26 Introduction to Variation

W. Edwards Deming, the famous quality improvement guru, claimed that the two most important things for managers to understand are:

1.     Variation and how to deal with it

2.     The forces that motivate and demotivate people

The subjects of the first 21 lectures, motivating, staffing and communicating, address the forces that motivate and demotivate people, i.e. the Theory Z portion of effective leadership. Forces mean the collection of perceptions, understandings and misunderstandings that influence the attitude and behavior of people. Lectures 23 – 25 introduced management of processes, part of the control function of managers, and treated the stand alone topics of managing risk and theory of constraints. Now we turn to variation and how to deal with it, the central theme of process improvement and process control. Managing in the presence of variation is also part of the control function of managers.

W. Edwards Deming claimed that the inability to interpret and use the information in variation is the main problem for managers and leaders. (See the book Out of the Crisis by W. Edwards Deming) When there is a problem with any work process the manager and the employees both must understand when the manager must act and when employees must act. It is through an understanding of variation and the measurement of variation that they understand when and who should take action and, just as importantly, when not to take action. Thus variation is involved in both improving poor processes and maintaining good processes.

Variation is just the reality that actual values of parameters, physical or financial, have some statistical spread rather than being exactly what we expect, specify or desire. For example, we may have a budget for supplies of $1000 per month. When we look at spending for each month it is typically close to but not exactly $1000. Over time the spending might look like that shown in figure 15.

Figure 15. An example of variation from planned budget by actual spending.

For our purposes the definition of variation is deviation from planned, expected or predicted values of any parameter. The parameter might be financial, as in the example shown in figure 15, it might be in units of production per day or minutes per service, or it might be a physical parameter, such as the dimension of a machined part. Thus variation occurs in all the work processes of any kind of organization. Therefore, as Deming implied, the effective leader must understand the information in variation and how to properly manage in the presence of variation.

Let’s start by returning to the work process illustrated in figure 12, the SIPOC diagram.  Where might we expect to see variation in a work process? The answer is everywhere. Deviations from ideal inputs are variation. Deviations from ideal outputs are variation. Deviations from expectations in use are variation. Variation in use can be due to either hidden variation in outputs or unexpected variation in the use environment or the use process.

Let’s define an effective process from a customer’s point of view. It is a process that produces outputs that meet or exceed the customer’s expectations for quality and cost. Customers can be internal or external to the enterprise or the organization that owns the process. Customers have stated and unstated expectations. Specifications, requirements, standards, and contract items are examples of customer’s stated expectations. Customer’s unstated expectations are typically suitability for all conditions of use and affordability. Therefore, for the purposes of process improvement discussions, we can say that an organization’s effectiveness is determined by the effectiveness of its processes in satisfying its customer’s expectations. (In general the effective organization must satisfy all its stake holders’ expectations, including managers, workers, owners and the community as well as the customers.)

Variation Drives Process Effectiveness

We can see the effects of variation by examining an ideal business process (figure 12, an ideal process is repeated in the top half of figure 16) and a typical process as shown in the bottom half of figure 16.

 

Figure 16. Comparison of a typical process to an ideal business process.

An ideal process converts all of the supplier’s inputs to outputs that satisfy the customer’s expectations. A typical process includes inspection steps to ensure that a defective input is not sent to the process or a defective output is not sent to the customer. The customer also adds an inspection step because of receiving defective outputs in the past. If outputs fail any of these inspections the failed item is scrap or must be reworked. It’s easy to see that the typical process is more expensive, and therefore less effective, than an ideal process because inspections cost money and scrap or rework cost money. In a typical chain of processes costs of failing inspection increases as the work progresses along the chain because more rework is required if an inspection is failed at processes near the end of the chain. Thus often the largest cost to the organization is warranty costs from customer returns. That is the reason for the inspection of the outputs before they are sent to the customers. The reason these inspection steps are added is the presence of variation. If there was no variation in the inputs or the outputs then there would be no need for inspection to find those items whose variation from ideal is larger than acceptable.

Notice that even the ideal process has inputs and outputs that exhibit variation but for the ideal process this variation is within acceptable limits most of the time. We need to define what we mean by “most of the time”. If there is variation then sooner or later a product will fail to meet customer expectations if there is no inspection. (Actually it will happen even with inspection since no inspection is perfect, i.e. inspection is a process that also has variation.) If the variation is small enough so that only rarely is there a customer return and the cost of correcting this return plus the cost of the disgruntled customer is less than the cost of including inspection then it makes business sense to not have inspection.

Now I hope the student is thinking that to make a valid decision to not include inspection takes data to establish that the variation is sufficiently low. The astute student is also thinking that collecting such data costs money also, perhaps as much as the inspection. This is an example of what is meant by a manager needing to know how to manage in the presence of variation. Next we examine how a manager can achieve such understanding and make good decisions in the presence of variation.

Variation is a Statistical Phenomenon

To understand managing in the presence of variation we must answer the questions how can the manager decide:

·       when to take action,

·       what action to take and

·       who should take the action?

Managing correctly in the presence of variation requires the use of methods based on statistics since variation is a statistical phenomenon. The statistics needed for 85% or so of a manager’s work is relatively simple and easily learned. The effective leader and all workers must understand and use these simple methods. However, there are situations that require more elaborate statistics. Every organization should have access to at least one person well versed in statistical methods so that managers and process improvement teams have a resource to check their work and assist on complex problems. This statistical expert can be a consultant or a worker that is well trained in statistics.

Here we are going to briefly look at some of the most important simple methods. As an example, figure 17 illustrates the daily averages of phone expenses for an organization plotted for each month of a year.

Figure 17 A graph of an organization’s daily phone expenses averaged for each month of a year.

Should the manger take action in response to the March expenses? The June expenses? If action is necessary in response to the March expenses, whose action is it? The manager’s? The workers? If the manager is expected to discuss unusual expenses in a weekly or monthly report what should the manager say about the March and June expenses?

Control charts are a visual method of answering the questions posed about the phone bills. A control chart for the phone expenses data from figure 17 is shown in figure 18. You can learn how to generate control charts later. For now I only partially describe how to interpret the data in a control chart.

Figure 18 A control chart for the example phone expense data.

The line with diamond markers is the same data shown in figure 17. The line with the square markers results from averaging the data over a whole year. The line with the triangle markers shows the range of variation of daily expenses for a given month. The two lines labeled Upper CL and Lower CL are upper and lower control limits, which are statistically determined from the data set. For the purposes of this introduction it isn’t necessary to know how to calculate the control limits. The control chart tells us that, with the exception of the March data point, the phone expenses are stable, that is they exhibit variation about a stable sample average, which is not steadily increasing or decreasing. A stable process is predictable, e.g. frequency of errors, efficiency, process capability and process cost are predictable. Deliberate changes to a stable process can be evaluated.  Note that some process improvement literature refers to a stable process as being “in control”.

Variation exhibiting a stable statistical distribution is due to the summation of many small factors and is called common cause variation. Changes to a stable process, i.e. one with common cause variation is typically the manager’s responsibility but can be the responsibility of trained and empowered workers. Knowledge workers should be responsible for common cause variation because they are usually more expert with respect to their processes than their managers. However, as is described in the next lecture, even knowledge workers should not be empowered to control their processes before they have been trained in statistical methods because mistakes can make processes worse.

Only the data point for one month, March, falls above or below the two control limit lines. Variation that is outside the stable statistical distribution, i.e. above the upper control limit or below the lower control limit, is special cause variation.  The point for March falls below the lower control limit. This means that the March data is special cause variation. Special cause variation is the workers responsibility; they typically know more about possible causes than the manager because they are closer to the process. But the workers need training in problem solving to fix special cause variation and they need to be empowered to make fixes to their processes.

The workers should review the data for March and examine the phone system to see if they can determine the reason the daily averages were so low. For example, the phones may have been out of order for a week, which would have lowered the daily expenses but require no action other than getting the system operating again. Properly trained and motivated workers can handle special cause problems, usually without any management involvement.

A stable process is a good candidate for process improvement. The goal of process improvement for a stable process is to reduce the variation and/or change the mean. Process improvement should not be attempted on a process that is unstable until the process is brought to a stable condition because changes in data taken on an unstable process cannot be uniquely attributed to the action of the process improvement. The special cause variation that makes the process unstable must be removed before beginning process improvement.

Note that the control chart also provides the manager information useful in considering process improvement. In the example shown in figure 18 the yearly average phone expenses are about $21 per day. A manager can evaluate the cost benefit of making a change to the phone service based on this data since it is stable over a year. If the manager can make a change without investment that promises a 10% reduction in phone expenses the manager can see that data will have to be monitored for about four to six months to determine if the mean daily expenses do indeed drop from $21 to $19 because the normal range of variation in monthly averages is larger than the expected change. However, if the change really works as promised then in about four to six months the monthly averages should begin to vary about a new long term average and the control chart will show this change.

Exercise

1.     Go to “Control Charts” in Wikipedia (http://en.wikipedia.org/wiki/Control_) and read the article. This material expands upon the introduction given in this lecture.

2.    Go to http://www.goalqpc.com/shop_products.cfm and buy yourself a copy of Memory Jogger II. This handy book teaches everything you need to know about problem identification and problem analysis. It is small enough to carry in your pocket and it is your guide to the details of process improvement. If you prefer a spiral bound version it is available from Amazon.com (Michael Brassard, and Diane Ritter, The Memory Jogger II: A Pocket Guide of Tools for Continuous Improvement and Effective Planning) There is also a Six Sigma Memory Jogger available.

The Memory Jogger book recommended here is so widely used and so effective for the practical user that there is no point in repeating the material in this course. The student is expected to study the Memory Jogger and put the techniques into practice. This means that the student and all the people reporting to the student are to have the Memory Jogger book , or an equivalent, be trained in the techniques summarized in the book and put these techniques into practice. This is essential if an effective organization is expected. The exception is if your organization is following the Six Sigma approach where only selected people are highly trained.

If you prefer not having to learn statistical techniques yourself you can attend training if your budget and schedule permits. One example workshop in statistical process control is offered by the American Supplier Institute. See: http://www.amsup.com/spc/1.htm. This workshop focuses on manufacturing but the techniques work for any type of organization. A web search reveals many other training organization offering similar programs. I have found it more cost effective when training all workers to bring the trainer to the organization rather than sending workers to outside training.

If you find that the pace of blog posts isn’t compatible with the pace you  would like to maintain in studying this material you can buy the book “The Manager’s Guide for Effective Leadership” in hard copy or for Kindle at:

http://www.amazon.com/Managers-Guide-Effective-Leadership-Organizations/dp/1449000673/ref=sr_1_2?ie=UTF8&qid=1346946310&sr=8-2&keywords=Joe+Jenney

or hard copy or for nook at:

http://www.barnesandnoble.com/s/Joe-Jenney?keyword=Joe+Jenney&store=book

or hard copy or E-book at:

http://bookstore.authorhouse.com/Products/SKU-000269270/The-Managers-Guide-for-Effective-Leadership.aspx