Scope of Discovery

The mercenary of our group has been heard to say, quite often, “One of the problems with idiotic traditional systems is scope of discovery is wrong.” Pressed to explain, he is apt to add, “When you analyse something, anything, you can only do an appropriate analysis if your underlying data was discovered on a broad applicable spectrum.” Luckily, translation of the idea is available for mere mortals who speak something like plain English.

Scope of discovery in traditional safety systems is wrong, because the statistical metrics that govern “whether you are safe” depend upon avoiding recording conditions that skew them. Consequent to that, traditional safety counts what benefits them more than what will harm them, since they are almost exclusively measured by post-event metrics. The problem, what makes this wrong, is that when you make subjective choices about what to record, you create a pool of data that, when analysed, is ignoring what is often the largest part of the data-set that should be analysed.

A case in point is the classification of incidents. The metrics that are used to declare a company safe will degrade that rating significantly if your near miss recording exceeds a certain ratio. While ignoring near miss recording is then almost a matter of commercial survival in some sectors, doing so actually trades off the perception of safety (by way of statistics) with the development of a safer workplace. You cannot fix what you have never seen.

This might not matter if the difference between a near miss and fatality was not often a matter of centimetres or seconds. The intention or recording near miss in a risk-managed environment is to identify the cause of the incident, basically to identify a control failure. Those failures define how to apply resources to better controls, and without the ability to analyse them, we cannot do that. By not recording them for fear of statistical self-destruction, we have no process that will avoid these failures creating unsafe conditions that increase risk until an encounter becomes a serious one, perhaps even a fatal one.

Good safety needs to be a side-effect of intentional management, not luck. This requires more data of a higher quality to perform better aggregate analysis, and if scope of discovery is being repressed the result is a skewed database. You will be analysing risk based not upon risk reality, but upon the encounter of risks where distinct failures created negative outcomes. There is no way to control preventatively with any effectiveness, since the best you will do is create a reactive control modification. What is required is a scope of discovery that provides a massive aggregate pool that can be used to execute predictive analysis.

Workers routinely identify risks in the workplace, and if you capture those identified risks you can control for them. The control mechanisms may or may not be efficacious, but the only way to determine it is to monitor effectiveness. Post-incident control failure analysis is important, but if the only incident types one ever analyses are negative impact ones (injury or fatality), then you are placing faith in the controls rather than assessing them. If you also analyse controls via inspection processes, and include a wide range of near miss and even better “risk identification” events in your analysis pool, you are creating a method to objectively create proactive preventative control improvements.

Scope of discovery is the key to better analysis and the provision of better safety.

In an asset inspection, if you check that a guard is being properly maintained, you are confirming control. If over a year the inspection is indicating the control is not being maintained, you have an opportunity to analyse that data effectively. If in 50% of the cases of asset inspections that control is failing, you have a serious risk pocket.

Now, if you have recorded a dozen near misses where that same control failed, you have a pool of data that can be analysed. Yes, there is no loss of property or personnel, but the reality is that if in a specific organisational location (welder shop floor in some shop, for example) half your near misses are indicating the same control failure, you can project with fair accuracy that at some point near miss will become something more. Maybe the metal fragment the guard is intended to deflect down to a safe pan will hit another machine and damage it, incurring property loss, or maybe that same fragment will cut an employee, blind one, or kill one.

When that happens your incident becomes an injury or fatality is that, unless you have this other data, you will be looking at a failed control in total isolation. The guard was down, the employee was negligent, case closed. Except, if half the inspections done are showing the guard is out of place, and you have a dozen near misses leading to this event, the context is probably different. Why is that guard such a problem? Is the control ineffective? Could you have implemented a modification that avoided the costly injury or fatality?

By limiting scope of discovery you create a false sense of safety, and you ensure your eventual control enhancements are done in a vacuum. Preventative measures are only possible if you have more data of better quality, and if your analysis crosses the limitations of cost-only incidents. The dozen near misses in our example would have alerted to the likelihood this control would be in a failure state eventually, but if the only time you hear about a control is after it failed, it will never be capable of providing preventative measures.

Risk-management is about embracing the range of your available data to ensure risk-awareness is real, and objective metrics exist to assist in focusing resources preventatively.