- Reactive Risk Management – Do something to address the risks identified in an accident or incident after it has occurred. This is how early aviation safety did risk management because they did not have any/enough data to do proactive or predictive risk management.
- Proactive Risk Management – Do something before an accident happens by utilizing data to identify risks from past accidents or incidents. Aviation safety has a well earned reputation for being a leader in proactive risk management.
- Predictive Risk Management – Do something based on potential risk as determined from normal operational data (i.e. not accident data) to reduce the risk of an accident that has not happened (yet).
Today there is a tremendous amount of normal operational data available along with many new, efficient data processing tools to effectively exploit that data. Basic risk management tells us that identifying hazards, and then evaluating the probability and severity of the hazard, enables us to determine the risk. The availability today of normal operational data and the tools to effectively use that data enables us to identify hazards previously only discovered by an accident or serious incident. To be predictive we must be able to use normal operational data and modern data processing tools to show the potential risk and identify the benefits of reducing that risk by addressing the probability or the severity of hazards identified in the data.
Regulators and safety investigation organizations are reactive by nature, so a shift to being predictive will not be easy. Our wealth of data today enables us to not only look at past accidents and incidents, but to also see what is happening in normal day to day operations. This allows us to identify hazards and see what and where the current risk trends are, and to identify potential or new risks. There are two keys to being successful at being predictive in a reactive world:
- Have the data to verify the risk exists and show it is worth addressing, and
- Have the support of the decision makers to address the risk without having an accident to react to.
To get the decision makers support, there is a critical requirement that must happen - a change in paradigm. We now have the data to look at events – both significant and normal – to identify hazards and determine areas for potential risk reduction. However, our current safety system focuses on negative outcomes, not events. If an event does not have a negative outcome, it is rarely investigated. For example, after the Air France 447 accident several similar events were discovered. However, they were just events, and none of them had a negative outcome so they were just part of the normal operational data that was collected and the hazard, and resultant significant risk, was not identified and acted on. In another example, in a recent fatal takeoff accident the aircraft was unable to rotate for takeoff due to the controls being locked. The investigation determined that the flight crew rarely conducted pre-takeoff control checks, which would have discovered the locked controls. Since the lack of control checks were not a caution or a warning and were not an exceedence of accepted parameters, and not doing a pre-takeoff control check did not result in a negative event, it went unnoticed.
Being predictive is based on events – and preventing negative outcomes. If we act on negative outcomes, it is at best proactive, and normally reactive risk management. There are two keys to being predictive, particularly in a reactive world:
- Have the data to verify the risk and show it is worth addressing, and;
- Have the support of the decision makers.
The key to both of these is data. Data will enable us to use our predictive capabilities to further reduce risk. Decision makers, including individuals, organizations, and the safety and regulatory systems themselves, are reactive by nature. However, with today’s data capabilities we can hopefully use predictive risk management to generate a risk reduction action before an accident. By utilizing incident and normal operational data in our prediction process we will be able to show that we reduced the risk of an accident and hopefully avoid having to react to one.