HOW DATA CAN TAKE
the NOISE out of the
OVERVIEW | A large manufacturing plant was concerned about safety. The work force had more than 5000 people. There were over 20 different departments. External and internal policies required a thorough analysis and report for every injury. Over a 10 year period the company had many records with detailed data about injury demographics, conditions, area, date and time, shift, overtime and other factors. This database was used to generate reports required by management and external agencies.
CHALLENGES | There was a safety oversight committee with ranking personnel from both union and management. There was a good working relationship. But, issues that surfaced tended to be ad hoc, random and generally hard to "get your arms around." When they existed the specifics about issues were almost always more qualitative than quantitative. There was a lot of emotion. This was NOT surprising given the consequences to the workforce. There was a lot of "noise." There had been approximately three major initiatives to "improve safety" over the years. Interestingly, there was an improvement with each initiative. Also, interestingly, when these initiatives ceased the injury frequency rose to the level prior to the initiative's beginning.
THE SOLUTION | There was a concerted effort launched to explore the very comprehensive database. Seasonal patterns, trends, special causes and many systemic themes quickly emerged. They had been locked in the data for many years. It was discovered that multiple departments had some of the same key issues, and some issues were revealed for the first time. Focused improvement efforts across multiple departments were started. The week-to-week and month-to-month variation in injury rates was huge, and analyzing trends was virtually impossible. But, a 12-month moving average became a norm for safety statistics, and the actual trends instantly were seen. This data was analyzed using Microsoft Excel's pivot tables and several statistical programs. Individual departments then had a way to quantify, prioritize and plan highly focused efforts. This did not directly reveal the "root causes," but knowing where to look in a methodical manner did. Pareto charts, Cause-and-Effect diagrams and a Plan-Do-Check-Act approach replaced the previous qualitative approach. When the data started to be used, a number of "bad entries" emerged. These were quickly corrected, and the accuracy of the data automatically improved with use.
OUTCOME | The safety statistics improved to previous "best" levels, and kept improving. The "noise" went away. Communication and behavior changed after making the data and Microsoft Excel widely available. The net impact was more than 100 fewer injuries each year. The reason was known. It wasn't related to a specific named program that may or may not continue. There was a cultural commitment to continuous improvement, and a tracking mechanism to monitor the effort.