Data Reenginering Case Study

Project Decision Analysis Context: Constructing and Linking the Uncertain Events

by Bill Girling

A. Legislative Support (EXHIBIT #1)


The Legislative Support event is defined by the following attributes:

Measurement: "Legislative Support" must be greater than "No Support": E(LS) >E(NS)

1. Probability Assessment - Economy


Fed rate strategy defended, "Now is the time to resist inflation, president of Richmond bank insists" Mary Beausoleil. Pg. C10 . Richmond Times Dispatch April 26, 1997.

According to Alfred Broaddus, Jr., president of the 5th District Federal Reserve Bank, "The current economic situation is about as good as it gets." "The economy is into its seventh year of expansion" "The consensus opinion for 1997 among forecasters, of which Broaddus is not one, is for more of the same: " However, others (64 lawmakers) were "nervous that efforts to slow the economy could upset the political landscape.

Likelihood: The state of the economy will affect the support for the project.

Good Economy: 50%

Bad Economy: 50%

2. Probability Assessment - Scandal


Within the political environment, the adversarial approach promotes the elevation of bureaucratic missteps to "scandalous" proportions. In each gubernatorial term, it is virtually certain that a major incident will make headlines and the merits of each case will be vigorously debated. While the ultimate impact is unpredictable, the visibility generated usually creates delays of a year or more for all new programs involving expenditures of more than one $ million. An examination over the past 18 years of Virginia state government reveals a "show stopping" incident about once every five years. Recent events (Exhibit #2) resulted in headlines that have a direct impact on this project. In 1995, the Department of Social Services became embroiled in controversy over claims of major cost overruns. The increased sensitivity of administration officials and members of the legislature resulted in delays and the real possibility of cancellation of the project.

Likelihood: There will be a negative impact on supporting the project as a result of a scandal.

Scandal 20%

Scandal Free: 80%

3. Probability Assessment - Shift in majority Legislative Party


This is a significant issue because the Senate sub-committee and the staffs that have purview over the project are very supportive. Changes in majority control create an automatic re-examination of priorities. It is not unusual for some programs to be dismantled because of their affiliation with the "opposing" party. During the three critical years of project implementation, there will be one election where all members of the House of Delegate and the Senate will be up for election. Political analysts, Dr. Larry Sabato and Dr. Robert Holsworth agree that each of the past two elections of the Legislature have been too close to call more than 60 days preceding these statewide elections. However, baring major events such as a shift in the economy, a Presidential election, war, etc. the party in power tends to stay in power. Since the current distribution is so closely balanced, it is a reasonable reflection of the probability that the party structure stay "as is" and not shift during the life of the project.

Likelihood: There will be a negative impact on supporting the project as a result of the Republican Party gaining control of the House or Senate.

Democratic majority: 51%

Republican majority: 49%

4. Attitude Toward Risk (Utility Ratings)

Note: The upper value of 1 is used in all Utility ratings. There are multiple instances of risk assessment which ultimately contribute to the final decision analysis and the utilization of comparable values is required.

The attitude toward risk is reflected in a bi-polar comparison for each of the three key elements: "good economy" vs. "bad economy"; Democrats vs. Republican; and "Scandal" vs. "Sandal Free".

a. Economy: The rating of 1 for "good economy" and 0 for "bad economy" is intuitively appropriate for this event. Funding is provided on a "biennial" basis and this project transcends several budget cycles. The legislative attitude about funding is directly related to the state's economic status.

b. Political Party: The risk attitude associated with the political party in the majority, reflects the assessment that the political party currently in the majority (Democrats) supported the original project concept and continues to support it after several challenges. If the Republicans gain control of the legislature, there will be a negative impact, simply because a change has occurred and all existing programs are subject to scrutiny.

c. Scandal: Members of the legislature are keenly tuned to public perception and rapidly distant themselves from the slightest hint of scandal. A scandal free environment is of high value to the key decision makers who seek to minimize the high risk situation.

5. Multiple Attributes

The relationships between multiple attributes are dealt with by means of a "weighting" scheme that establishes a relative value between the attributes. In the "Legislative Support" event, the weighting mechanism has become critical in displaying and adjusting to the sensitivity of legislative support to volatile issues such as public perception. The assigned "weighting" values were based on the judgment of subject experts who determined that the least significant factor was the "Scandal" attribute and it was given a value of 25. The "Economy was determined be three times important and was given a value of 75. The majority legislative "Party" was judged to be slightly more important and it was given a value of 100. This proved to be an "initial" assessment. As the project unfolded, several events unfolded which changed these subjective assessments dramatically. More about this specific subject is addressed in the summary.

B. Budget Analysis

Measurement: The project budget can not exceed $12 million and the key elements, consulting and software cost cannot exceed $4 million combined.

The Budget Analysis event is defined by the following attributes:

1. Present Value

A present value calculation (Exhibit #4) is prepared for the five elements that comprise the four year project budget. Adjustments are made to reflect elements that will be expended on a monthly or annual basis. The projected expenditure "range" was generated through several different devices. The range for the "Consulting Fees" were estimations provided by the only viable vendor. "Software" "Platform", and "Training" fees were determined by knowledgeable procurement experts estimating the degree to which bid proposals could be negotiated. Maintenance fees are a function (percentage of contract) of the software and platform costs.

2. Cost Risk Analysis (Exhibit #5.1, 5.2, & 5.3)

The purpose of this analysis was to determine if any specific cost area(s) could be considered "high risk". The extended Pearson-Tukey method was use to represent the probability distribution . It is appropriate for it to be used in the assessment of "subjective probability" where range assessments have already been made. As displayed in the exhibit, the standard deviation for "Consulting" and "Software" displayed an extremely high range of variability. These two elements were then subjected to further analysis as described in the next section.

3. High Risk Expenditures (Exhibit #6)

In the assessment of this area, consideration was given to the probability assessment techniques, attitude toward risk and the existence of multiple attributes. In a continuation of earlier steps, the extended Pearson-Tukey method was use to represent the probability distribution. In this project, funds could be applied where needed and limits were defined by the total budget available. Therefore, the probability for the combined distributions had to be defined. Since software and consulting services are independent events it is appropriate to create the nine different "pairing" possibilities as defined in Exhibit #6a. The respective probabilities were multiplied together to generate the "combined" probability.

The handling of multiple attributes was addressed by calculating (Exhibit #6) the proportional cost for each element in relationship to the total cost for the consulting and software services cost.

Another area of concern is the sensitivity to the "Budget" factor, of the primary decision to proceed (Go) or delay (No Go) the project. Exhibits #6c and #6d display the use of a continuous distribution function, that provides for the determination of a variety of discrete probabilities associated with a specific cost figure. Budget sensitivity can be assessed as the funding picture changes. The ramifications are discussed in the section on "sensitivity" as well as the section on the "Primary Decision Analysis" (Section E).

C. Deployment Decision (Exhibit #7)

Measurement: Organizational Deployment has greater value than Functional Deployment.

The selection of a "deployment" strategy is a critical issue. It will satisfy one of the requirements for providing a "safeguard", if the project should falter. An "organizational" deployment provides for developing all of the functional capabilities of the project and implementing these functions simultaneously in one organization at a time. A "functional" deployment staggers the development and installation of the three key applications, human resources, benefits and payroll. If supported by the availability of funding and parallel system maintenance facilities, an "organizational" deployment will allow the project to slow down or halt (without losing the current investment), if it is no longer technically, financially or politically viable.

The comparison provided in Exhibit #7, quantifies the relative value of the important elements. A weight, developed with the "swing weighting" method, was assigned to each element in order to establish its relative value in comparison to the other elements. A utility function was used to determine the value that experienced users felt was generated when a choice was made between the deployment strategies. For example, "knowledge transfer" gains more value through "organization" deployment rather than "functional" deployment. This preference assessment method indicates that an "organizational" deployment strategy has a greater value.

D. Primary Decision Analysis

The Primary Decision required, "Does the Commonwealth of Virginia proceed with or stop the project", is exercised in Exhibit #8. This decision tree reflects the assimilation of three uncertain events that individually, are comprised of multiple attributes. Attributes developed earlier in the analysis of the "uncertain events", were used to populate several elements the decision tree. However, as independent entities, each of the uncertain events has its own probability, utility, and weighting characteristics.

1. Probability Assessment

a. Legislative Support:

An assessment on the probability of maintaining legislative support for the project reflects the mean of the appropriate event attribute probabilities i.e. "Good Economy", "Democrats retain majority", "Scandal Free". It is assumed that these are independent events of equal weight.

(P(Good Economy) + P(Democrats) + P(Scandal Free))/3 = (.5 + .51 + .8)/3 = .60

b. Within Budget:

The analysis of High Risk Expenditures (Exhibit #6) resulted in the development of a continuous distribution function. Through this mechanism, it is possible to project the probability that the "controlling" cost factors, i.e. consulting and software expenses, will be within or over budget. The project budget of $4 million for these two elements has a 66% probability of being within budget.

c. Deployment:

The probability of success was established by utilizing the results of the 1996 PeopleSoft Implementation Report. This major software vendor (also a finalist for the project) surveyed 259 clients as to their preferred method for system deployment. The successful installation of the application can be directly related to the selection of a preferred deployment approach.

The distribution is as follows: Organizational Deployment - 45%; Functional Deployment- 55%. This figures are plugged into the final the project Decision Tree.

2. Attitude Toward Risk (Utility Ratings)

The attitude toward risk is reflected in a bi-polar comparison for each of the three key elements:

"Legislative Support"; "Within Budget"; "Operational Deployment". The upper rating of 1 for all of these elements and the assignment of zero to their inverse is intuitively appropriate for this event.

3. Multiple Attributes

As in the preceding events, relationships between multiple attributes are dealt with by means of a "weighting" scheme that establishes a relative value between the attributes. The values assigned to each of the three decision elements was "swing" weighted to reflect their relative value. A variety of different arguments were considered in assigning weights. Generally, negative attributes were given higher value because they could more substantially contribute to the "failure" of the project, whereby "positive" attributes would not necessarily contribute as much to success. The relationship to the subordinate "uncertain events" is straight forward for the Budget and Deployment attributes and derived directly from the analysis that occurred at the subordinate level. Integrating the "Legislative Support" event into the decision structure is difficult. Earlier, this event was determined to have a 60% probability of occurring. Conversely, "No Support" has a 40% chance of occurring. For this event, it important to remember the measurement criteria is that success ("GO") is defined when Legislative Support (LS) is greater than or equal to "No Support" (NS). These conditions are reflected in the values assigned to the "Weight" values.

Navigate from here: returning to the project project home page, or jump to another part of the decision analysis case study (Situation Description, Possible Decision Outcomes, Decision Analysis, Constructing & Linking The Uncertain Events, Decision, Summary) - or jump to Peter Aiken's home page, download some reengineering articles, or access other reengineering links.

This page and all web site contents were last updated and are copyright 10/24/04 by Peter Aiken - all rights reserved.