Professional Safety Journal Asks Chuck Pettinger
Your company specializes in helping companies predict injuries before they occur. What are predictive analytics and how they are being applied to workplace safety data?
Predictive analytics is the practice of collecting raw data (often in very large quantities), and turning that data into actionable information. Based on this information, inferences about future outcomes, or predictions, can be made. By predicting workplace injuries, they can be prevented from ever happening!
How do you know the predictive models developed based on these data work?
Safety experts here at Predictive Solutions along with artificial intelligence experts from the School of Computer Science at Carnegie Mellon University (the team that helped build Jeopardy's Watson and IBM's Deep Blue), have been analyzing our 100 million pieces of workplace safety data across 15,000 workplaces and building predictive and advanced analytics models. The models have proven to be very effective with accuracy rates of 85 percent with and R2 correlation of 0.75 in predicting incidents.
How can SH&E professionals help management make better decisions and avoid decision traps?
I have been doing change management consulting for many years and the answer to this varies based on the culture. In the best cultures in which I've worked, simply providing actionable information is sufficient to elicit the desired change. In other cultures, it's all we can do to have senior level leadership mention safety. In either case, however, we need to be more prescriptive with the information we provide our leadership team. If we simply ask managers to support safety, then we are not going to make a difference. One of my clients with an advanced safety culture took the guess work out of leadership support; they made leadership engagement activates part of their yearly "scorecard." From the president down to the supervisors, they all had to perform one of 15 engagement activities per time period. These activities ranged from simply attending a safety meeting to participating in change efforts. These were tracked, trended and fed up the leadership ladder. This elicited much behavior change.
Observations are a key element of most behavior-based safety processes. What are some recommended practices for ensuring that the quality of an observation is being measured objectively?
First of all, feedback is THE key element of a BBS process. To me, the data gathered during the observation is "icing on the cake." Over the 20 years of researching (NIOSH grants on BBS, a dissertation on BBS), implementing and now helping sustain BBS processes, I see a variety of initiatives people are calling behavior-based safety. Similar to many popular management initiatives (TQM, Lean, 6 Sigma), well-intentioned SH&E professionals read about BBS, try to get the funding to implement proper change initiatives and fail to get the needed resources (unlike the funding put up for management initiatives). So, the eager safety professional, on a meager budget, changes the current inspection checklist and adds a few behaviors to it and calls it behavior-based safety. They try it, fail and say BBS doesn't work.
There are many ways to ensure quality in BBS:
- Teach people the difference between an inspection and observation. Inspections focus on rules and regulations; observations focus on behaviors. It is possible to follow every rule and regulation and still be performing risky behaviors. I still see "paying attention" on BBS checklists. This is not a behavior.
- Go out in observer pairs and conduct independent observation of the same employee, then compare results and calibrate.
- Add a "severity" metric to add more "color" to your risky observation. If you see someone doing something risky, I have my clients rate that risky behavior as low, medium, high or life threatening. BBS processes get so few risky behaviors, that I want to glean as much data as possible. Once you have this, you can start to look at how frequently people record medium- and high-severity behaviors, or are they all recording low-severity PPE use?
- Teach people how to deal with confrontation. People avoid it and this is the main reason BBS data are artificially high in percent safe.
- Look for a pattern of 100% safe observations. I am not saying it's impossible, but if one person (or department) never records a single risky behavior, that is telling you something about the culture.
- Look for the percent of observations with comments.
- Identify your best observers (the ones you know are doing "real" observations) and use these people as your comparison group. Now assess how the rest of your observers compare to your exemplars.
- Make data entry easier. If it takes too long to enter the data, via a paper form, computer or phone, people won't put down as much observation intelligence as they could or would.
- Even though the benefit is in the conversation between employees, we still need to use the data and get the leadership team to publicize the successes and use the observation intelligence to make visible changes. If employees know their data are being used, this will increase the quality.
What are some effective strategies for getting employees to participate in the observation process initially? Strategies to sustain that initial interest level?
To me, it's all about sustainability. Many change initiatives attempted by organizations fail to make a sustainability plan and declare victory too soon. So, once again, use the data! If there is once crucial flaw in BBS initiatives, it's not planning for sustainability by making and using a "data-use plan." Once you collect the data, what is going to be done with it, by who and how often? With our busy work life, getting management and employees involved and engaged is a value proposition. If we ask a manager to support our new BBS initiative and don't provide them with any good safety fodder, it will fail. We are all busy and if people don't use the data and communicate positive changes, the BBS initiative will wither into the flavor of the week. It's a value proposition. We have to demonstrate the value these observations are providing the organization, this will motivate the leadership team to visibly support this change and then employees will see the value in completing quality observations.
What are some common roadblocks to acting on trends that are revealed through these observations?
One great roadblock is that it points the finger back at organizational weaknesses. I know that some unions feel that BBS is "blame the worker"; in some implementations, it might well be, but only because the process wasn't implemented correctly. So, if I get this observation intelligence on my safety processes--will this make me, my department or site look bad? So, don't overreact to a risky trend in data. Use this as an opportunity to identify which processes are not working as intended.
The other roadblock is that too many BBS processes focus on percent safe. In healthy BBS processes, I want to start to see the number of risky behaviors increase. This doesn't mean that we are having any more risky behaviors, it simply means that our culture is opening up. Now that I haven't been fired for recording something I thought was risky, not gotten my head chewed off for giving the 40-year tradesman feedback and seen management truly demonstrate leadership by using the information in a proactive, leading indicator way, I just might increase the quality of my observations.
How can companies best overcome those roadblocks?
Make sure that people know the why, not just the how. I think we put people through too much training without giving them that vision of WHY we are doing observations, for instance. Safety professionals and leadership trumpet incident rate as a safety statistic. It's NOT a safety statistic, it's an injury statistic. If we don't have any injuries, does that mean we are totally safe? We need to make safety personal again. It's about people going home to their families and loved ones, not about a 1.23 recordable rate. If we want to reduce roadblocks to culture change, try not mentioning an incident rate to employees, ever. Most employees don't know what the recordable rate means and they don't care about it. They care about their friends and coworkers getting hurt. Our vision at Predictive Solutions is to eliminate death on the job, in this century. That is the "why" I go to work every day and the "how" is using predictive analytics to provide actionable information to organizations to get up-stream of the injury.