Mountains of available data can describe how workplaces have or have not become safer over time. From OSHA to the U.S. Department of Labor to the CDC, you can find detailed historical information on serious injuries and fatalities (SIFs), days-away-from-work injuries, work-related illnesses and much more.

What this data can’t tell you – at least, not by itself – is what might happen next week, or next year, in your facility. But by marrying this historical incident data with other relevant measurable factors, and then applying it in the right contextual tool, you might be able to get ahead of incidents before they happen.

This is predictive analytics. Investopedia defines it as using “current and historical data patterns to determine if those patterns are likely to emerge again.” The key word in that definition is “current,” and the explosion in real-time manufacturing data through the Internet of Things (IoT) provides opportunities that previously didn’t exist to create forward-looking models for a range of manufacturing topics, including workplace safety.

But data can overwhelm as easily as it can inform. Improving safety through predictive analytics requires a systematic approach to achieve desired outcomes.

Understand Your Goal

An article in HVACR Business suggests to first gather all the relevant stakeholders to discuss the outcomes of a predictive safety analytics approach. This includes not only the obvious goal of reducing incidents, but also what the program should look like, who should be involved and how the organization will act on the learnings.

Choose Your Data Points

At this step, your goal can help you decide what data is useful, and just as importantly what is not. A mix of leading and lagging indicators, along with real-time data, will be part of your predictive safety analytics recipe.  

Just a few examples of data points to consider collecting include:

• Previous reported incidents
• Training history
• Employee schedules
• Employee behavioral surveys
• Weather data
• Equipment sensor data
• Smart PPE data

Even observational data from employees can be incorporated. Canadian Occupational Safety reported on one company that urged employees to report potential safety hazards through the facility on a mobile app, resulting in thousands of data points. An added benefit of this approach, it was noted, was the result of spreading a safety-first culture throughout the organization.

Select a Software Tool

Predictive analytics works through machine learning. Pen-and-paper recording or even simple spreadsheets are insufficient to provide context to your data and bring you unexpected insights. You’ll need a predictive analytics software tool to convert the information into action.

Factors to consider when selecting a predictive analytics tool include:

• Integration with existing platforms
• Ability to visualize data
• Mobility
• Security
• Scalability
• Cost

Act and Evolve

Predictive analytics is about identifying the why, rather than the what, when it comes to safety incidents. The next step is to adjust your approach based on the analytics, measure results and continue to adjust.

HVACR Business noted a case in which data showed apprentices and employees within 90 days of their start date were connected to the most, and most serious, safety incidents. It allowed the company to adjust its training and emphasize the need for field leaders to reinforce that training.

According to Canadian Occupational Safety, sharing results of the modeling with employees is important for creating buy-in and building a safety culture. They will understand why changes are being made, and they will be able to see if results are achieved.

For more assistance in improving safety at your workplace, check out Grainger's Managed Safety Programs.