Data Analytics Are Becoming More Prominent in Safety Programs Across America’s Workplace Facilities and Changing The Way Safety Professionals Operate.
In October, the Colorado Safety Association presented the annual Rocky Mountain Safety Conference in Denver. Despite an early snowstorm, the event was well attended by safety professionals from across the region. I have always found the most valuable part of these events to be the chance to get perspectives from the trenches. As much as LinkedIn posts can be educational and entertaining (thank you Ed Davidson), nothing beats talking to safety professionals directly about their thoughts, experiences and challenges.
Recently I have been especially interested in the growing use of analytics and the search for helpful leading indicators of probable safety performance. The concept has been around for quite a while. Most safety professionals have long recognized that if they could collect reliable close call reporting, they would be able to zero in on potential high-risk locations, teams and behaviors – if only it were that simple! Close call (or near-miss) reporting is by its nature hard to get. The general sense is often, “well that was a close call – glad nothing happened – let’s get on with it…”. Add the fact that many systems make this reporting an ordeal, and no wonder it is unreliable.
Another issue with close call reporting is that much of it’s assumed value comes from the familiar “Accident Triangle” relationships. Originating from studies by safety pioneers as far back as 1931, a more-or-less fixed relationship between unsafe acts, near-misses and injuries was posited in the triangle, so if we could reduce unsafe acts and near-misses, then it followed that injuries would go down, which in fact did happen.
This is helpful, but dependent on probably incomplete reporting of near-misses and observations – what could we measure to fill out the picture? I have ideas for this, as did several safety professionals with whom I talked at the conference. Some might be kind of obvious, for example, incidence of repeat inspection failure items; time to respond to, and develop remedies for, incidents, close calls or observations; consistency with which job hazard briefings are conducted. But there may be other less obvious occurrences that correlate to incidents that we don’t know about. And these may well vary significantly between industries and companies within industries, which means that effective solutions may have to be tailored in part to specific companies for maximum effectiveness.
OSHA is off to a great start in taking a serious look at possible metrics we should be thinking about. I would guess that we will learn more about how activities and occurrences relate to injuries and property damage when the current generation of safety and operations software systems have collected enough data to give artificial intelligence algorithms something to work with. That is not as far away as you might think. Already AI is driving predictive preventative maintenance, which in itself enhances safety.
I am optimistic by the number of safety professionals out there who are also thinking of these possibilities and opportunities, and look forward to continuing the conversations with all of you!