The latest version of IncidentXP & BowtieXP, software release 9.2, allows a new way of showing incident data on a bowtie diagram. We think that the bowtie is a communicative tool and as such, we should add as many features as possible that can help you communicate and visualize information. This blog post offers a look behind the CGE scenes, as we explain our thought processes behind implementing this new feature. In turn, we ask you as our users, what you would think about the data and how you would interpret it.
If you have both BowTieXP and IncidentXP, it is possible to show incident data on your bowtie diagram. The new software feature is called “Show Failure Percentages” and works using the Incident filter (see figure 1).
Figure 1: Incident filter
For example, in figure 2, you can see that most of our barriers have a failure percentage of 100%. This means that from the incidents that have been analyzed and involved these barriers, these barriers always failed when asked to perform their function. The idea behind adding this feature is to show incident information in a more succinct manner.
Figure 2: Failure percentages shown on barriers the new way
Comparing this ‘new’ visualization of data to the ‘old’ way of showing it like in figure 3, we think the new way captures the same essential information, only in a much more condensed way. The below picture shows incident information shown on barriers the ‘old’ way.
Figure 3: Incident information shown on barriers the old way
Interpretation of failure percentages
One potential pitfall however, is that people forget that the percentages have a bias. The only time when the functioning of a barrier is examined is when an incident, near miss or near hit occurs in which the barrier is involved. As the flow of events is usually stopped by a single barrier, most incident diagrams will only have a single working barrier and quite a few failed barriers. This leads to a clear case of selection bias, where the picture is more negative than it would be in reality.
The danger is that people might interpret a ‘100% failure percentage’ as a barrier that never ever works, i.e. it fails 365 days a year. Instead of what it actually means, which is: from all of our incident analyses, for each incident the barrier was involved in, it did not work. In other words, the barrier may have functioned perfectly for 363 days a year, and only failed on the remaining two days.
In order to tackle this biased picture, it is important to have more than one way of getting information about the functioning of your barriers. After all, failure information is a lagging indicator, whereas leading indicators might be able to give more realistic predictions of barrier functioning. For this reason, it is important to also consider visualizing inspections and audits via the audit module.
What would you prefer?
Which way of showing incident information do you prefer to use? This ‘new’ way, or the ‘old’ way?
Let us know in a comment below.