Canadian Occupational Safety (COS) magazine is the premier workplace health and safety publication in Canada. We cover a wide range of topics ranging from office to heavy industry, and from general safety management to specific workplace hazards.
Issue link: https://digital.thesafetymag.com/i/358706
T he same predictive analytics techniques that have been used in other business functions like finance, sales and marketing, and customer loyalty are now being employed by safety professionals to predict and prevent workplace injuries. A research study by Carnegie Mellon University (CMU) showed that these safety prediction models have accuracy rates as high as 80-97%. Companies that employ these predictive analytics tools have achieved dramatic results: • A Fortune 150 energy company reduced its injury rate by 67% within 18 months • A Fortune 150 manufacturer reduced its lost work day rate by 97% within one year • A top 20 construction company achieved significant safety improvements including 90% of worksites experiencing no lost-time incidents Big Data, Big Computers, anD moBile teChnology Several factors have aligned to allow for predictive analytics in safety. First, data is being collected at a tremendously high rate. According to IBM, 2.5 quintillion bytes (that's a 25 with eighteen zeros after it) of data are created daily. Safety data collection is contributing to this phenomenon. One multi-national company with significant Canadian operations records nearly 2 million safety observations per year. Supporting this big data phenomenon is the rapid adoption of handheld and mobile computers, driven by the iPhone and Android-enabled devices. This rich and growing data set collected in the field and on the shop floor can be immediately uploaded into databases that fuel predictive analytics systems. Second, computing power and storage continues to grow exponentially. Moore's law, stating that the number of transistors on integrated circuits doubles approximately every two years, has held true. This ever-growing computing capability allows for extremely advanced analytics on big data sets called machine learning. Machine learning is the field of study that gives computers the ability to learn without being explicitly programmed. Machine learning techniques are used to create predictive models. how the safety preDiCtion moDels are Built anD work CMU helped build some of the only predictive models known to be in use in safety. They did this by feeding computers armed with machine learning techniques four years of real world safety data – mainly safety observations and inspections, as well as safety incidents. The computers, after reviewing millions of data points, identified patterns in the data that are not discernible by humans. After several "training" or "learning" sessions, the computers built a model that could predict the number of safety incidents that would occur at a location over the next 30 days, based solely on the safety inspection and observation data collected over the last 90 days. These models proved to be 80-97% accurate. After the model was fully trained, tested, and was predicting accurately, it was put into use within a broader safety software system. Once put into production, all that the model needs to predict incidents over the next month is safety inspection and observation data from the last three months. the Benefits of preDiCtive analytiCs in safety Rather than simply reacting to lagging indicators, or trying to track and interpret numerous leading indicators, predictive analytics jumps straight to the front of the class and allows safety professionals to answer the question, "what will happen next?". As you can see in figure below, according to Tom Davenport, one of the foremost experts on analytics, predictive models are the highest form of advanced analytics allowing business leaders to answer their most difficult business questions. Basic data access and reporting techniques, readily available in fairly simple data software systems like Microsoft Excel, only allow safety managers to track what has happened in the past – usually termed "lagging" indicators. Only when safety professionals move up the pyramid into advanced and predictive analytics can they start to identify leading indicators and predict future outcomes. Once they can predict where and when their safety incidents will occur, they can optimize their response to those predictions and prevent injuries from ever occurring. Employing advanced and predictive analytics allows companies to take a proactive versus reactive approach to safety. No longer are these advanced analytics techniques confined to other business functions like finance, sales, and marketing. They can be used to not just generate higher profits, but to save lives. So stop investigating workplace injuries, and employ advanced analytics to predict and prevent them. Griffin Schultz is the General Manager of Predictive Solutions Corporation (PSC), formerly DBO2. Griffin has an MBA from the Wharton School and has extensive experience using technology to help improve business outcomes. For more information, contact Griffin directly at gschultz@predictivesolutions.com stop investigating workplace injuries… start predicting and preventing them! BROUghT TO yOU By