History Matching and Relevance Analysis
In many projects, it is hard to determine which historical data should be used as an analog for future analysis. For example, in many research and development projects, new projects may differ significantly from the previous projects.
Many risks can be similar for different projects. For example, the risk may have already occurred in a different project within an organization. As a result, historical information about risks can be more comprehensive than information about the duration and cost of tasks as many research and development projects have unique activities. The selection of an event with its respective probabilities and impact from the historical data is based on an analysis of evidence regarding how relevant the event is to the current activity or project. Relevance analysis is performed using the different criteria. If an event is a full or partial match according to the selected criteria, it will contribute to the overall evidence that this event is relevant to the current activity.
Selection criteria can include risk events and event chains that:
- Belong to similar projects (similar cost, duration, or objectives), managed by the same project manager, and performed by the same organization
- Belong to similar tasks or group of tasks (similar name, duration, and cost)
- Occur during work performed by similar resources
- Have similar names, probabilities, and impacts
If the historical data has been properly collected, it may contain information about how particular events affected a previous project. This data can be used to calculate the probability and impact of the risk in the current project. In addition to getting evidence of relevance for risks from historical data, project managers may define a relevance coefficient or belief that the event is relevant to the current activity. For example, a specific event chain has a 50% relevance according to the historical data. In addition, project managers may define that it has an additional 90% relevance based on their own understanding of the event. In this case, both numbers will be used for calculating the probabilities and impact of risks.
Sometimes events can cause the restart of an activity that has already been completed. This is a very common scenario for real life projects because a previous activity must be repeated based on the results of the succeeding activity. Modeling of these scenarios using event chain methodology is simple. The original project schedule does not need to be updated, as all that is required is to define the event and assign it to an activity that points to the previous activity. In addition, a limit to the number of times an activity can be repeated, or recycled, needs to be defined.