Sociotechnical Reasons for the De-evolution of Statistical Process Control

January 2002
Volume 9 • Number 1


Sociotechnical Reasons for the De-evolution of Statistical Process Control

by Harrison W. Kelly III and Colin G. Drury, State University of New York at Buffalo

Increasing customer demands and the need to stay competitive create challenges in high-volume manufacturing companies competing in global markets. Increasing pressures to produce high-quality products with fewer resources can cause degradation in procedural compliance, particularly when compliance conflicts with the ability to meet production schedules. Reaction to statistical process control (SPC) out-of-control conditions is one such procedure. SPC procedures often require that processes be shut down until the cause of an out-of-control condition is identified and removed. Poor integration of SPC procedures into existing business systems makes compliance with this type of procedure stressful for many levels of the organization. Using focus group data, sample surveys, and focused interviewing, a model is derived that describes the SPC system changes that can occur. The interrelationships between these organizational functions, their respective knowledge levels, and what motivates them is critical to understanding why attempts to integrate SPC into business systems can fail. This article proposes a sociotechnical model describing the de-evolution of SPC in organizations where careful consideration is not applied to the integration of SPC into existing business systems. Advice to practitioners on the avoidance of SPC de-evolution is provided.

Key words: cluster variables analysis, focus groups, focused interviews, sample surveys, statistical process control


Consider a hypothetical organization partially consisting of local management (providing direction in matters of schedules and production rates), engineers (providing technical guidance and equipment redesign), quality technicians (providing data analysis support), and operators (providing actual labor to manufacturing product). The organization has determined that it needs to reduce the existing process variability (Ott and Schilling 1990; Duncan 1986). The quality technician suggests the use of statistical process control (SPC) and arranges training for the local management, engineers, and operators on the use and interpretation of SPC. Taking a standard approach, the quality technician stresses that out-of-control conditions require prompt action to correct the problem and return the process to a state of control. This “prompt action” includes shutting the process down, notifying local management and engineering support staff, and completing a report form to record the actions taken to return the process to a state of control. Once satisfactorily trained, data are collected, initial control chart limits are calculated, and a control chart is generated and introduced to the manufacturing floor for use (Juran and Gryna 1980). Once introduced, operators record data on the control chart, and in accordance with their training, use the control chart to identify when special causes of variation in the process occur (this will be referred to as the classical or typical SPC system for the remainder of this article).

When the control chart indicates that the process is out of control, the operator shuts down the process, notifies local management and engineering support staff, and completes the out-of-control action report form. Local management, concerned with meeting production schedules and assuring that the production line is making enough parts to fill orders, contacts the engineers to resolve the problem. In their eyes, the operator’s reaction to out-of-control conditions is counterproductive and disruptive, particularly if the parts produced are within the customer’s specification limits. Local management initially complies with the SPC procedures and involves the engineers to help resolve the process problems. The engineers, typically charged with supporting multiple manufacturing machines, have a hidden objective to quickly restart the equipment so they can effectively support all of their equipment rather than take the time to understand the cause of the problem. After a quick analysis, they ascertain whether the process is producing parts within the customer’s specification limits and either adjust the process or instruct the operator to continue to run in an out-of-control state. Observing these actions, the operators begin questioning the importance of SPC and the value in reacting to its signals.

Initially, the operators react in accordance with their training when the SPC chart indicates a process change, and they tolerate the organizational pressures that result. After several out-of-control signals occur, however, the operators realize that procedural compliance is an unpleasant experience and should be avoided. Since operators are typically the first people to observe the SPC out-of-control signals, they often receive pressure from local management and engineering to “keep the process in control.” With the support of local management and engineering, the operators avoid the unpleasantness of reacting by taking actions to prevent going out of control.

W. Edwards Deming (1982) used the term “tampering” to describe actions people take to reduce process variability where the action results in an inflation of the process variability. People generalize this concept to include any reaction to variability in a process, whereby process inputs are adjusted, but the underlying cause of the variation is not understood. Therefore, any action taken to correct an out-of-control condition without knowing its cause constitutes tampering. With the best of intentions, the operators ameliorate their SPC training and social pressures by tampering with the process. Their tampering falls into one of three categories: selective sampling, falsification of data, or unnecessary process adjustments.

Selective sampling, or the process of critically choosing parts to include in the sample recorded on the control chart based on a criterion other than consecutive random selection, is rationalized by overlooking the importance of subgroup sampling. Falsifying data, or the act of fabricating data to create a “good” sample, is rationalized with the perception that performance to specification is more important than performance to SPC limits. Process adjustments, or adjustments made to the manufacturing equipment in an attempt to “improve” the sample by forcing the process to produce “good” sample measurements, is rationalized by believing that the process is corrected and verified by the sampling. These three actions satisfy the operator’s immediate need to relieve the stress caused by out-of-control conditions, but they have a negative effect on the SPC and ultimately the process.

Quality technicians, unaware of the process tampering, examine the control charts and mistakenly conclude that process improvements have improved the process by reducing the intra- and inter-sample variability. The quality technicians mistakenly recalculate the control limits and release the charts to manufacturing. The new charts, while statistically correct, show control limits that reflect the operators’ ability to tamper and not the processes’ ability to produce consistent output. This breakdown in the SPC system reduces job stress while still giving the appearance of SPC procedural compliance and is the focus of the authors’ research.

The research presented in this article consists of focus groups, sample surveys, and focused interviews conducted in a United States manufacturing facility with trained employees who were actively using SPC for more than 10 years. The authors’ research suggests that when SPC is inappropriately integrated into an existing business system, it de-evolves into a control system that is significantly divergent from SPC as classically defined.


Focus group research is a qualitative research method consisting of small groups of people lead by a facilitator. Focus groups openly discuss topics that are important for a particular study or project (Stewart and Shamdasani 1990). Focus groups, by design, are informal, which encourages participants to share their views with the other group participants (Erlandson et al. 1993; Holstein and Gubrium 1995; Lee 1993). This within-group interaction can reveal complex social themes, illuminate variation in viewpoints, provide the opportunity to probe for clarification, and provide extensive and detailed information on a specific topic without having to conduct a full-scale anthropological investigation (Greenbaum 1987; Krueger 1994).

The authors’ focus group consisted of 15 volunteers who agreed to examine if (and why) “SPC was causing them trouble.” The individuals represented a broad organizational cross-section with three line operators, five technicians, three engineers, and four managers, and since all of the volunteers came from different manufacturing areas, none was subordinate to any other. A seven-question dialogue was used in the focus group sessions. The seven questions, created using a prior knowledge of typical SPC problems and complaints, provided a direction for the focus group discussion. As one can see in the first column of Table 1, the questions were constructed to capture the prevailing social, attitudinal, and comprehensive themes regarding SPC. The questions were intentionally designed to be open-ended, nonthreatening, and nonleading. Each question was asked in the order shown in Table 1 and audible responses were captured in the order they were uttered.

Transcripts of the responses were created from the focus groups and studied, being careful not to disrupt the order of the responses. Transcript analysis consisted of a careful review of emergent sentiments and central themes found in the responses. Since measuring strength of opinion from focus group data can be difficult, emphasis was placed on adjectives and adverbs used by the participants and were subsequently carried through to the analysis summary.

Responses to the first question, “What is SPC to you?” begin with traditional textbook definitions and concepts pertaining to SPC. The words and phrases used have a positive connotation that accurately describes the theoretical purpose of SPC. This suggests a level of knowledge and awareness of SPC among the participants but does not indicate attitudinal preferences. The eighth utterance stands in sharp contrast to the first seven. It starts a series of connotatively negative responses from the focus group that continues throughout the remainder of the focus group sessions. A properly functioning SPC system as envisioned by a statistician is one where SPC is used as a tool to help manufacturing identify when a process changes. In this ideal system, there is no conflict between the users and the tool. Words like “nuisance,” “cumbersome,” “unnecessary,” and “afraid” indicate that there is a breakdown in the ideal SPC system and conflict exists. This was confirmed when the participants were asked about the contrast in the responses to which they responded that they initially provided definitions of SPC and later provided descriptions of its use.

Examining the responses for the remaining six questions, one can see that people largely have a negative attitude about SPC, there is a lack of organizational support for the SPC system, and there is confusion about SPC in general. People largely understand that the purpose of SPC is to identify “problematic” process conditions and that the company makes money when product is shipped to the customer. Most agreed that when business pressures to produce product outweigh the perceived importance of pursuing the root cause of an out-of-control condition, people are put in an uncomfortable position.

Knowledge of the problem, and the risk taken by ignoring the signal, initially causes distrust in the organizational leaderships’ support of the SPC system and later in the impression of SPC as an effective problem-solving tool. Karasek (1981) and Karasek et al. (1990) refer to this state as one of high information and low control. People in this situation take action to alleviate this stress. The manner in which this is accomplished will be discussed later in this article. Based on the participant responses and the perceived emergent themes (see the third column of Table 1), the authors concluded that the focus group responses suggest that the lack of social and technical support is both causing the SPC confusion and resulting in the negative SPC attitude. Recognizing that the focus group constitutes a limited population representation, a larger sample survey was used to determine if the focus group participants shared approximately the same understanding, attitudes, and beliefs as the study population.


The sample survey was designed to further investigate the prevalence of the beliefs and attitudes discovered in the focus group analysis. The survey consisted of 22 unbiased, nonthreatening, nonleading questions aimed at studying the perceived emergent social, attitudinal, and comprehensive themes related to SPC. The authors’ underlying hypothesis, largely derived from the focus group analysis, was that a poor organizational SPC social and technical support structure leads to cognitive dysfunction and counter-productive control actions.

Of the 853 surveys distributed, 158 were returned (18.5 percent of the overall population), with no less than 15.8 percent representation from any one organizational position (operator, technician, engineer, and local management). Eighty-two percent of those who responded are employed by the company for four or more years, and 78.5 percent of the people have some knowledge of SPC for four or more years. Fifty-four percent of the respondents do not regularly record data on an SPC chart, and 59.5 percent do not regularly analyze SPC (see Table 2).

Hierarchical cluster analysis of variables using an average linkage method and absolute correlation for a distance measure was used to analyze the survey responses. Cluster analysis sequentially amalgamates variables (in this case, responses to questions) based on the similarity in population responses. In other words, questions that are answered similarly by the population are grouped together. This pairwise amalgamation continues forming clusters of variables until only one cluster remains containing all variables.

The authors’ analysis revealed a statistically significant similarity between specific organizational positions and the frequency in which they record data on an SPC chart. Examination of the Pearson correlation coefficient revealed a statistically significant positive correlation (0.581, P<0.001), indicating that positions with greater authority were less likely to record data on an SPC chart. This finding provided support for separating survey responses into two groups: people who regularly record data on SPC and those who do not. Once split, hierarchical cluster analysis of variables using an average linkage method and absolute correlation for a distance measure was conducted on each group of survey respondents. Dendograms for these analyses are shown in Figure 1 and Figure 2, respectively.

The analysis for regular recorders resulted in seven clusters. Likewise, the analysis for the nonregular recorders resulted in seven clusters. While the analyses were performed independently, both analyses resulted in four similar clusters that were of interest. The first of the similar clusters indicate that both groups of individuals believe control limits specify what a process should produce, and both groups believe they are paid to keep a process in control. Confusing the descriptive nature of a statistical limit with the prescriptive nature of a specification limit could be the result of a lack of technical support. Lack of social support could be the cause of the pressure to “keep a process in control.” In either case, these findings appear to support the focus group findings.

The second set of similar clusters indicated that both regular and nonregular recorders learned how to use SPC while in the work environment from either consultants or short courses, and they both think that more SPC training is needed. The authors’ collective experience has been that short courses usually fail to stress in-depth conceptual or mathematical SPC concepts favoring a more cursory, conceptual one. Additionally, it has been their experience that these courses spend little (if any) time on the integration of SPC into existing business systems.

These sample survey findings confirmed the focus group findings by suggesting that inadequate social and technical support structures existed and that there was a general lack of SPC understanding. Even with the additional supportive information, questions remained regarding individual SPC comprehension compared to that of a statistician. To capture this information, focused interviewing of study participants was conducted.


Focused interviewing is a qualitative research method for gathering very detailed information from a small group of individuals in a one-on-one format. (Fowler and Mangione 1990; Erlandson et al. 1993; Holstein and Gubrium 1995; and Lee 1993). Randomly selected individuals are interviewed to openly discuss topics that are important for a particular study. While the interview discussion is planned, the interviewer probes for better understanding and clarification throughout the interview. Qualitative focused interviews capture the subtle nuances in individual interpretation of terms and definitions. The current application of focused interviews is intended to clarify individual comprehension of the control system and to determine specific differences in control methodology from that of a statistician. The questions used were open-ended and designed to qualitatively assess whether the survey population shares the same understanding of the objectives and mechanics of SPC as a statistician. The focused interview participants consisted of three operators, three SPC technicians, three engineers, and three local managers.

Analysis of the interview results indicated that operators, technicians, and engineers colloquially refer to the process control methodology in use as SPC. The actual control method, however, is divergent from classical SPC. “SPC” was used as a general term, indicating a process of collecting and graphing data, yet it did not necessarily resemble classical SPC. The SPC technicians indicated that data collection sheets with “tentative control limits” are used to collect data for the creation of a new control chart. When asked how these limits were derived, answers ranged from “based on other similar processes” to “the process engineer told me what to use.” They indicated that the data were then analyzed, control limits were calculated, and a new SPC chart was generated using the new calculated limits. Once the control chart had been implemented on the line, it was monitored for out-of-control conditions and other problems.

One technician indicated that if the control chart goes out of control too soon after implementation, there is immediate suspicion that the problem lies in the control chart and not in the process. All SPC technicians indicated that if the control chart goes out of control too often, the operators and engineers would either remove the control chart (against procedure) or would insist that the control limits be widened. One technician volunteered that operators often complain that “control limits are too tight” when a new control chart is introduced. One engineer admitted that control limits had been initially statistically calculated, however, these limits tended to generate “too many out-of-control signals” and could not be supported by the technical support structure. Another engineer described the process of creating new control limits as follows: “Data are collected, control limits are generated from the data, and then these limits are reviewed for appropriateness in the process.” He went on to state “If the limits are too tight for the process to hold, or if the customer doesn’t need very tight control, then the limits are widened.” Another indicated that “the best SPC limits are created through trial and error, if one set of limits say the process is out of control and it clearly isn’t, then another set can be tried to see if they work better.” In areas where the control limits had been statistically calculated, most agreed that processes with a longer history using SPC received more complaints from SPC users regarding the control limits being “too tight.”

When asked if a control chart should go out of control the operators agreed that it should not. One individual volunteered that if the chart went out of control it would indicate that her process was producing “bad product.” When asked why this was a problem, she said she thought she could lose her job (moved to a less desirable, more tedious job) if she did not produce product within the control limits. She indicated that “if other people on other shifts could produce product with the same equipment within the control limits, then she to should be able to produce within those limits.” Another stated that out-of-control conditions “mean that my process is producing bad parts, so the engineer and team leaders get all interested in the process—the rest of the time, no one cares.” Responding to the same question, an engineer stated “only when there is a problem that will affect the customer. The rest of the time, out-of-control conditions don’t tell me anything and I don’t have the time to worry about nonissues.” This sentiment was echoed by a local management: “I don’t care if the SPC chart says the product is bad, if it is in specification limits, it’s good enough for me.”
Additionally, engineering interviewees conceded that the control chart signal was of no consequence “as long as all parts were within specification.” This lack of distinction between control limits, specification limits, and the perceived excessive out-of-control signals suggests that engineering staff members lack appropriate understanding of the SPC limits. Local management also indicated a general distrust in SPC and a lack of appreciation for the purpose of SPC. This may be in part due to engineering input confirming local management’s bias on the business value of SPC. One operator, when asked how often a control chart goes out of control, indicated, “never—my current control chart has not gone out of control since I started working on that process (one year ago).”


This research, in conjunction with knowledge of actual implementation and use, has led to the creation of the proposed model shown in Figure 3 describing common problems and socioenvironmental issues that can lead to a de-evolution of the classical SPC system.

The model is entered with the inclusion of new data into the existing data set. Given the users understanding of the control method and process, the new and historical data are interpreted and a decision is made to either take an action or continue running the process. The actions individuals can take (or their potential actions) are determined largely by their understanding of the control objectives, control mechanics, process objectives, and process mechanics. These knowledge groupings became apparent as a result of the focus group, sample survey, and focused interview responses.

The first of these, control objectives, describes an individual’s awareness and understanding of the information provided by the control chart. It describes an individual’s comprehension of a classical SPC system. As noted previously, the authors’ research revealed objective comprehension differences between users and statisticians. They also uncovered differences in how SPC is used. The authors refer to this as control mechanics. Control mechanics describe an individual’s understanding of how to properly collect and record data on a control chart or the physical use and reaction to SPC in a classical SPC system. Process objectives, like the control objectives, describe an individual’s understanding of the purpose of the manufacturing process. In other words, does the individual understand the purpose and importance of the process? Process mechanics, like control mechanics, describe an individual’s understanding of how the process operates and how to make adjustments to the process. An individual’s knowledge of the control objectives, control mechanics, process objectives, and process mechanics describes the totality of potential actions for that individual. As such, potential actions will naturally vary between individuals.

The authors’ research also indicated that socioenvironmental factors affect individual control actions. Notice in Figure 3 that the control system is externally disturbed by potential action modifiers. Potential action modifiers are circumstantial and environmental conditions that can modify the user’s potential actions during the collection of data and/or reaction to SPC signals. In other words, the potential actions a user takes may be modified depending on the circumstances, the amount of pain expected by reacting, the existence of competing objectives, and the level of trust an individual has in the SPC signal received. Focus group, sample survey, and focused interview results designate these as the primary causal factors affecting potential actions.

Circumstances alter an individual’s potential action when it is perceived that the individual is isolated from technical (or nontechnical) assistance. This can ultimately limit the individual’s ability to react to a signal derived from the new and historical data. Individuals in this situation are forced to compromise their potential actions, which may affect the quality, frequency, and timeliness of the reaction. An example that demonstrates circumstantial modification of potential actions is an individual, working on an off-shift, in need of technical assistance. The individual will likely react to SPC signals differently knowing that there is no technical assistance to help resolve the problem.

Reaction pain is the perceived or expected stress associated with (or resulting from) the reaction to SPC signals. Given a history of painful events, an individual may have a tendency to compromise a potential action. An example demonstrating reaction pain might be the overloading of an individual responsible for many activities. In other words, an individual may be responsible for resolving SPC out-of-control signals in addition to performing his or her regular tasks.

Objectives that compete for prominence alter the potential actions of an individual when the need to meet another objectives is perceived as having greater importance than conforming to the SPC system. For example, a competing objective might be failing to shut down a process in response to an SPC signal so the process can continue producing parts for the customer. Compromising SPC actions affects the individual’s trust in the historical data, which leads to a lack of trust in the SPC system signals. A lack of trust in SPC system signals can affect an individual’s potential process actions.

These potential action modifiers have an effect on potential actions, the way in which new data are collected, and the way SPC signals are used. Individuals make final control decisions using their knowledge of the control method and the process (potential action knowledge) and by (consciously or subconsciously) considering the potential action modifiers. This has an effect on the data set when next interpreted as the final control decision and only the sample data remains. Future potential action modifiers can thus be altered, and this can potentially exacerbate the influence that the potential action modifiers have. These potential action modifiers thus form the social constituent to the technical capabilities (potential actions). Potential action modifiers therefore have a significant effect on the overall control system.

As described, SPC de-evolution occurs over time with the addition of every new piece of data and the decision that is made on how to react to the SPC system signals. Iterative cycling through the model thus describes how SPC can de-evolve. This proposed model and the conditions observed is partially substantiated by other authors studying human behavior using quality, statistical, and graphical methods.


In 1965, Stok published groundbreaking work linking the behavioral and industrial/statistical scientific fields. Stok investigated the effects of the visual presentation of quality data on the personal attitude of production operators and on quality improvement. He concluded that there are both incentive and informative effects associated with the display of data and that each of these “affects quality in a favorable way.” He believes the incentive effect derives from an individual’s desire for self-evaluation compared to established standards or norms, whereas the informative effect derives from the individual’s desire to use the information to direct process changes, thereby improving the quality of work. The authors suggest that the informative effect may serve to encourage needless process adjustment so as to make the recorded data look better on the control chart and the incentive effect serves as a catalyst, increasing the amount of adjustment and lessening the informative effect. Despite the action taken, the recorded intra- and inter-SPC sample variability is reduced, and there is a subsequent reduction in out-of-control signals. The most visible impact of this action is a control chart that falsely indicates process improvement.

Drury (1998; 1975) examined the breakdown of SPC from a human factors perspective. He introduced two models; the first describes how process control, generally speaking, works, and the second describes what SPC is supposed to do in a control system. In the first model, Figure 4, Drury explains process control as a system composed of command inputs (or new information) and prior process outputs (that is, a feedback mechanism). The totality of these two sources of information are considered relative to the understood process model and control system and become actions on the process. Outputs from the process become process feedback and the cycle begins again. What Drury is expressing is that many companies view process control simply as rule-based decision-making process requiring no knowledge of probability or statistics, which leads to SPC usage problems.

In the second model, Drury describes the function that SPC is supposed to serve within the control system (see Figure 5). Here Drury shows the cognitive process that an individual follows in determining which of many control decisions is decided upon given rules and knowledge of the control system and SPC. In this model, an individual uses the SPC chart and knowledge of the rules of SPC use in conjunction with a new sample to determine a control decision. One of many control decisions is made given the known SPC rules, operator knowledge, and job aids. Drury further contrasts these two models to examine failures that can occur at the various points in the control system and the consequences of these failures. These models describe a system that is highly dependent on the operator’s knowledge, and available job aids for the operator’s understanding and interpretation and influence the rules that are used at various points in the control system. System inputs can be subject to change, and as Drury points out, the changes can result in systemic failures at each step of the process.

Considering the findings discussed in this article and the effect that the human operator has on the overall control system (as described in Figure 3), it is reasonable to extrapolate from these earlier models a similar pattern of degradation in the control system. The authors therefore suggest that these earlier models support their recent findings and their proposed model of the SPC de-evolution.


Box and Kramer (1992), and Vander Wiel et al. (1992) suggest that SPC is a misnomer and that a more accurate term might be “statistical process monitor,” since the SPC chart itself does not physically control the process—people do through their actions. Wheeler (2000) proposes the terminology “process behavior chart” for largely the same reasons. In addition to supporting this body of research, their work additionally suggests that people may also control the SPC system and unintentionally cause it to devolve into a system that only superficially resembles an SPC system. Subsequent informal interviews with several manufacturing plants, both in the United States and abroad, have indicated similar problems with SPC systems (both automated and manual). The authors believe enough evidence exists to indicate a need for intervention. Interventions can range from awareness training through design of better technical solutions, to full sociotechnical systems redesign. Further observational research using a broad cross-section of industries is planned to evaluate the nature of SPC de-evolution.


It is currently unknown if SPC de-evolution can be prevented or even slowed. Practitioners interested in improving the effectiveness of SPC and hindering de-evolution could infer from the authors’ model a need to reengineer the functional responsibilities associated with SPC. Specifically, the analysis of and reaction to an SPC chart should be separated from the task of data collection. The nature of SPC is one of exploration and inquiry rather than control and routine; therefore, data collections should remain in the hands of the operator, given only a specification target to suggest where the process should run. The analysis, interpretation, and associated problem solving should be conducted by another functional position. Intuitively, an engineer’s function is one of problem identification, root-cause analysis, and solution implementation, a natural position to perform SPC analysis and reaction. The authors therefore suggest separating data collection from the analysis of and reaction to SPC and reassigning these tasks to different organizational functions. Additionally, an alternate process control more in line with the needs of the process operators should be substituted for SPC on the manufacturing floor.

This reallocation of SPC in conjunction with a different type of control mechanism for the operators to use has several sociotechnical design benefits. Operators would maintain holistic work and would not suffer from a loss of information or a lack of control, as nothing would be taken away from them–only replaced with charts that are more meaningful and useful. Several charts would be suitable replacements, namely pre-control, modified control limit charts, or some other form of integrated process control to determine when the process is at risk of producing unshippable material.

Precontrol charts and modified control limit charts calculate control limits differently than standard Shewhart-type control charts. Shewhart-type control charts start with an historical estimate of location and spread and assuming a Gaussian distribution, estimate control limits describing 99.7 percent of the expected outcomes given an unchanging distribution. Both precontrol charts and modified control limit charts use an estimate of standard deviation and an assumption of a Gaussian distribution, to calculate “control limits” by working backward from the customer’s specification. This type of control limit permits an operator to vary in ways standard Shewhart-type control charts do not. As a result, there is a reduction in the number of signals (that is, out-of-control conditions) requiring a response.

As such, both precontrol and modified control limit charts are more in line with the way operators, engineers, and local managers perceive the operators controlling task. In other words, rather than describing and reacting to the historical state, (that is, Shewhart charts) both precontrol charts and modified control limit charts provide warnings when the process drifts too close to a customer specification limit. Signals obtained from these controls indicate a need to adjust the process. It is important to note that quality professionals claim that neither of these control charts promote continuous improvement–a tenet of SPC. If, however, precontrol or modified control limit charts were given to the operator and Shewhart SPC charts were given to the engineering staff, (that is, parallel application of control methods), these control charts could not befall this criticism. Use of these charts in conjunction with a well-designed process would result in a reduction of signals indicating the existence of a problem in need of being solved by the operators. This reduction in signals would enable the operator to focus on using the information contained on the new control chart in an adaptive and controlling manner, forming opinions regarding what causes what (controlling the process) and minimizing the distractions of determining what might have caused what (diagnostic problem solving). Additionally, it provides engineers with the ability to isolate the application of SPC.

SPC implementation will also be more effective if the scope, term-of-use, and graphical display of SPC were modified. In other words, the number of potential causes associated with an out-of-control condition would be reduced (that is, scope) if SPC were applied to process input parameters instead of output parameters. Clearly, in an extreme situation where SPC were applied to a process variable wherein only one causal factor could be associated with an out-of-control condition, any directional changes of any magnitude would signal the necessary process adjustment to fix the situation. Reducing the length of time SPC is used on a process (that is, term) would serve to limit the extent of SPC de-evolution and, in-turn, any stress involved with reacting to fictitious out-of-control conditions. Accomplishing this implies that there are no existing quotas, mandates, or guidelines that must be followed in the application of SPC, and a clearly stated purpose for the SPC chart would need to be established to use as a criteria to terminate SPC use. Again, reducing the term of SPC usage would mitigate de-evolution (and therefore reaction stress and coping strategies) by preventing long-term use.

Without further research it is difficult (if not impossible) to estimate the effect of the suggestions made here on SPC de-evolution. The authors therefore plan to conduct comparative research to evaluate these alternative process control methods and other control methods to determine what effect they have on process comprehension and stress.


The authors are grateful to the reviewers for taking time to share their views on improvements to this article and for providing the valuable remarks on its content and structure.


Box, G., and T. Kramer. 1992. Statistical process monitoring and feedback adjustment: A discussion. Technometrics 34: 251-285.

Deming, W. E. 1982. Out of the crisis. Massachusetts: Massachusetts Institute of Technology, Center for Advanced Engineering Studies.

Drury, C. G. 1998. Human Factors and Quality: Update from the Trenches. Keynote address at the ERGON-AXIA meeting, Japan.

Drury, C. G., and J. G. Fox. 1975. Human reliability in quality control. New York: Halsted Press.

Duncan, A. 1986. Quality control and industrial statistics. Chicago: Irwin.

Erlandson, D. A., E. L. Harris, B. L. Skipper, and S. D. Allen. 1993. Doing naturalistic inquiry. Newbury Park, Calif.: Sage.

Fowler, F. J., and T. W. Mangione. 1990. Standardized survey interviewing: Minimizing interviewer-related error. Newbury Park, Calif.: Sage.

Greenbaum, T. J. 1987. The practical handbook and guide to focus group research. Lexington, Mass.: Lexington Books.

Holstein, J. A., and J. F. Gubrium. 1995. The active interview. Thousand Oaks, Calif.: Sage.
Juran, J. M., and F. M. Gryna. 1980. Quality planning and analysis from product development through use. New York: McGraw-Hill.

Karasek, R. A. 1990. Lower health risk with increased job control among white-collar workers. Journal of Organizational Behavior 11:171-85.

Karasek, R. A., D. Baker, F. Marxer, A. Ahlbom, and T. Theorell. 1981. Job decision latitude, job demands, and cardiovascular disease: A prospective study of Swedish men. American Journal of Public Health 71:694-705.

Krueger, R. A. 1994. Focus groups: A practical guide for applied research. Newbury Park, Calif.: Sage.

Lee, R. M. 1993. Doing research on sensitive topics. Newbury Park, Calif.: Sage.

Ott, E. R., and E. G. Schilling. 1990. Process quality control troubleshooting and interpretation of data. New York: McGraw-Hill.

Stewart, D. W., and P. N. Shamdasani. 1990. Focus groups: Theory and practice. Newbury Park, Calif. Sage.

Stok, T. L. 1965. The worker and quality control. Ann Arbor, Mich.: The University of Michigan.

Vander Wiel, S. A., W. T. Tucker, F. W. Faltin, and N. Doganaksoy. 1992. Algorithmic statistical process control: Concepts and an application. Technometrics 34: 286-297.

Wheeler, D. J. 2000. A modest proposal. Knoxville, Tenn.: SPC Press.


Harrison W. Kelly III is a doctoral candidate in industrial engineering human factors. He has spent 12 years applying statistical methods in manufacturing and has consulted with many organizations on quality and statistics. Kelly’s research interests include the study of quality systems, quality management, the application of statistical methods, process improvement, decision-making, and strategy. He is a member of the American Society for Quality and the American Statistical Association.

Kelly holds a bachelor’s degree and a master’s degree in applied and mathematical statistics. He can be contacted at the Department of Industrial Engineering, School of Engineering and Applied Sciences, 307 Lawrence D. Bell Hall, Box 602050, State University of New York at Buffalo, Buffalo, N. Y. 14260-2050, or via e-mail at .

Colin G. Drury is professor of industrial engineering at University at Buffalo, where his work is concentrated on the application of human factors techniques to manufacturing and maintenance processes. Formerly manager of ergonomics at Pilkington Glass, he has more than 200 publications on topics in industrial process control, quality control, aviation maintenance and safety. He was the founding executive director of The Center for Industrial Effectiveness, which works with regional industries to improve competitiveness and has been credited with creating and saving thousands of jobs in the region. Drury is a Fellow of the Institute of Industrial Engineers, the Ergonomics Society, and the Human Factors Ergonomics Society, and received the Bartlett medal of the Ergonomics Society (for contributions to human factors and quality) and the Fitts Award of the Human Factors Ergonomics Society.

Drury can be contacted at the Department of Industrial Engineering, School of Engineering and Applied Sciences, 307 Lawrence D. Bell Hall, Box 602050, State University of New York at Buffalo, Buffalo, New York 14260-2050.

If you liked this article, subscribe now.

Return to top

Featured advertisers


(0) Member Reviews

Featured advertisers

ASQ is a global community of people passionate about quality, who use the tools, their ideas and expertise to make our world work better. ASQ: The Global Voice of Quality.