Year after year, retail workers are less likely to get hurt on the job. The industry has seen a steady decline in its injury rate, dropping every year since 2012. But retail work has also gotten more deadly with the number of fatalities up five percent in 2016, the most recent year for which data is available.
Automation reduces the risk of many types of injuries, including the top two causes of injuries in retail: overexertion and lifting and handling materials. However, introducing large machines into the work environment creates hazards of its own, and the proximity of workers and heavy equipment can be a deadly mix.
No retailer is immune the to the risk of a tragic incident. Three months ago, an employee was killed when a shelving unit fell on him at a North 40 Outfitters store in Colville, Washington—and it happens at the biggest companies as well. In August 2017, for example, an Amazon warehouse worker in Indiana was crushed by a machine he was servicing; a month earlier, at an Amazon distribution center in Pennsylvania, a worker was crushed to death on the loading dock.
Problems leading to significant injury events increasingly have their origin in human-machine interactions, but why do they happen? Rather than workers simply making errors or wrong decisions, incidents often arise from mismanagement and failures in supervision of automated resources. This fact can muddy injury investigations and make it more difficult to discern the complete range of reasons for why a worker may have acted as he or she did.
An Investigation Challenge
As retail operations change, it’s vitally important for retailers and LP leaders to monitor the impact on worker safety. That places pressure on the nearly 400,000 workplace injury investigations that retailers must conduct annually.
Unfortunately, there is an inherent problem in conducting them. That is, investigators conduct them by looking back on a sequence of events. The result is that workers’ actions can seem to lead to an inevitable outcome, with investigations concluding where they went wrong.
This “hindsight approach” has problems, according to a growing body of occupational safety research. For one, conclusions about how workers’ behavior contributed to an event can easily become counterfactual and judgmental—stressing what people should have done to avoid an event but failed to do. As such, incident analyses often don’t explain much—and may not help prevent the next event.
It’s important to be mindful that there are numerous shortcomings to this injury analysis model:
- Saying what people failed to do or should have done does not explain why they did what they did.
- Investigators already know the outcome of workers’ actions, so investigators evaluate and make assessments about workers’ actions and performance in light of it. This “hindsight bias” often makes a complex history of leading events seem much more determinant. It often seems that workers’ actions led inevitably to the outcome.
- Current understanding of human error is rather unscientific. As such, investigators tend to rely on generalized concepts to explain an incident, such as “a ‘casual atmosphere’ contributed to the incident.”
- Investigators understand more about a problem event than the people who were involved in it. With hindsight and additional perspective, investigators see a situation more completely than the individuals involved did at the time it happened. They may have been operating under very different assumptions at the time of the event, so investigations often misunderstand workers’ actions or decisions.
Another common problem is that injury investigations typically scrutinize worker behavior at critical decision points and then look—in isolation—at whether those decisions were reasonable. In doing so, it’s typical to compare worker behavior against written procedures for operating a machine or completing a task, such as stocking a store shelf. But that is not very informative, warn safety researchers. There is almost always a mismatch between actual behavior and standard operating procedures (SOPs) that investigators can locate after the fact. Pointing out divergencies sheds little light on the “why” of the behavior in question. Example: how many investigations of store associate injuries identify “hurrying” as a condition during the incident? But “hurry” is a condition you identify in hindsight, one that tends to guide a search for evidence about itself.
Is There Any Other Way?
Sort of. Workplace incident analyses can’t avoid looking back in time, but they can minimize problems associated with a “hindsight mentality.” The goal is to focus on how people’s actions and decisions may have made sense at the time, and the key is to move assessments of worker actions back into the flow of events of which they were part—and which helped bring them forth.
In order to understand how human error or action may have contributed to an incident, investigators need to step into the past themselves. Human performance is fundamentally embedded in—and systematically connected to—the situation in which it takes place, research shows. Human errors are not made in isolation—they are connected to the tools they work with and the environment they operate in. An assessment of worker action in an injury event, therefore, is only meaningful if it is made in reference to the environment in which the worker performed his or her actions.
6 Best Practices for Safety Investigations
How can you do it? Here are several helpful ideas outlined in safety research and by safety consultants we interviewed:
- Begin by adopting the “local rationality principle.” That is, start with a mindset that—at the time of the accident under investigation—workers were probably doing reasonable things given their point of view, their available knowledge, their objectives, and their available resources.
- Identify “shiftmarkers.” There is often a point in events leading to an accident where a worker realizes that the situation is different from what they believed it to be previously. These shifts are markers that investigators should focus on to understand what was unfolding around workers that they may have been responding to.
- Use time and/or space to organize investigations and describe activities and events. Worker actions unfold over time and in some physical space. By organizing data spatially and temporally—through the use of maps, timelines, or both—it can be easier to couple workers’ behavior to the process and location at the time of the incident. Such organization often yields further clues about why wrong actions may have made sense to people, say researchers.
- Identify the start point of your investigation. Technically, there is no such thing as the “beginning” of a workplace incident. For example, you might trace a current mistake all the way back to when senior leaders wrote the company mission statement and left out any reference to safety. Of course, an injury investigation has to start somewhere and making it clear where that is—and explaining the choice—puts a human performance investigation in a proper framework.
- Learn what goals people were working toward at the time of the incident. These often directly connect to how the process and events were unfolding around them. Ask workers, “What were other people in the operating environment doing?” People who work together on common goals often divide the necessary tasks among them in predictable or complementary ways.
- Take a new attitude into investigations. Don’t look for an accident’s “cause.” Instead, search for mechanisms by which your organization is allowing failure to succeed.
Finally, to the extent possible, retail safety programs should aim to be proactive—not just to identify problematic human factors after they trigger events. For example, supervisors, safety committees, or safety work teams might undertake a concentrated effort to identify which job activities workers can perform in a risky manner, how often workers succumb to risk-taking while conducting them, the potential for risk-taking to result in injury, and strategies to reduce risk-taking associated with the activities.
This article originally appeared in the December 2018 edition of LPM Online. You may see the animated version at LPM-online.com.
