One of the most commonly used project control tool is earned value management (EVM). EVM controls the progress of the project from three perspectives:

  1. Planned Value (PV) is the approved time phase budget of the project. The PV defines the work to be completed at a specific time and their monetary value.   
  2. Earned Value (EV) at a specific time defines the work completed up to that time and their monetary value.
  3. Actual Cost (AC) shows the total cost up to that time at a given time. This cost is the total cost incurred when doing the work measured in EV.

The following performance indices and deviations are calculated using PV, EV and AC to summarize the performance of the project at a specific time in the EVM to the project manager:

  1. Timeline Deviation (TD) EV differs from PV. It summarizes how far ahead or behind the project the timeline is at a given time. For example, if the TD is positive, the value obtained at a given time will be higher than planned, so the project will show that it is ahead of time.
  2. Cost Variance (CV) EV is different from AC. It summarizes the budget performance of the project at a certain time.
  3. Time Schedule Performance Index (TSPI) is the ratio of EV to PV. Summarizes the time performance of the project
  4. Cost Performance Index (CPI) is the ratio of EV to AC. Summarizes budget performance of the project.

One of the reasons for the widespread use of EVM is the time and budget spent measuring cost and time performance together with the value of completed work. Estimates can be made about when and how much the project will be completed at cost, according to the values of TSPI and CPI calculated in the EVM. When this is done, the future performance of the project is modeled as a required mathematical function according to the time spent on that core, cost, and TSPI and CPI. However, Caron et al. (2013) likens it to using a car just looking at the rearview mirror. It is not a realistic assumption to assume that future events in an area of risk and uncertainty, such as project management, will resemble events in the past. To make more accurate estimates, the events that cause problems in the past, and the risk events that may cause future problems, should be considered.

The usefulness of EVM in high risk and uncertainty projects is limited by three reasons:

  1. EVM is largely ignoring the uncertainty in projects. Only the point values of EV, PV and AC are used when calculating performance indices and deviations of EVM; parameter uncertainty of these variables is not considered. Since projects are complex and ambiguous, it is unrealistic to assume that the point values of variables are known or guessed correctly. Specifying the EV as point value, especially for completed projects, is a difficult task for project managers and restricts the use of EVM. For example, it is often not possible to pinpoint exactly what percentage of a job in progress is running in a software project. Pinto (2016) stated that determining the exact value of KD is not possible in most projects, and has shown that the different assumptions used to set the value of KD reach very different results.
  2. EVM may signal to the project manager that this is a problem with cost or time, but it does not allow for numerical analysis of reasons for this and for individual risk events. In order for the project manager to make the right decision about the real problems, it is necessary to examine whether the problem persists or not. For this, the project control model should also allow analysis of the causes of the problems.
  3. It does not provide a direction on how to determine the EVM parameters. Considering that the projects are not similar and the projects are often limited data, a good project control model should have an infrastructure that will use both numerical data and expert knowledge.

In this project, a project control tool will be developed that combines the parameter uncertainty of EVM and numerical analysis of individual risk events, which would exceed these constraints of EVM. The developed tool will use Bayesian network models.