GERT Modelling in Project Management

Project delivery plan is an attempt to describe processes required to deliver scope with available resources to achieve desired goals.

One of the most critical limitations in project delivery systems is lack of ability to describe processes as they could be performed. Deterministic CPM-based modelling is missing ‘OR’ and ‘IF/THEN’ logic that is essential for describing complex processes. GERT modelling addresses this and some other limitations.

GERT (Graphical Evaluation and Review Technique) modelling is a project management and network analysis technique that allows for the probabilistic treatment of both the activity duration and the project structure. Unlike the traditional CPM/PERT method, which assumes a single, deterministic path through a project, GERT is capable of modelling complex, stochastic networks that include loops, conditional branches, and probabilistic events.

Key features of GERT include:

Probabilistic Approach: GERT allows for activities to have probabilistic outcomes. For example, a task might have different possible results, each with its own probability of occurring.

Conditional Branching: GERT can model scenarios where the path forward depends on the outcome of a previous task. This is useful in projects where decisions are made based on the success or failure of earlier tasks.

Loops and Iterations: GERT can handle loops, meaning it can model activities that may need to be repeated multiple times based on certain conditions, something CPM does not easily account for.

Complex Networks: GERT is suitable for projects with highly complex workflows, where tasks are not simply linear or sequential.

However, GERT is more complex and less commonly used than CPM due to:
– its advanced probabilistic nature,
– the specialised knowledge required to effectively apply it, and
– enforcement to apply CPM-based planning and delivery tools that don’t support GERT.

GERT is particularly useful in projects, where uncertainty and variability are significant factors.

Any process can be described by five operations: AND, OR, XOR, IF/THEN.

Input:

• AND: all paths have to be performed
If Task A and Task B must both be completed before Task C can start, Task C is dependent on an AND condition.

• OR: alternative (one or more paths can be performed)
If either Task A or Task B must be completed before Task C can start, Task C is dependent on an OR condition.

• XOR: alternative (only one path possible)
If either Task A or Task B can lead to Task C, but not both (only one path can be chosen), Task C is dependent on an XOR condition.

Outputs:

• Deterministic: every outcoming activity has probability equal to 1, which means that everyone will be performed.
E2E Testing commences after the testing of all components is completed.

• Probabilistic: each outcoming activity has some probability of appearance.
There’s 70% probability that the UAT will pass and the project can move to deployment, but 25% chance that the UAT will fail leading to additional rework and testing, and 5% that the third UAT cycle is required.

CPM-based modelling doesn’t support OR, XOR operators and probabilistic outcomes.

GERT modelling has its language: three inputs (AND, OR, XOR) and two outputs (deterministic and probabilistic) create six node combinations.   

Similar to CPM modelling, GERT nodes are connected with arrows that represent logic between nodes, but outputs can be predefined or not. It can be based on predefined probabilities or current project conditions. 

More project planning examples:

• The project delivers multiple outputs. Release notes commence when the first output is ready, but due to uncertainties, it is unclear which outputs will be ready first (XOR logic).
• If the risk event (30% probability) is materialised, additional activities have to be performed (probabilistic output).  
• If the project is more than ten days late additional resources are going to be deployed to accelerate testing (conditional decision).
• Three bulldozers with different productivity rates (volume units per hour) could be assigned on Task 1, Task 2 and Task 3. Activity durations depend on the assignments, but there are technological limitations that must be considered:
  • Only one bulldozer can be assigned to ‘Task 1’ (XOR logic);
  • One or two bulldozers can be assigned to ‘Task 2’ (OR logic);
  • Exactly two bulldozers are required to perform ‘Task 3’ (AND logic), but it could be any two bulldozers.

Such cases can’t be described with CPM-based modelling easily, so estimators and planners must guess. 

PAST

The history of GERT Modelling in Project Management is well described in “Network Scheduling Technique for Construction Project Management” book published in 1997:

“There were hundreds of remarkable pieces of work that appeared in the sixties and seventies. Some of these works made significant improvements to the original techniques. The most interesting development was the addition of decision nodes to the network. The aim of these works were to handle several alternatives of the project in one network.

One of the pioneers was Freeman whose work was generalised by Eisner and Elmgrabhy. The result of their work was the addition of three types of decision nodes: AND nodes; EXCLUSIVE-OR nodes and INCLUSIVE-OR nodes.

Their works are considered as the basis of some research activities that were led to the development of GERT (Graphical Evaluation and Review Technique). The initial GERT developments were made by Alan Pritsker who published his results in 1966. The growing of the GERT family has started with Pritsker’s work and soon became huge. The Precedence-GERT that was able to handle several types of activity relationships was developed by Pritsker in 1974. Arisawa and Elmagrabhy devised a technique that was able to perform time cost trade-off in GERT network in 1972. In 1975 Herbert developed his R-GERT method, which solved resource constraint scheduling problems in GERT”.

This wasn’t just a theory. GERT modelling systems were developed and applied for project delivery. 

PRESENT

From widely applied planning and delivery systems, only one supports GERT modelling: Spider Project. Spider Project Team developed unique methods to simulate GERT-based processes:

Trigger: a dummy activity used to describe probabilistic events. A trigger may have multiple outputs. The event has a probability to occur, and each output has its probability but all probability of outputs has to be 100%.

Switch: a dummy activity used to simulate conditional situations usually caused by cumulative effects. Conditions could be described based on actual or forecast project data by applying simple or complex formulas.

Team: Resources could be assigned directly or via teams. A team may have one or multiple resources. One or multiple teams could be assigned to an activity. This allows for the simulation of complex conditional assignments.

FUTURE

Mature planning requires a correct description of project processes, and it is only a matter of time before GERT modelling becomes the new standard for planning and delivery. Two main factors impact growing attention to GERT planning: Quantitative Risk Management and AI-based planning.

• Quantitative Risk Management
Project risk consultants recognise the power of Monte Carlo Simulation (MCS) analysis and use it to integrate risks into project delivery models. However, the community still lacks consensus on the correct application of MCS in project management. While there is general recognition that it is crucial to have a reliable model, there is not a joint agreement on what a ‘good model’ actually means.

It is easy to describe but hard to develop. A good model correctly simulates project processes and possible decisions. CPM-based models are not sufficient for that. To be reliable, Monte Carlo Simulation requires the development of a GERT model.

• AI planning
Predictive AI analysis requires probabilistic conditional models. Deterministic CPM-based models powered with AI may show a length of delay if the project sticks to the old plan, but they can’t find optimised solutions as the model is missing crucial data. Progress to the next level of maturity requires the development of easy-to-use GERT-based planning systems. Such systems may also help crack two major planning challenges: time-cost balancing problem (TCBP) and Resource Constraint Scheduling Problems (RCSP).

I found that the application of GERT modelling is enjoyable and beneficial.