of a remediation strategy for a
contaminated waste site. The polluted area developed when a large pipe carrying
toxic chemicals developed a leak approximately 20 years ago. The leak went undetected
until just recently. When the leak was detected, the site was assessed to be
very polluted, and a decision regarding the best strategy for cleanup was necessary.
To model the decision process, the decision analysts met with the site owners
and technical experts. They identified six different tactical pieces that can
be bundled into strategies. DPL calculated cost, time, and performance attributes
for each strategy, using a spreadsheet. Once final cost, time, and performance
values were known for each strategy, utility values were calculated using utility
functions and weights derived from the site owners and the key decision makers.
The site owners plan to use the results as part of their Remedial Investigation/Feasibility
Study Report to the EPA.
One interesting task was modelling the decision-making process for the site
owners. By using an influence diagram, the team was able to step logically through
the key decisions, values, and uncertainties that are encountered when making
a decision of this magnitude.
The number of attributes to model added complexity to the project. The site
owners are required to evaluate potential strategies on nine separate criteria,
three of which are cost, time, and performance. The remaining six were not modelled
in this effort, due to a lack of sufficient data. Nonetheless, the team gained
a great deal of insight by modelling just these three attributes.
Top of page
Plutonium Problems
Dr. James Dyer and John Butler at the Graduate
School of Business at the University of Texas recently used DPL to help the
Department of Energy prepare a policy concerning the disposal of surplus weapons-grade
plutonium.
A significant challenge arising from the end of the Cold War is the need for
safe, secure and verifiable management of weapons-usable highly enriched uranium
and plutonium from the disassembly of nuclear weapons. Global stockpiles of
these materials pose a danger to national and international security if they
are not managed and disposed of in a manner that precludes their reuse in weapons.
DOE has announced a strategy for managing these materials by reducing
the number of locations where they are stored and by pursuing a dual-track plutonium
disposition strategy that allows for immobilising plutonium in glass or ceramic
forms and burning plutonium as mixed oxide fuel in existing reactors.
Technical, institutional and cost uncertainties exist with both the immobilization
and reactor options. Dyer and Butler were on a decision analysis team helping
DOE examine these options (the other members of the team were Thomas Edwards
of Lawrence Livermore National Laboratories and Jianmin Jia of Chinese University
of Hong Kong). They were able to demonstrate that a hybrid approach was superior
if the probability of a key event was within the range of 0.16 to 0.81.
The power of DPL and Logical Decisions helped verify intuition and graphically display the problem.
As a result of the analysis, DOE will complete the necessary tests, process development,
technology demonstrations, site-specific environmental reviews and detailed cost proposals for both approaches.
Top of page
Water Resource Planning
Jane Ratchye, a Resource Planning Engineer, and a team of planners from the
City of Palo Alto Utilities used DPL and the decision analysis framework to analyse
water resource options for the city of Palo Alto, CA, over the next 20 years,
and specifically to determine an appropriate long-term contractual commitment
with the San Francisco Water Department (SFWD). Alternatives included conservation
measures, obtaining water from another water supplier, refurbishing Palo Alto's
groundwater wells, and a major wastewater-reclamation project. The project used
a systematic decision analysis process to determine critical variables, screen
the initial set of alternatives, incorporate uncertainty, and analyse the most
promising alternatives. Decision analysis also provided a framework for enabling
and encouraging the participation of a citizen advisory panel in each stage
of the analysis. The analysis provided guidelines for the level of commitment
to the SFWD and concluded that the value of the wastewater reclamation project
was very low compared to its cost.
Top of page
Going to Mars
Ryan McCorvie of the Jet Propulsion Labs (JPL) in Pasadena used DPL to analyse the probability
of mission success. Since recent discoveries have renewed interest in the exploration
of Mars, the JPL is currently designing a Mars Sample Return mission in which
unmanned rovers will collect rocks from the surface of Mars and return to Earth
for study. Initial back-of-the-envelope calculations suggested that the probability
that a rover on the surface of Mars could meet up with the return vehicle was
about 0.5 _ which is not a very good outlook. The Engineering Economics and
Costing group of the Mission and Systems Architecture Section has set out (with
models of the terrain of Mars, the mobility of martian rovers, the reliability
of these rovers and the landing footprint of the return vehicles) to determine
exactly how feasible such a mission is. DPL is used to combine all these variables
and calculate the overall probability of success.
Top of page
Budgeting for Pollution Prevention Projects
Charlotte Hudson and Rich Houghton of the Air Force Institute of
Technology developed a user-friendly DPL model for helping environmental managers
with limited budgets prepare funding strategies for pollution prevention projects.
Decision analysis is the basis for a capital budgeting process that takes into
account several of the key budgeting issues facing environmental managers, including
combining environmental projects with other business projects in the budget
evaluation process, and considering the costs, savings and general environmental
benefits associated with pollution prevention projects. An important component
of the process is the ability to use sensitivity analysis to understand the
budget implications of changing project economic and environmental factors.
Top of page
|
