Do operational risk taxonomies have a future?

ORX has recently published a new reference taxonomy for operational risk. It has been positively received by members, the industry and by supervisors. We’re investing in the extension of that taxonomy in 2020. But, given the rise of advanced analytics, do taxonomies have a future?

Our recent work has been focused on bringing the industry operational risk taxonomy up to date. We started with the current event-based taxonomy. Many event types have stood up well and we simply reflected more user-friendly language. Some event types have proven to be more popular and so these we expanded. Some event types weren’t significant in 2007 but are very much more significant now, and these we added.

It is interesting to note how our method, in reaching these conclusions, differed from previous efforts. Historically the formation and amendment of a taxonomy was expert-led. We chose instead to be data-led and then apply analytics and expert judgement. We have also explicitly stated that we expect the taxonomy to continue to evolve. These decisions reflect the more dynamic nature of risk. So the effort becomes ongoing, making small changes more often.

This year we’ll be investing in testing how well our reference event types work and developing a taxonomy for cause and for impact. We, along with many in the industry, believe that the combination of cause, event and impact offers a much better base for categorisation of events and solves many of the more vexing categorical issues.

Why then, now, question the value of these efforts? The answer lies in the fundamental changes we are seeing in the way that technology allows us to collect, label, analyse and report data. Previously you needed a top-down taxonomy as an instruction to enable the bottom-up semi-manual collection and analysis of data. Within firms, this made data comprehensible and, across the industry, comparable.

Technology enables machines to pull unstructured data directly from business systems, develop bottom-up categorisation and unconstrained analytics in real-time. We can therefore learn and make changes much more quickly than in the past.

Why then the need for a taxonomy? Why not just let the machines just do their work? The answer lies in who you’re producing analysis for. People. We expect taxonomies to change a little more often and we anticipate that new technology will make it easier to make those changes. But people understand stories best and the best stories require a constant cast of characters. So, taxonomies do have a future, just different from the past.