The Opportunity
- Two companies were looking to identify developing trends in safety incidents with their employees, across regions and business units. They wanted to find leading indicators of a change of recordable or lost time incidents.
Our Approach
- For both companies we developed time series modeling methods that excel at both forecasting incident rates and alerting to changes in the underlying trends. These models are able to do so for thousands of low levels of aggregations, such as working locations, business unit, and job function.
The Impact
- For both companies, the task of navigating a large spreadsheet or BI report with lots of data did not lend itself to focusing their attention on areas where problems were starting to develop. The models were (1) able to alert to those contexts and (2) forecast trends out into the future, should no remedial action be undertaken.
- In the case of one company, as part of a knowledge transfer engagement, we taught their analysts how to build and maintain these models internally.