How would you recognize reoccurring key-lessons learned from over 200.000 annual safety reports? The solution: Big data text analysis for bowties.
When you are dealing with a vast number of text-based safety records the challenge lies in obtaining relevant information formatted to initiate genuine operational change. As stated by Prof. Coen van Gulijk from the University of Huddersfield, “one of the challenges for safety managers is dealing with vast amounts of data. Especially when it comes to near-miss or close call reporting the numbers can be vast and reading into all of them to gain a systematic overview is often impractical.”
The last thing you need is more shelf-ware or, heaven forbid, a procedure in which you are urged to create documentation on the documentation. Nevertheless, turning data into relevant information for most is not a straightforward process. In fact, a trend spotted among the risk management community is that of being very data-rich, but information-poor.
So, what is data?
The raw facts and statistics we collect for reference or analysis are what is known as data. Before data can be applied, it must first be processed, analyzed, and interpreted to produce important intelligence and insight that help to improve business performance. If you are unsure where to start you can begin by answering 3 fundamental process safety questions:
- Do you understand what can go wrong?
- Do you know what systems you have in place?
- Do you have information to assure this works effectively?
The first question requires you to really understand your risks. A risk is the probability of an occurrence that adversely affects the realization of the organization’s business objectives, or causes serious damage to the environment, or leads to hurting people. Ask yourself: do we know the most significant risks and are they assessed and aligned with the organization’s risk attitude? Whereby risk attitude is defined as “an organization’s approach to assess and eventually pursue, retain, take or turn away from risk” (ISO 31000).
The second question will assist you in identifying which preventative and corrective measures your organization has in place to prevent incidents from happening. Who is responsible? Which barriers or controls are critical? This question is about the implementation of your barriers and controls. It is a combination of utilizing applicable compliance frameworks and assessing your process in a barrier-based manner, using the bowtie methodology.
In terms of big data analysis, the third question, which is focused on collecting and processing assurance data, proves relevant. Having established that turning data to information is key, this final process safety question concerns itself with the actual performance of your barriers and controls. It asks you to use all available data to generate information that should help you understand the availability and effectiveness of your critical barriers and controls. This type of assurance data often comes to us in the form of a report and as I’m sure you can imagine over time this documentation piles up. Now recalling those 200.000 annual safety reports, depending on your organizational capacity, a phenomenon known as information overload could occur. Overexposure to data can make it difficult to understand an issue or grasp key-lessons learned and effectively make decisions.
From data to information
How can this data be compressed? One way to go about it would be to plot safety reports against a bowtie. Utilizing a bowtie diagram as a framework provides you with a visual display that pictorially communicates relevant information about the quality and status of your barriers.
Moreover, The University of Huddersfield has started to develop a method that can be used to seriously compress big data sets and extract key hazards from text, using the familiar bowtie framework. In their recent efforts scholars from the University of Huddersfield worked on adding more functionality to a bowtie, namely: machine-assisted text analysis. In their publication “From free-text to structured safety management: Introduction of a semi-automated classification method of railway hazard reports to elements on a bow-tie diagram” published in Safety Science (2018), they used large text-based close call reports to interpret the text with computers and map them against an existing bowtie. To state an example of the effects the combined approach ‘big data and bowties’ can have on extracting key insights, see figure 1 in which six of the most common cause pathways are identified from monthly produced close call documentation.