Why ignore extreme events?

For much of my analytics work, disasters, force majeure, and large breakdowns are excluded from scope. It’s not because they’re not important! Recently an LNG tanker docked at the Curtis Island terminal near Gladstone was unable to be moved and blocked the terminal for 11 days, likely costing tens of millions of dollars. So why exclude such extreme and rare events?

The first reason is that we’re far more interested in business as usual (BAU) operations which will be how the system runs the vast majority of the time. There is little to be learned by blindly applying BAU rules to an extreme event. People will work overtime, schedules will be rearranged, perhaps even new logistics methods will be employed in the short term to keep running what can be kept running. All of the analysis that is done for BAU operating rules will need to be redone for every considered rare event. Keeping the scope to BAU avoids doubling or tripling the process-discovery period of any analytical project.

The second reason is that estimating the likelihood of extreme events is very, very difficult. There is no reason that the chance of a major disruption such as the one at Curtis Island could be predicted from looking at outages data for the terminal unless such problems are a common occurrence. The incidents of one or two hour breakdowns shouldn’t be extrapolated to give the likelihood of a week and a half shut. Due to the nature of extreme events being rare and the difficulty of estimating how many you’re expected to have in any given period, it’s best to leave them out entirely of a BAU analysis where they can overly skew results.

The proper way to handle the analysis of a disaster or other extreme event is to treat these scenarios separately. First, identify possible issues, “What if the liquification plant is down for X days?”, “What if the port is closed for Y days?”, or, “What if a vessel cannot move for Z days?”. Then consider mitigations that can be employed before estimating impacts. Finally, don’t tally up all these scenarios weighted by likelihood, but rather present them as their own results to avoid giving the impression that we can predict them with accuracy. Perhaps mitigations overlap and we can protect against multiple extreme scenarios with one action, or perhaps we’ll find that some scenarios actually don’t hurt as much as we expected. In all cases, these insights will not impact your BAU rules and should be left to their own project.

If you’d like help analysing your BAU processes or help analysing how your system can handle extreme events, reach out to us for a chat at Hello@NorthCardinal.com.au.