This month we have been reviewing the latest release of Primavera Risk Analysis, version 8.7.

Standard practice in planning is to determine the required contingency for inclement weather, and place an activity bar towards the end of the schedule, prior to the Practical Completion milestone. Whilst this is an accepted norm, it does not really take into account the fact that not all activities are affected by inclement weather. Really, the “inclement weather bar” is a simple mechanism for a contractor to inform the principal of how much contingency they think they require to compensate for bad weather conditions on site.

Accepted also, is the fact that should an activity be delayed due to weather subsequently impacting the critical path (pushing the Date for Practical Completion out), the planner is then supposed to reduce the contingency duration to compensate. One could essentially accept this method as ‘placing a cushion’ on the end of the critical path.

However, suppose a weather delay is incurred on the activities that are not on the critical path? How can we model this effect to determine the most appropriate amount of contingency? How can we optimise our plan to account for poor weather? Enter Primavera Risk Analysis:

Here, we define the different weather events that may occur. For example, we have added July 2010 and have stated that there is a 25% chance of rain which can cause between 1 and 10 days delay. We would typically get statistics such as these from the Bureau of Meteorology.

We then can, for the project, define multiple collections of weather events. Say for example you had a project where work was being carried out in South Australia and Japan simultaneously, you could define a weather events profile for both regions.

Next, we proceed to allocate the weather profiles to respective activities:

Note that above, we need to select activities one by one and allocate the event calendar. Whilst this appears cumbersome at first, it forces us to actually ‘think’ about what we are doing – thus producing more accuracy in our results.

Once completed, our ‘weather events’ are loaded into the schedule underneath our project plan. During risk simulation, these individual risk activities will activate thus producing delay across our programme. We can define the number of iterations required and actually physically step through each iteration of the project:

Remember that “Risk Analysis” is about analysing risks in your programme/project. “Monte Carlo Simulation” is about running the project in theory, multiple times over to understand what the most likely outcomes will be. Whilst it may seem appropriate to run a simulation 20 million times, it actually does not have much effect on the end result. In conjunction with another consultancy firm,we ran a Risk Simulation on a project in late 2009 (with another product) and after leaving the simulation to run over a weekend, the end result was that the results mimicked those of a typical 1,000 iteration simulation.

You will notice that as we step through each iteration, small blue shadows appear on the Gantt chart itself. you will also notice that the Finish Date for the first activity changes with each iteration. We are simulating the activity occuring multiple times and each time the weather patterns are different. This makes the analysis more realistic of real world events.

Once the simulation is run, we can click on the whole project or a single activity to review the statistics from the analysis. In the above screenshot we can clearly note that there is an 80 percent chance (P80) that given our knowledge of weather conditions in the state, we will most likely finish that activity on 22/11/2010.

We can also review the data for the weather events as well to understand what actually happened with the weather over the simulation. We simulated 1,000 years of the project essentially so what happened in July for each iteration? The above screenshot depicts that in 1,000 simulations it did in fact rain 250 times. Of those 250 times, 80 percent of them occured prior to the 26/7/2010.

So there you have it. Weather Modelling using Primavera Risk Analysis. But what now? Where to from here? What about that contingency bar? Considering we’ve already done the analysis we only need to use the analysis results to determine our probabilistic dates. These dates are then built up using weather modelling across the whole schedule and not just putting a “calculated bar” at the end of the critical path. We then place these dates in our Contract to ensure that our risk of inclement weather is well and truely covered.