Applying SPC in a Service Environment

Applying SPC in a Service Environment

Lon Roberts, Ph.D.
Roberts & Roberts Associates
Plano, Texas
Phone: 972-596-2956
E-mail: [email protected]
Website: www.R2assoc.com

Abstract
Statistical Process Control (SPC) has long been used to monitor and control the
performance of manufacturing processes. Its benefits in this arena are well known. By
comparison, little attention has been paid to the use of quantitative tools for monitoring
and controlling service processes, other than accounting data in financial reports and the
use (and abuse) of numerical ratings in personnel evaluations. This monograph discusses
the need for establishing objective measures to monitor and control service processes—
especially at the point where taking action can result in maximum benefit. It also
describes the use of SPC to accomplish this intent, including the practical realities and
considerations for doing so. It is assumed that the reader has at least a cursory
knowledge of the concepts of SPC.

Opportunity Knocks
Perhaps as a legacy of the industrial era, which began in the late 19th Century, it seems that most
of the emphasis to distinguish ourselves in the global marketplace continues to be placed on
product superiority. Don't get me wrong, I am a firm believer in striving for product
superiority—but this should not be held up as a "red herring" to avoid facing the reality that:

• 80% of the U.S. economy is linked to jobs in the service sector

• Customers are five times more likely to take their business elsewhere due to poor service
as opposed to poor products1

• Even in a manufacturing environment, only some 40% of the cycle time required to fill
an order is, on average, attributable to the manufacturing component—the bulk is
consumed by so-called "white collar" processes

When speaking on the subject of this report at conferences and other public events I sometimes
remark, perhaps a bit irreverently, that many of us see ourselves as being in the "widget
business" when in fact we are (or should be) in the total customer solution business. Seldom, if
ever, do customers in today's world buy a product without at least a tacit expectation that there is
some degree of customer support to back it up. If this is true for so-called "off-the-shelf"
products it applies at least tenfold to products where after market support is inextricably linked

www.R2assoc.com
© Copyright 1995 (revised 2005) Lon Roberts Page 1
 Applying SPC in a Service Environment

to the product itself. It applies even more so to products where the customer is intimately
involved in specifying and designing—or in other words, customizing—the product itself.

Considering the importance of service processes, coupled with the relative opportunity margin, it
is somewhat of a mystery that so little science has been applied to the design and management of
such processes. Some would argue that service processes are heavily reliant on the "human
factor" and therefore not compatible with the ideals of scientific management. Those who take
this position in the extreme are apt to say that management (specifically management of so-
called "white collar" processes) is an art, not a science. I am convinced, however, that it is not
an issue of one or the other, but both. The challenge is how to keep one point-of-view from
overriding the other. Unfortunately, managers often align themselves toward one extreme or the
other. Some would insist, for instance, "If you can't measure it, you can't manage it." Those on
the other extreme would argue that when it comes to services, "The most important factors can't
be measured." Since there an element of truth to both positions, one of the challenges in
applying SPC to a service environment is to prevent one point of view from totally dominating
the other.

The issue of balance is a prevalent theme in the discussion that follows. In any case, we should
not allow philosophical differences to mask the fact that there is tremendous opportunity to
improve the performance of our service processes—assuming we don't knowingly abuse or
unknowingly misuse the tools (including SPC) that might allow us to do so.

Hear No Evil, See No Evil

Despite the fact that service processes are the mainstay of today's economy, we still have an
inclination to let these processes fend for themselves. I submit this is largely due to the fact that
the work-product as well as the work-in-process (WIP) of such processes is intangible, and
therefore difficult to measure and monitor. It is difficult, for example, to objectively determine
how much progress is being made toward the completion of such tasks as "gathering
information" or "coordinating resources"—especially when the end-point and/or the success
from doing so in each case is not clearly defined. Consequently, upper management seems to be
content with the attitude, "If it's difficult to measure, why bother?" This condition is
compounded by the fact that many managers are uncomfortable dealing with "people issues"—
self-contradictory though this may be.

But, I firmly believe it is for this very reason that the opposite attitude should prevail. Because it
is difficult to build accountability into our service processes we should be ever more vigilant to
do so. Taking the opposite stance eventually leads to fostering highly inefficient processes that
are destined to this end, if for no other reason because the external environment—including
customer wants, needs, and values—are continually changing.

In today's competitive world, profit margins are such that few companies can ill-afford to sustain
inefficient processes. As one metaphor suggests, we are now flying closer to the ground and
therefore have less room for maneuvering, less margin for error. Sooner or later, companies that
ignore this reality will—if we are allowed to stretch the metaphor—face the unpleasant prospect
of experiencing "n" takeoffs and "n-1" landings.

www.R2assoc.com
© Copyright 1995 (revised 2005) Lon Roberts Page 2
 Applying SPC in a Service Environment

Law of Unintended Consequences

Related to the “difficult to measure and monitor” issue is the issue of what I call the Law of
Unintended Consequences. By this I refer to the fact that no matter how careful we may be in
identifying objective indicators of how a service process is performing, there is frequently a
downside to doing so. As it turns out, this "law" is especially relevant to service processes since
the human component is vitally significant to services. In other words, when there is a
breakdown in a service process, the problem can almost always be traced to human error or
human inadequacy. By contrast, when a manufacturing process goes out of control, there is a
good chance that the problem is due some mechanical failure.

As an example of the Law of Unintended Consequences in action, let's assume, that a certain
company believes it is important to monitor and control the time-duration of the service calls
taken by its Customer Service Representatives (CSRs). If minimum time per call is given top
priority, those who establish this standard should recognize the high possibility that the CSRs
will take excessive short cuts in handling customer problems. The outcome in this example
would ultimately contradict the intent of having CSRs in the first place—that is, to demonstrate
the company's genuine concern for its customers.

The previous example also highlights the fact that the Law of Unintended Consequences is
exacerbated by misuse of the data. If CSRs are evaluated on the basis of minimum time per call
or number of calls handled, it is unrealistic to expect them to spend time on a difficult problem
that a customer wishes to resolve. Bound by these criteria, CSRs have been know to defeat the
system by having the customer repeatedly call back rather than solve the problem initially.

To avoid falling prey to the Law of Unintended Consequences when planning your measurement
system and interpreting the results, I suggest the following:

1. Carefully consider the downside any parameter you wish to monitor in your service
processes. Anticipate possible ways that your "system of measurements" can be defeated,
undermined, or rigged. Also anticipate outcomes that are directly opposed to what you
wish to accomplish.

2. Take precautions to ensure your measurement system will be used to improve process
performance and certainly not penalize employees for desired behavior.

Taxonomy of Service Processes

There are, perhaps, numerous ways to classify service processes. Table 1 describes at least one
method of classifying such processes that I find to be beneficial from the standpoint of being able
to clearly identify the process boundaries and the critical links to the customer as well as the
system-at-large that a particular process supports.

Why should we concern ourselves with classifying services? Let me respond by way of
example—in this case involving certain training services. Training could, for instance, be

www.R2assoc.com
© Copyright 1995 (revised 2005) Lon Roberts Page 3
 Applying SPC in a Service Environment

bundled with a new computer system or else be purchased as a stand-alone service through an
independent, third party provider. In the first case the process boundaries may extend from the
point of where the system is purchased to the point where the training is delivered. In the second
case the process boundaries might extend from the point where the training is purchased to the
point where the training is delivered. The intra-process interfaces, customer expectations, and
system of controls are necessarily specific to the application and circumstances, regardless of any
similarities in the expected outcomes in each case. This suggests that, when applying SPC in a
service environment, among other considerations it is necessary to be crystal clear on whom the
service will support and what the expectations are from point-to-point within the process.

Classification Description Examples

Pure Services that are an end unto • Overnight courier

Services themselves. In essence, the service
service is the product.
• Lawn maintenance

Integrated Services that are "bundled" • New automobile/

Services with a certain product. In warranty service
this case, the customer is package
interested in the overall
package–typically involving • Software/telephone
a product and service support package
combined.

Product-Support Services provided to directly • Engineering design

Services or indirectly support the
design, development, • Product distribution
delivery or sale of a certain
product. These are often
internal services.

Table 1

In the strictest sense SPC also requires that we be able to identify and monitor certain
characteristics of the process that are relatively stable. Here the expression "relatively stable"
implies that the characteristic in question will, within certain limits, vary in a random fashion
around some pre-determined baseline (or average) level. But, as we shall see, there are some
important reasons why such a restriction is more than we can expect from most service
processes. This does not necessarily invalidate the use of SPC in such processes—not, that is, if
we are willing to ease up on the restrictions in how we traditionally define SPC. In other words,
we need to agree in advance that we are interested in functional pragmatism rather than idealism.

www.R2assoc.com
© Copyright 1995 (revised 2005) Lon Roberts Page 4
 Applying SPC in a Service Environment

What's So Special About Services?

Services are distinguished from products in several important ways, one of which has already
been described: namely, that services are intangible while products are tangible. Another
factor—perishability—is related to intangibility but distinctly accounts for the fact that services
cannot be saved, accumulated, or inventoried.

Either of these factors—perishability or intangibility—could have bearing on the application of

SPC to service processes. For instance, the relative intangibility of the work-product may limit
us to monitoring only those factors that relate to process execution, since there is often no
partially completed work-product available for examination while the process is underway. By
contrast, in a manufacturing environment we typically have access to a broader array of variables
and attributes that can be monitored, including those that relate to the actual work-in-process.

From an SPC point of view, the perishability factor is significant due to the fact that the
customer "experiences" the service at the point of execution. In other words, there is often little,
if any, delay between the time the service is produced and the time when the customer benefits
from the service. The real-time nature of service processes suggests that, in contrast to product
manufacturing, there may be little, if any, opportunity to correct deficiencies before the outcome
of the process reaches the customer. Consequently, outcome-indicators, such as the number of
defectives, often have limited worth in a service environment. In the same respect, I would
challenge the notion that customer retention is a key measure of service quality, as some
contend.2 Customer retention, while perhaps the ultimate indicator of service quality, is a weak
index for controlling service quality. By the time you are able to observe and measure customer
retention you are past the point of improving it—much in the same way that product inspections
are an after-the-fact proposition that offer limited utility from the standpoint of process control.
This highlights the fact that if in-process indicators and controls are important in a
manufacturing environment, they are even more so in a service environment. Loosing a
customer is more consequential than rejecting a product.

There are a couple of other distinguishing characteristics of services that deserve special
attention from the standpoint of applying SPC. Let's examine these in more detail.

High Human-Factors Component

Manufacturing processes are dependent on an array of devices and sub-processes that harness the
laws of physics and chemistry. These may take the form of servos, actuators, sensors, gears,
switches, or a host of other electro-mechanical devices, tools, or complex machines. To a large
extent, process performance in a manufacturing environment is defined by the reliability and
accuracy of the devices, tools, and machines themselves. In fact, as the dependence on
automation increases, the so-called human factor diminishes in importance, at least in those
portions of the manufacturing process that are so affected.

With service processes, on the other hand, the outcome and resulting quality of the outcome are
highly dependent on the human contribution. Therefore it is important to recognize that service
processes have a significant labor content—where the term "labor" is the prevalent way of

www.R2assoc.com
© Copyright 1995 (revised 2005) Lon Roberts Page 5
 Applying SPC in a Service Environment

distinguishing employees from capital equipment, even though knowledge and intellect have
largely displaced manual labor in the contemporary service process. Be that as it may, we are
somewhat conditioned to view all processes by the generic form represented in Figure 1. This is
perhaps due to the fact that process flow diagrams have traditionally been used to represent
manufacturing processes, which in turn produce products rather than services. In other words,
Figure 1 depicts the process as a transformational entity, wherein inputs are translated into
outputs through a series of well-defined steps or stages. Using the traditional form of
representing a process, suggested by Figure 1, the process essentially consists of those things that
happen within the box, while the rest is viewed as being peripherally related to the process itself.

Figure 1

As far as service processes are concerned, I submit that Figure 2 provides a more accurate
representation of reality than Figure 1. The point here is that the process and the people
supporting the process cannot be isolated in most service processes. The distinction between the
process representations in Figure 1 and Figure 2 is more than symbolic or semantic, it ultimately
has bearing on the way we choose to monitor, control, and enhance the process.

SPC Implications: No two people are alike in terms of their psychological, emotional, social, or
intellectual makeup. Consequently, given that service processes are heavily dependent on human
performance, it is reasonable to expect that the baseline (or average) performance will shift when
and if different individuals are brought into, or removed from, the process. Furthermore, even
the same individual does not perform consistently from one instance to the next—thus adding to
process variability. When planning a measurement and tracking system for the service process,
it is important to be aware of any variance and/or any temporary shifts in process performance
that are introduced as a result of individual performance. If data are collected in such a way that
it aggregates or fails to account for the contribution of two or more individuals, it may be
impossible to pinpoint and remediate problems on the basis of the available "facts." Collective
indicators of team performance also tend to mask the impact of individual contributors—which
may, in some situations, be the desired intent.

www.R2assoc.com
© Copyright 1995 (revised 2005) Lon Roberts Page 6
 Applying SPC in a Service Environment

Figure 2

High Degree of Customer Involvement

As indicated in Figure 3, in a service environment, the customer is often intimately involved in

the process that results in the service in question. In other words, the customer and the service-
provider may work hand-in-hand. For instance, the person receiving a haircut from a certain
barber is directly involved from start to finish in the process that leads to the outcome. By
comparison, when it comes to products, the customer typically has little, if any, involvement in
the internal manufacturing process that actually results in the product itself.

To appreciate the significance of what this means, it may help to compare the similarities of a
process in which there is a high degree of customer involvement to that of a cafeteria serving
line. In the case of the latter, the customer moves from point to point, selecting from the
available salads, entrees, vegetables, breads, desserts, and beverages. Here each process-point
may be as important as the next. In other words, regardless of the outcome at the end of the
serving line, from the customer's point of view, the process is no better than its weakest link. If,
for instance, the person serving the salad is inconsiderate or otherwise inattentive, the customer
will most likely view the process as a whole as a failure. These, and analogous process-points in
which the customer is intimately involved, are essentially what one author refers to as moments
of truth.3 It suggests that the customer is "tied into" various points in the process—not simply the
end-point. But it also suggests that the customer may bear some responsibility (if not
accountability) for the actual quality of the service. For instance, a particular customer may be
having a "bad day" or perhaps has unreasonable expectations, depending on how the term
"unreasonable" is defined in a given situation.

SPC Implications: In the case of services, the customer often introduces an element of variability
to the process, depending on his or her involvement in the process. Furthermore, the degree of
variability induced will differ from one customer to the next. Some customers, for instance, are
more cooperative than others in making the process work effectively and efficiently—as we
might expect in the haircut example. Even from one day to the next the same customer will
experience psychological, emotional, and physiological shifts that can affect his or her
contribution to the process. Therefore, depending on the process and which aspects of the
process we wish to monitor and control, it may be necessary to account for—or at least be aware
of—the variance that is introduced by the customer. Any decision to take action based on SPC
indicators of process performance should consider the relative degree of customer involvement
factor—not for the purpose of affixing blame, rather for the sake of making rational decisions on
how and when to modify the process.

www.R2assoc.com
© Copyright 1995 (revised 2005) Lon Roberts Page 7
 Applying SPC in a Service Environment

Figure 3

Key Performance Indicators

The earlier discussion regarding the Law of Unintended Consequences described the potential
downside of choosing improper indicators to measure and monitor process performance—
especially the human concerns. When setting up an SPC system to support a service process, it
is also important to consider the following:

• The benefit of getting the process team involved in selecting the appropriate performance
indicators

• The selection of key performance indicators that are capable of providing "early warning"
of how the process is performing—the leading indicators, so to speak

• The danger of selecting process performance indicators on the basis of available data
rather than meaningful data

Elaborating on the last point, it is better to avoid SPC entirely than to establish a system that
measures and monitors superficial or meaningless indicators of process performance. For
instance, I question the value of monitoring "keystroke errors" in an environment where
everyone has access to computerized spelling and grammar checkers. At most I would limit this
practice to certain critical documents, such as sales orders, rather than every document produced.

A key performance indicator should be just that. To the extent possible, I suggest monitoring
parameters that provide an early indication of potential problems that may, if allowed to pass,
have a significant impact on the downstream customer. Naturally, in those processes for which
there is a high degree of customer involvement, this may be a challenging task. One approach
for identifying the key performance indicator(s) is as follows:

1. Start by identifying a problem condition that has a "reasonable" chance of occurring and
that would significantly impact the customer should it occur.

2. Using the Fishbone Diagramming technique4 or the Process Classification method5 have
the process team create a list of possible causal factors.

www.R2assoc.com
© Copyright 1995 (revised 2005) Lon Roberts Page 8
 Applying SPC in a Service Environment

3. Using Pareto Analysis6 or the Q-Sorting technique7 have the process team isolate the
critical few causes.

4. If a direct indicator is available for representing the behavior of the process at the point
where the causal condition occurs, this would likely serve as an acceptable performance
indicator. But, in some cases, it may be necessary to devise an indirect metric that
mirrors the performance of the process at the control point in question.

In addition to subjecting the short list of potential performance indicators to the test for
unintended consequences, it is desirable to know the cost-effectiveness associated with each
potential performance indicator. In other words, we would likely wish to have some idea of the
degree of difficulty of obtaining and recording the data in question.

In some cases, the most direct indicator of the parameter we wish to monitor may also be the
most difficult or most costly to obtain. For instance, at some stage in the process of filling a
sales order, we may be interested in tracking the time consumed before the order begins to
receive attention—depicted by the time differential, T2 minus T1, in Figure 4. If for some reason
it is infeasible to time-stamp the sales order when it begins to receive attention (i.e., the T2 data
were for some reason costly or difficult to obtain), an acceptable alternative may be to subtract
T1 from T3 and then subtract from this the pre-determined "average time" required to process an
order at this stage of the process. This might provide us with a "reasonable" and less costly
estimate of the time delay, T2 - T1, especially if the average processing time, (T3 – T2)avg, is
relatively constant.

Figure 4

A Word or Two About Control Limits

Earlier I discussed the fact that service processes can experience shifts in the process baseline as
well as added variance as a result of both the high labor content and the high degree of customer
involvement. Since these are technically induced, rather than random, sources of variation, it
may be desirable to expand the process control limits—when control limits are indeed relevant—

www.R2assoc.com
© Copyright 1995 (revised 2005) Lon Roberts Page 9
 Applying SPC in a Service Environment

and somewhat relax the "when-to-take-action" criteria that would otherwise apply to
manufacturing processes.

Unfortunately, there are no hard and fast rules on how much the control limits might be
expanded. Often it comes down to a matter of judgment based on intimate knowledge of the
process and the specific characteristics of the service environment. For instance, if both of the
factors mentioned have little bearing on the stability and reliability of the process, then the
control limits may be established using a standard process capability study. Also, in some cases,
as when Run Charts are involved, control limits are not an issue.

Figure 5 depicts a condition where the ±3σ control limits of a certain SPC chart have been
extended by the amount of ± ∆ to compensate for anticipated shifts in the process that are
induced from the sources identified above. If, in this case, the chart plots the sample averages of
a certain process parameter, it is possible to determine the magnitude of ∆ from the following
equation:

where k is the amount (in standard deviations) by which the process parameter in question is
expected to shift in either direction, and n is the sample size used in establishing the sample
averages. For example, if we anticipate an induced shift in the process of ±0.5σ and the sample
size is 4, then the process control limits in Figure 5 would be extended in both directions by
0.25σ (i.e., 0.5σ ÷ √4).

Figure 5

The Right Tool for the Job

As far as SPC in theory is concerned, there is no distinction between service processes and
manufacturing processes. The tools are as valid for one as they are for the other. But, from a
practical point of view, based on what we know about the nature of service processes, certain
charting tools are more appropriate than others.

www.R2assoc.com
© Copyright 1995 (revised 2005) Lon Roberts Page 10
 Applying SPC in a Service Environment

Here are some practical considerations relevant to the "typical" service process that may have
bearing on the setup and maintenance of your SPC system:

• The data that are available for characterizing the performance of service processes often
are not conformable to sampling schemes. For instance, the data from service processes
tend to occur aperiodically and somewhat infrequently rather than continuously and
frequently, as with manufacturing processes. This lends support to the use of such tools
as Run Charts, Scatter Diagrams, Moving Range Charts (for charting individual values),
Pre-Control, as well as plots of certain Critical Ratios8 that are able to account for relative
changes in two or more interrelated factors.

• Attribute data relating to "number or percent of defectives" should be used with caution
in a service environment. As discussed earlier, due to the real-time nature of service
processes, it may be unacceptable to tolerate any defectives—especially those occurring
at points in the process where the customer could immediately experience the resulting
impact.

• Service jobs, in contrast to manufacturing jobs, require a high degree of physical mobility
on the part of the employees. Likewise, the processes themselves are often not tied to a
specific physical location, as is almost always the case with manufacturing processes.
Consequently, the task of collecting and recording data in a service environment adds a
measure of difficulty. These, and perhaps other workplace realities, should be carefully
considered when establishing the charting and data collection requirements.

A Final Point to Ponder

SPC has much to offer any service organization that is serious about fact-based decision making.
On the other hand, there are some critical pitfalls in attempting to apply SPC in a service
environment. My advice? Carefully review the points described herein while assessing their
implications to your situation. Also, to the extent possible, I recommend involving the process
team when planning and setting up your SPC system. I can assure you that if you pay heed to
this advice that you will be well on your way to enjoying the benefits that SPC has to offer.

References
1. H. J. Harrington, Business Process Improvement: The Breakthrough Strategy for Total
Quality, Productivity, and Competitiveness, (New York: McGraw-Hill, 1991).

2. A Business Week Guide: The Quality Imperative, (New York: McGraw-Hill, 1994), pp.
140-141.

3. Jan Carlzon, Moments of Truth, (Cambridge, Mass.: Ballinger Publishing Co., 1987).

www.R2assoc.com
© Copyright 1995 (revised 2005) Lon Roberts Page 11
 Applying SPC in a Service Environment

4. Lon Roberts, Statistical Process Control for Intuitive Thinkers, (Plano, TX: Roberts &
Roberts Associates, 1992), pp. 34-35.

5. Ibid., p. 36

6. Ibid., p. 37

7. Problem Solving & Decision Analysis, A training program licensed through Roberts &
Roberts Associates, Plano, Texas, module 12, p. 10.

8. Project Management, A training program licensed through Roberts & Roberts Associates,
Plano, Texas, p. 97.

About the Author

Lon Roberts, Ph.D. is the president of Roberts & Roberts Associates, an international consulting
and training firm based in Plano, Texas. He is the author of numerous publications and training
programs, as well as four books, his latest titled SPC for Right-Brain Thinkers: Process Control
for Non-Statisticians (ASQ Quality Press, 2005). Dr. Roberts is noted for providing “rapid-
solution” consultation to teams, team leaders, process owners, and executives. He also conducts
workshops in the areas of SPC, process reengineering, project management, problem solving and
decision making, and quality excellence. He holds a Ph.D. from the University of Oklahoma and
B.S. and M.S. degrees from Oklahoma State University.

For information on workshops, keynote presentations,

publications, or consultation, contact Dr. Roberts at:

972-596-2956
[email protected]

www.R2assoc.com
© Copyright 1995 (revised 2005) Lon Roberts Page 12