4. A comparative study of the CCPM excluding bad human
4.2 Project execution—CCPM vs. PERT
Project execution is designed to evaluate the mean project time and plan reliability of both CCPM and PERT methods. Our execution tool is a simulation model of PMsim developed by Goldratt (1997). Each simulation is replicated 1000 times. The computer randomly generates task duration time for each task based on the task time distribution shown in Figure 4.2. Data collected are mean project duration, its standard deviation, medium, and the 90th percentile. Bad human behaviors such as bad-multi-task, student syndrome, and Parkinson‘s Law, do not exist.
Single project simulation
The CCPM plans for the non-critical chain path to start as late as possible, therefore the simulation was designed to start the first task of each path no earlier than its planned start time even if it can be started early (as late as possible, ALAP). The PERT method simulation was designed in two ways, One way starts the first task of each path immediately when it can be started (we call it PERT-SP-AEAP). Similar to CCPM, the other way starts the first task of each path no earlier than its planned start time even if it can be started early (we call it PERT-SP-ALAP).
Table 4.3 summarizes the results of our single project simulation. From the statistical hypothesis test of the population mean by the student t-test, no matter whether the uncertainty is low, medium, or high, the data show that the CCPM achieved significantly better mean project time than PERT-SP-ALAP did. However, from the statistical hypothesis test of the
population mean using the student t-test, no matter whether the uncertainty is low, medium, or high, the data show that the CCPM is not significantly better than the PERT-SP-AEAP in achieving mean project time. Concerning planned reliability, CCPM achieved higher reliability than both PERT-SP-AEAP and PERT-SP-ALAP did.
Table 4.3 Simulation results of single project
N=1000
Uncertainty Low Uncertainty Medium Uncertainty High
PERT-
The CCPM plan method adds a synchronization buffer to prevent releasing projects too early (does not encourage starting a project early even if it can be started), therefore, the simulation was designed according to the scheduling rule, in which the first task of each project path starts only at the planned start time, even if it can be started early (ALAP). For the PERT method, the schedule rule within every project will be as early as possible (Table 4.3 shows the PERT-SP-AEAP achieved a better result). However, the scheduling rule among projects was designed in two ways. One is the same as the CCPM (we call it PERT-MP-ALAP). The other is that except for the tasks of B1-B, G1-R, and H1-P, where the first project will start at the planned start time, the rest of tasks of all projects will be started as soon as possible (we call it PERT-MP-AEAP).
Table 4.4 summarizes the results of our multi-project simulation. From the statistical hypothesis test of the population mean by the student t-test, no matter whether the uncertainty
is low, medium, or high, the data show that the CCPM does not perform significantly better than PERT-MP-ALAP does. However, the statistical hypothesis test of the population mean by the student t-test shows that no matter whether uncertainty is low, medium, or high, the data show that the PERT-MP-AEAP achieves significantly better mean project duration than CCPM does, in terms of projects B and C. Concerning plan reliability, CCPM demonstrates higher reliability than PERT does. The higher the uncertainty, the better the planned result of CCPM is.
Table 4.4 Simulation results of Multi-project
N=1000
Uncertainty Low Uncertainty Medium Uncertainty High
Project A Project B Project C Project A Project B Project C Project A Project B Project C
Results finding
The project plan and execution results show that if excluding bad human behaviors, we can draw several findings as follows:
1. No matter for a single project plan or a multi-project plan, with 90% confidence level, the CCPM plan is much more conservative (longer project time and longer project completion date) than the PERT plan. The higher uncertainty, the more conservative it is.
2. For single project execution, no matter whether the uncertainty is low, medium, or high, the results show that the CCPM is not significantly better than the PERT-SP-AEAP in achieving mean project duration. For multi-project execution, no matter whether the uncertainty is low, medium, or high, the results show that the PERT-MP-AEAP significantly achieves better mean project duration than CCPM does in terms of projects B and C.
3. Although from the mean project time result, CCPM is no better than PERT, however, from plan reliability, no matter whether uncertainty is low, medium, or high, the simulation result shows that CCPM achieves higher reliability. This means that using the Equation (3) to estimate the project duration time and not adding a synchronization time buffer to the schedule of the highest loaded resource among projects such as CCPM did, PERT allows for too short a project duration time and too tight a completion date. The higher uncertainty, the worse the result will be.
4. Realistically, few project practitioners will use Equation (3) to estimate task time and project time. They typically take the 90th percentile of task distribution of Figure 4.2 as the task time (CPM (Critical Path Method) typically takes the 90th percentile of task distribution as the task time). Table 4.5 illustrates the plan results using this procedure and re-planning the project with the PERT plan method. Comparing the CCPM and PERT plan with the project time estimate of equation (3) yields a much longer project time and longer project completion date. Comparing the planned results with the simulation results of Tables 4.3 and 4.4, no matter whether uncertainty is low, medium, or high, projects can be completed with nearly 100% reliability. This means that directly taking the 90th percentile of task distribution of Figure 4.2 as the task time, the PERT plan will result in too conservative a plan, making it less competitive.
Table 4.5 Plan results
Uncertainty Low Uncertainty Medium Uncertainty High
Project A Project B Project C Project A Project B Project C Project A Project B Project C
CPM PERT CCPM CPM PERT CCPM CPM PERT CCPM CPM PERT CCPM CPM PERT CCPM CPM PERT CCPM CPM PERT CCPM CPM PERT CCPM CPM PERT CCPM
Estimated Project
time
119 87 90 187 137 158 221 162 192 133 97 100 209 153 176 247 181 214 182 114 137 286 179 241 338 212 293
Reliability 100% 80% 89% 100% 68% 86% 100% 70% 80% 100% 84% 89% 100% 75% 87% 100% 80% 80% 100% 80% 97% 100% 70% 96% 100% 67% 95%
5. From the simulation, if excluding bad human behaviors, the expected task time estimation method, the schedule rule (within project and between projects), and task time distribution are the three major factors that affect the result of both methods.
From the above findings, if excluding bad human behaviors, and if the schedule rule for PERT is AEAP within project and between projects, in terms of mean project time, the CCPM method is no better than the PERT method because of logistical change. However, from our study, we identify two merits of the CCPM method over the PERT method.
1. Concerning the project plan, CCPM logistical change can plan a higher reasonable and reliable project plan than the PERT method because PERT either underestimates the project completion date (using Equation (3)) or overestimates (by directly taking the 90th percentile of task distribution of Figure 4.2 as the estimated task time).
Simulation results support that no matter whether uncertainty is low, medium, or high, CCPM demonstrates a higher reasonable and reliable project plan due to logistical change.
2. The scheduling rule that CCPM uses is as late as possible (within project and between projects). Scheduling a non-critical path and projects as late as possible is advantageous in delaying costs and avoiding bad multi-tasking. However, with the PERT plan, scheduling a non-critical path and projects as late as possible increases the probability of delaying the project because of no safety buffer to handle uncertainty (simulation results support this point), so scheduling as early as possible is always preferable. The CCPM with project and feeding buffers can tell when not to start and will not hurt the project being delay. This is also the contribution of CCPM logistical change.