We all want to get home safely, right? For some of us, the worst risk we face at the office is a sore neck from bad posture. But for many others, the risk of serious personal injury on the job site is very real.
One of the best approaches to safety management begins from the ground up, through a safety-driven culture that focuses on getting every employee to be aware of their surroundings and actions. But just providing a set of workplace rules is not enough. Organizations need a relentless focus on continual improvement to ensure that they get better.
Continuous Improvement is a framework that recognizes that small improvements, compounded over time, can have a huge impact. It was Einstein after all who stated that compound interest is the most powerful force known to humanity. He was on to something.
Talk to any programmer and she will intuitively understand this concept. Large, successful projects aren’t the result of a single stroke of genius but rather the cumulative result of thousands of hours of work to make small improvements to a system. Unsurprisingly, the same holds true when looking at how companies deploy data analytics in general.
Back to workplace safety
One of the challenges associated with tracking safety comes from the interpretation of safety data. Although administrative fields like name, location, time, etc. tend to dominate safety incident forms, it is the written or verbal description of what happened that houses the truly important information. Naturally, a person can easily understand human-readable descriptions of safety incidents. But it’s difficult for a single person to ready hundreds of these reports and pull out trends.
How can we “read between the lines” and get to a point where safety teams can understand root cause for ALL safety incidents across an organization? In other words, how can we accelerate the feedback loop to drive true continual improvement?
By going digital, of course.
Workplace safety from the ground up
The Human Factors Analysis and Classification System (HFACS) was developed by Shappell and Wiegmann to address the overwhelming impact of human error in flight accidents. It considers four categories of errors and failures, including:
- Unsafe acts
- Preconditions for unsafe acts
- Unsafe supervision
- Organizational influences
For any safety incident report template, much of the insight related to each of the four HFACS categories is buried in the nuance of the written descriptions. For example, phrases like “was told to do” or “didn’t know” are very important in determining root cause. As mentioned earlier, while it’s easy for a person to understand this information when reading a single incident report, things become difficult when reading hundreds or thousands of reports.
Sample HFACS analysis
One approach to extracting nuance is to apply a method known as latent semantic analysis to text descriptions of safety incidents. By processing all text for hundreds of safety incidents in one project, we were able to learn that 74% of all incidents could be grouped into three categories:
- vehicle incident
- procedure not followed
- equipment defect
Note that these are not standard fields found on incident report forms. Rather, they were uncovered by processing all the reports and extracting the common themes from the incident descriptions. Not surprisingly, these categories change from company to company, and even from site to site.
Making workplace safety stick
Tying workplace safety back to continuous improvement, two important conditions must be met. First, the company must be able to get safety feedback in a timely manner. Once this data is analyzed for root cause, corrective actions must be taken to prevent future incidents. From a data analytics perspective, we see this as dashboarding and predictive analytics modeling, respectively.
By integrating safety into standard business reporting processes, companies can ensure that they get continuous feedback. At 3AG Systems, we achieve this for our clients by integrating bottoms-up incident categorization into corporate dashboards. Because the data is digitized, it can be updated with each new safety incident, ensuring up to date information for leadership.
Most importantly, predictive models can be developed that not only recommend follow-up actions to a specific incident, but that can identify when unsafe conditions are likely to cause another accident. Acting on “digital near-misses” may prove to be just as important as responding to accidents for the purpose of driving continuous safety improvement.
Companies can apply data analytics to reinforce their safety-driven cultures, provided they can parse through all the safety data at their disposal. We believe that a bottoms-up approach not only improves the quality of safety initiatives, but it allows companies to tailor them to their specific industries and environments, thus making programs more meaningful to all employees.