The Behavioral-Based Safety programs currently in use on most project sites are focused on making observations or completing checklists in order to identify unsafe behaviors and modify them. The goal is to make the project site, and the workers on it, safer. While this approach has worked for decades and has brought contractors’ incident rates down significantly in the process, the same tools that got us here won’t take us to the next level of the safety performance required to achieve the coveted goal of zero incidents.
This is where Predictive-Based Safety (PBS) comes in. PBS harnesses the power of technologies like big data, predictive analytics and artificial intelligence (AI) to collect and analyze project data with the goal of identifying and reducing risk.
Predictive-Based Safety Defined
Predictive-Based Safety involves the collection of job site data from a variety of sources including interactions with craft workers for the purpose of applying prescriptive analytics. The data is analyzed by artificial intelligence and results in a list of actions that project teams can take in order to change behaviors and improve safety. Predictive-Based Safety is a compliment to Behavior-Based Safety techniques.
Where Does the Data Come From?
Data is at the heart of Predictive-Based Safety. Data comes in many forms and from many sources. One must weed out the sources of noise to find the true indicators of risk that drive safety. Each data source serves as a channel to strengthen the predictive model. Some of these include:
Safety Observations
If people and behaviors are at the heart of Behavioral-Based Safety, then observation data collected on people and behaviors is the life blood of Predictive-Based Safety. This is why a rock solid safety observation program is imperative to a high-performing Predictive-Based Safety program. Safety observations of yesterday, often collected on paper, are transformed into data-rich interactions where project staff and craft engage with one another to improve safety conditions and behaviors together. Using tablets or smartphones, the data from these interactions are collected and stored in records for analysis. When collected in the right way, safety observation data can be AI-ready the moment a user hits the “save” button.
Photos and Video
Photo and video data from a variety of sources including progress photos, quality inspections, daily reports, and from the safety observations mentioned above are evaluated by AI to spot subtle indicators of risk that previously went undetected. Because AI makes observations automatically and with no bias, these insights from visual data uniquely complement and augment observations made by humans.
Traditional Project Data
Project data, such as schedule, cost, and man power levels that previously resided in silos, is now available for predictive analytics thanks to pre-built integrations that connect the myriad software tools used every day on job sites.
Predictive-Based Safety in Action
These streams of data are combined and turned into analytical models that seemingly give project teams super powers. Teams are not only able to see that they are at risk of having an incident, but they are able to pinpoint the actions they must take to prevent one from taking place. Building on their training and experience in executing Behavioral-Based Safety, these same teams engage the craft workers more effectively. They share insights and findings from the data enabling the craft workers to correct a condition or modify a behavior before an incident occurs rather than make the same change based on an incident review meeting.
From a management standpoint, dashboards containing key metrics and ratings pulled directly from the same streams of data are accessible to every level in the organization. This democratization of metrics provides transparency and enables each business unit to recognize high-achieving projects and individuals or to formulate action plans to address any areas of underperformance.
Final Thoughts
Although Predictive-Based Safety is a new approach, the results are undeniable. Companies on the leading edge of adoption have seen prediction accuracies as high as 86% and commensurate reductions in the occurrence of recordable incidents as high as 60%.
How would results like these impact your organization?
Check back regularly for new articles from our Predictive-Based Safety series including:
- What makes an effective Predictive-Based Safety Observation?
- Is my company culture ready for Predictive-Based Safety?
Newmetrix (formerly Smartvid.io) enables the AEC industry to significantly reduce jobsite risk by combining the best of human and artificial intelligence. Newmetrix's Safety Suite includes Safety Observations, Safety Monitoring and Predictive Analytics products that combine to give project teams and management the ability to predict and prevent incidents - saving lives and preserving timely project delivery.
Our AI engine, Vinnie, has been trained to recognize construction risks in photos and other project data creating an unbiased, automated risk assessment that enables teams to have better visibility into risk. Our Safety Observations product combines an easy-to-use mobile application with risk scoring and workflow that enables the entire company to engage in gathering safety data. The Predictive Analytics product analyzes both the AI-based Safety Monitoring and Safety Observation data sources, in addition to other project data, to create prioritized project rankings so teams know where to focus attention. By implementing this Predictive-Based Safety approach, our customers have seen reductions in incident rates of 10-30% or more. Getting started is easy with pre-built integrations to Autodesk BIM360, Procore, Oracle, Egnyte, Box, and other data sources.
