The seventh step of a typical Oracle | Primavera Risk (Pertmaster) Monte Carlo analysis is to use standard risk reports for proactive planning and project management. This section will focus on generating and successfully presenting the tornado chart for project drivers.
Video Walkthrough of the Tornado Chart in Oracle | Primavera Risk (Pertmaster) (Duration: 17 min: 28 sec)
Navigating the Tornado Chart
The risk analyst will open the tornado chart in one of three ways:
- Follow the menu path Reports | Tornado Graph… (F11).
- Click the green Tornado (F11) icon on the toolbar.
- Right-click on an activity or milestone and click Show Drivers (logical predecessor option usually recommended when selecting a reporting milestone or activity).
Right-clicking on the graph gives the project analyst formatting options, copy options, and the advanced settings interface.
Right-clicking on the statistics will show options for formatting and copying the stats.
The Bookmark Visible Tasks button (Bottom-Right corner) will set bookmarks in the Gantt chart view. This will often help for quick mitigation scenario modeling.
Task uncertainty and risk events are coded in different colors. Risk events will be in red on all tornado charts.
Tornado Chart Menu Options
- File – options to print and export data.
- Edit – allows the analyst to change the sensitivity calculation settings, use various copy functions, and adjust the bookmark options.
- View – users can switch between task or risk mode. A risk analyst can also switch between tabs (cost, schedule, etc).
Tornado Chart Tabs
- Duration Sensitivity Tab – this chart is very confusing for many users. In simple terms it shows what activity is driving another activity or milestone. It will look at two activities in a bubble and calculate the correlation of their duration. This can be shown looking at a scatter plot on one driving activity versus the reporting milestone. If the driver duration is placed on the y-axis and the duration to the reporting milestone is on the x-axis, then creating a scatter plot for each iteration will allow a project manager to see how the data set trends. If each time the driver increases in duration, the duration to the milestone increases, then the correlation percentage will be high. Simply put it sees that we can predict what the milestone will do based on what that activity does. If an increase in the driver duration seems to have very little impact on the reporting milestone duration, then the correlation should be quite low. The slope of the best fit or regression line through the data set can be used to determine the correlation percentage although Primavera Risk does have a few different correlation metrics that may give very slightly different answers. The larger the slope of the line, the stronger the correlation will be.
- Cost Sensitivity Tab – the cost sensitivity and duration sensitivity chart are read in exactly the same manner.
- Criticality Tab – this metric simply shows the percentage of iterations or simulations that an activity was on the critical path. If a risk analyst is reporting to a milestone or activity that is off the critical path, then the key drivers may show as 0% critical. They are still driving the milestone, however they are not on the overall critical path. An example is a contractual milestone that is off the critical path. The milestone will still have drivers shown on the duration sensitivity tab, however none of the drivers may jump onto the critical path.
- Cruciality Tab – this metric simply takes the duration sensitivity percentage and multiplies it by the criticality number. The attempt of this chart is to show not only if the activity is a driver to a reporting milestone or activity, but whether it is on the overall network critical path. A key driver for a contractual milestone that is not on the critical path may show up as 0% because the criticality might be 0%. It may be a key driver to that activity, however it is not driving the overall critical path.
- Schedule Sensitivity Index (SSI) – the criticality index is multiplied by the task standard deviation and that number is divided by the project standard deviation. It does calculate very quickly. Instead of looking at the duration to come up with a correlation coefficient like duration sensitivity, this metric is using the standard deviations of a driver and the standard deviation of the project to calculate the output. This metric is only useful for overall project drivers because the standard deviation of the entire project and not the standard deviation of a reporting milestone (or activity) is being used..
Options Located Under the Graph
- Task vs Risk (Display Mode) – task mode will show a tornado chart of all activities and risk events. Risk mode will show a tornado chart of risk events only.
- Task types to display – users can choose to show all or only certain types of tasks. Showing normal tasks is most common because they are the only activities that can be loaded with three-point estimates. The left column will have all activity types and the right column will be grayed-out unless a risk register has been integrated into the Monte Carlo Simulation. The right column allows users to hide or show risk register data on the tornado chart.
- Risk Summary Task – shows a bar for the task plus the connected risk event.
- Base Task – shows a bar for the original task a risk is tied to.
- Risk Impact – shows a bar for the risk event.
The filter options include:
- Show top <x> risks – this option allows the analyst to choose how many activities are shown on the tornado chart.
- Ignore values smaller than + or – <x> – the project team may want to ignore very small values. Very low values serve more as noise than usable data. Low numbers indicate that the task is not a driver.
- Ignore negative values – the project team may also choose to hide negative numbers. Generally small negative numbers are seen as noise. The risk analyst may want to investigate if a large negative number appears. Large negative numbers are usually caused by opportunities. This is often correct and not a statistical issue.
What Does It Mean?
Which chart should I use? – The answer depends on what is trying to be shown, but duration sensitivity is often the best metric because it is simply showing the correlation between a driver and a reporting activity. I do not need to take the overall network critical path into account if I know that the milestone delivery is inherently critical. Flaws in logic, linked sub-projects, maintenance activities or reports after the project completion, and many other issues can cause issues with the calculation of the critical path. The duration sensitivity is not affected by any of that. The duration sensitivity will show the correlation between two activity durations to determine if one is driving the other. This is often the best metric for isolating a milestone and seeing what is driving its completion date.
- What do the numbers mean on the criticality chart? – The numbers on the criticality chart are the number of iterations that an activity was on the critical path.
- The definitions of the charts are above, however simplify the charts for the audience. Nobody wants to hear a class on statistics and correlation coefficients, however people seem to understand words like driver, impact, best fit line, trend curve, etc. For example, a project manager may highlight a piping activity as being the most critical project driver as it has the greatest impact on the project completion. Keep the explanation simple to help a broad audience with diverse backgrounds. If someone wants detailed information on the calculations, then it is usually better handled outside of the interactive session or project team meeting.
- What do the numbers on the duration and cost sensitivity chart mean? – Many users get tripped up on these charts. The numbers on the duration and cost sensitivity chart are a correlation metric. It shows how strong the correlation is between to activities to predict if one activity is driving the other. Showing a scatter plot with a regression or best fit line can often simplify the concept for an audience, however avoiding stats talk and focusing on what the data means can be extremely helpful.
- The numbers are on the sensitivity chart are confusing me. Why don’t they add up to 100? – The fact that the numbers are a correlation metric and not a percentage of impact is hard to understand for many people. Users can normalize the percentages in the graph options (Format Menu | Graph | Drawing Options Tab). This will put the drivers impact on a scale to 100.
- I normalized the percentages and it is still confusing. Can I make it simpler? – The numbers can be completely hidden in the graph options as well (Format Menu | Graph | Drawing Options Tab). Most audiences can relate to big bars are bad, small bars are not so bad.
- Where do I go from here? – Use the data to run scenarios. The histogram told you where you are at. The tornado is telling you how to fix it. Use the data to run scenarios. The purpose of the analysis is to give proactive information to improve the probability of project success. The purpose of a risk analysis is not to give terrible news with no insight on how to fix it. Monte Carlo analysis is often seen as useless because the answer does not make sense or because we do not like the answer. The answer must make sense. Not liking an answer that makes sense is a different issue. If you don’t like the answer, then the graph indicates we need to put steps in place to avoid the answer instead of sweeping the results under a giant carpet.