Not all variation is created equal. Planned variation, like that in an experiment, is a process improvement strategy. Unplanned variation, however, is nearly always bad.

Two types of variation concern a Six Sigma team:

  1. Common cause variation – All processes have common cause variation. This variation, also known as noise, is a normal part of any process. It demonstrates the true capability of a process.
  2. Special cause variation – This variation is not normal to the process. It is the result of exceptions in the process environment.

In a process improvement project, special causes of variation should be the first target. Eliminating special causes of variation brings the process into a state of control and exposes the sources of common cause variation. The next step  is then to reduce common cause variation.

Typically, the processes or process steps most suspect for introducing variation are those that are manual or judgment oriented in nature. If an individual applies personal judgment within a process, you would expect to see bias or higher variation in the process output. Automated processes typically have more consistent performance and lower variation.

Fuel Feed System Example

Consider a project on a fuel feed system, in which there was too much variation in the concentration of petcoke (a carbon product) being blended with coal and burned in a plant’s boilers. The result was either 1) too much petcoke, resulting in a violation of environmental parameters or 2) too little petcoke, which increased the cost of operation.

Fuel Feed System
Fuel Feed System

The fuel feed system was comprised of two conveyer belts that fed a third conveyer belt. One of the feeder belts fed coal and the other fed petcoke. The third belt, called the silo belt, fed the boilers. The concentration of petcoke loaded to the boilers was controlled by the belt speed of each of the first two belts.

The silo feed belt was designed to start empty and, as a result, was the last belt to be shut off so that it would be empty when stopped. The other two feed belts were designed to start empty or full.

Process Mapping

One of the best ways to find these manual or judgment-based steps in a process is through the use of a process map. As a process is mapped, decision points are represented as diamonds. These are the first places to look for variation.

When mapping a process, information both from the process owner and from the Six Sigma team’s observations are used. There are situations in which the process as described by the process owners is as-we-think-it-is or as-it-should-be view; a process owner may know the standard process, but chooses not to not follow it. A Six Sigma team, however, must focus on the as-is world.

The standard process for starting up the fuel feed system was to start the silo feed belt first, the coal feed belt second and the petcoke feed belt last. All belts were to be empty when started. The shutdown process required that the petcoke belt be shut down first, after it was empty. The coal feed belt was to be shut down second, when it was empty. The silo feed belt was shut down last, when it was empty.

Analyzing the Problem

The first step in the team’s analysis of the variation issues was to compare the start-up and shutdown processes of each shift. The Six Sigma team did not assume that all shifts complied with the standard processes, since the computer systems allowed them to change the order of start-up and shut down.

What the team found was that one of the four shifts (Shift A) shut down the coal and petcoke belts full, and then shut down the silo feed belt when empty (see table below). When restarting the system, this shift would start the silo feed belt first and then start both the coal and petcoke belts simultaneously. This shift had the lowest variation in petcoke concentration (within variation) compared to the other shifts.

Another of the shifts (Shift B) would first shut down the petcoke feed belt full, then the coal feed belt full, then the silo belt when empty. On start-up they would start the silo feed belt first, then the coal feed belt and end with the petcoke feed belt. This group had the highest within variation in petcoke concentration.

The other two shifts (Shifts C and D) followed the standard process of start-up and shutdown. Their within variation in petcoke concentration was higher than the first shift, but lower than the second shift.

The process improvement team had learned that Shifts A and B did not follow the standard process of operation. The variation within each shift, however, remained within tolerance.

Belt Start-up and Shutdown Processes – Before
Start-up Shutdown Within Variation of Petcoke Concentration
Shift A 1. Silo (empty)
2. Coal and petcoke (full)
1. Coal and petcoke (full)
2. Silo (empty)
Lowest
Shift B 1. Silo (empty)
2. Coal (full)
3. Petcoke (full)
1. Petcoke (full)
2. Coal (full)
3. Silo (empty)
Highest
Shifts C and D 1. Silo (empty)
2. Coal (empty)
3. Petcoke (empty)
1. Petcoke (empty)
2. Coal (empty)
3. Silo (empty)
Middle

Finding the Source of Variation

The team’s next step was to match emissions logs with the petcoke feed logs over a three-week period. What they learned was that although each shift alone was within tolerance, the interaction of the different shifts created significant swings (between variation) in petcoke concentration. The use of three different methods for starting up and shutting down the fuel feed system was the root cause of the variation problem.

The solution was clear – to get all shifts to follow the same process. At a team meeting, representatives from each shift agreed to follow the first shift’s start-up and shutdown processes. This became the standard process for the plant’s fuel feed system.

After the agreement, the Six Sigma team monitored the fuel feed process for four weeks to measure the results. They found all shifts in compliance with the standardized process and a low between variation in petcoke concentration.  The low variation allowed the plant operations group to incrementally increase the petcoke concentration and thereby reduce the plant’s operating cost.

Key Takeaways

One conclusion that the Six Sigma team came away from this experience with was that compliance with standardized processes is higher when process owners are part of the dialog that creates the standardization. Processes are developed to serve the process owners (people), not the other way around.

Another conclusion the team made was that communication between groups needed to improve. Different groups within an organization need to understand why standardization is necessary and they need to know how to recover from unforeseen process upsets. In this case, what is the standardized process for start-up and shutdown when system maintenance required a different shutdown condition than normal?

The reduction in variation was accomplished through cooperation between different shifts and a common goal of improvement. The common goal was shared by the entire fuel management group, not just the Six Sigma team. The final takeaway for this Six Sigma team was that Six Sigma is about people, not math. Calculating variation simply gave the team the “what.” The “why” came from the use of people skills and leadership.

About the Author