oculai provides two methods for predicting when a project will be completed. They are based on different assumptions, use different inputs, and answer slightly different questions. Understanding these differences helps in choosing the right method and interpreting deviations between them.
What each method does
The Critical Path Method works at the activity level. It takes the current actual state in the schedule and runs the planning engine again, fixing already completed and active activities. Future activities are treated as if they will take exactly as long as planned. The result is a forecast progress curve: a single deterministic trajectory showing how cumulative progress will develop from today until project completion — including forecast end dates for all future planned activities.
The Monte Carlo forecast works based on the planned/actual progress curve and ignores individual activities and their dependencies — the decisive factor is the pace achieved so far. Based on this past pace, 50,000 possible future scenarios are simulated and a distribution of possible progress curves is generated.
The key difference
The critical path method asks: Taking into account accelerations and delays that have already occurred — when does the project end if all remaining activities are carried out exactly as originally planned?
The Monte Carlo method asks: If the team continues at the pace achieved so far — when does the project reach 100% progress?
These are genuinely different questions. A project where activities are largely completed on schedule, but the overall pace is consistently below plan, shows hardly any delay in the critical path method, but significant delays in the Monte Carlo forecast. Conversely, structural delays in the critical path can shift the critical path method far back, even though the general work pace appears acceptable.
When to use which method
| Critical Path Method | Monte Carlo |
Required inputs | Complete activity structure with dependencies and actual data | Planned/actual progress curve |
Result | Forecast start and end dates for each activity | Distribution of possible progress trajectories |
Suited for | Analyzing which activities are driving delays | Assessing whether the work pace is sufficient for on-time completion |
Early project phase | Reliable even with little history | Unreliable with few data points |
Late project phase | Can become too optimistic or pessimistic when planned values consistently deviate from actual values | Reliable with sufficient history |
Pace-driven projects | Can underestimate delays | Maps systematic under- or overperformance well |
Dependency-driven delays | Correctly accounts for delays along the chain | Does not account for bottlenecks from dependencies |
When both methods agree
When both methods deliver a similar end date, the reliability of the forecast is higher. Both approach the same result from different perspectives: once via the structure of the plan, once via the observed work pace.
When both methods diverge
Divergences are often more informative than agreements.
Monte Carlo later than critical path method: The team is consistently working slower than planned, but the schedule has not yet fully reflected this. The planned durations are probably too optimistic.
Critical path method later than Monte Carlo: Structural delays in the dependency chain are shifting the end date, even though the general pace is good. This happens when a few critical activities are delayed while parallel work is proceeding faster than planned. The problem then lies in the sequence, not the pace.
Limitations of both methods
The critical path method assumes that remaining activities will take exactly as long as planned. If the plan is unrealistic, the critical path method also remains too optimistic. It also requires a maintained and up-to-date activity structure — an outdated plan is not a reliable basis.
The Monte Carlo method assumes that the future resembles the past. However, if the project moves into a qualitatively different phase (e.g. from basement floors to standard floors), historical values may no longer be representative. It also requires a sufficient number of data points (= recorded weeks of progress data) and cannot explain why a project is progressing faster or slower.