In businesses where the owners or employees have little experience with the many tools that Six Sigma offers, run charts are one of the few analytical tools employed. It is not uncommon for a Six Sigma professional to dismiss run charts as relatively weak analytical tools as they become enamored of the more sophisticated statistical tests and models. Despite their relatively simple nature, run charts can be a valuable tool in the early stages of Six Sigma project selection. Perhaps more importantly, they can be helpful in teaching colleagues about the value of more powerful tools like statistical control charts and trend analysis.
A run chart is simply a chronological line graph of a process characteristic–the X axis identifies the product or characteristic in chronological order and the Y axis represents the value for the specified product or characteristic. For example, I might measure the number of customer support calls for a particular type of issue I receive each day over the course of sixty days. The X axis would consist of days in chronological order and the Y axis would represent the number of calls for the specified issue.
At my company I received a request to track the variance of various types of technical issues experienced by our customers over time. The goal was to enable the operations team to watch the trends over time and address the issues that trended up.
I set to work and created a simple web-based tool that plots simple run charts for the top 100 most common issues experienced by our customers. I showed a prototype to the operations team who was satisfied that this tool would make the question of which issues to address much easier.
And for the first few weeks they were satisfied. We quickly found that looking through all 100 plots was time consuming and tedious. They then wondered whether there was a way to simply mark any of the issues that were trending up or down. This was a fantastic opportunity to introduce the power of statistical process control charts and simple trend analysis.
I proposed that we could identify spikes by marking single data points that fell three standard deviations above the mean in red. I implemented this in the web-based report and the operations team was again delighted–now that had a more sophisticated tool that would quickly tell them whether any issues had increased to the point that it was out of “statistical control.”
A few months later, the team realized that there were still some limitations to the report. If a particular issue increased only slightly over time, the report would not detect the trend unless they looked at each and every run chart. This may sound like a finicky complaint but looking over 100 run charts daily becomes quite tedious and time consuming. This was the perfect opportunity for me to introduce the concept of statistical correlation.
I added to the report a correlation coefficient for each issue series along with a probability statistic. I also calculated a regression line for each issue and displayed its slope. I then added color alerts to any cases where the probability was less than 0.05.
With this final addition, the operations team was now also able to quickly identify more subtle upward trends. They have been quite satisfied and effective at choosing how to spend their time.
While run charts are not the most sophisticated of analytical tools, they provided me with tremendous opportunities to arm our operations group with more useful and effective information.