This article is a sequel to “Total Cost of Quality,” which appeared in the August 2002 issue of Imaging Spectrum magazine. That article explored the concept of quality in a company’s products and services and the costs involved, whether quality is emphasized or ignored completely. The article demonstrated that the cost of ignoring quality issues is actually higher than the costs associated with the prudent use of methods to ensure quality. We will refer to these methods as “quality tools.” We also discussed in detail the concepts of prevention and appraisal costs in quality control.

This month’s article will focus on specific quality tools that are commonly used in the search for quality. The tools one would use to analyze product quality (design) and process quality (manufacturing) are the same in many cases, but the applications differ slightly. Often, the root cause of a product returned under warranty turns out to be a production process that is too variable. Therefore, this article will focus on root cause analysis techniques and manufacturing process improvement.

Defining and Measuring Quality

Quality has been defined, at least in part, as inversely proportional to variation. Everything in the world has some variation associated with it. No two of the same model cars from the same factory will perform exactly the same. No two toner cartridges from the same production line will have exactly the same yield. The purpose of applying quality improvement tools is to reduce variation of attributes that affect performance. Therefore, it is useful to consider the application of quality tools to processes or products in order to reduce variation.

A fundamental consideration in improving quality is determining what to measure. It is much easier for a customer to define quality than it is for a manufacturing or design engineer to do so. Customers simply state, “I want this product to work all the time.” Design and manufacturing engineers must translate customer requirements into measurable quantities. Kodak calls these product or process attributes key characteristics (KCs). Xerox refers to them as critical parameters (CPs). Whatever they are called, these attributes of a process or product are intimately tied to quality.

To select meaningful attributes for a manufacturing process, one must consider the function of the process. What is it that this step is trying to accomplish? What is a good output? What is a bad output? These questions are good starting points for determining what about the process should be monitored.

Appraisal Tools

The first step in solving any issue is to understand the actual problem. This may sound like common sense, yet many people are prone to “solution-jumping” before the problem has been fully understood. On the other hand, some people are disposed toward “making a career” out of every issue that comes up. Using established quality tools helps strike a balance between quick fixes and paralysis by analysis.

Customer Warranty Questionnaires

Suppose a toner cartridge remanufacturer is interested in understanding, analyzing and reducing the defective units returned under warranty. When evaluating a cartridge returned under warranty, the first source of information may be a questionnaire that should be returned as part of the warranty claim. Typically, the questionnaire asks for information based on the customer’s observations. This information can help direct the in-house evaluation effort. For example, if a particular cartridge were returned with a questionnaire that specified “wrong cartridge type,” it would not make sense to diagnose the internal components of the cartridge.

Customer feedback is sometimes difficult to interpret because customers lack in-depth knowledge about cartridges. What they typically look for is limited to printer function and image quality. It is useful, therefore, to develop a questionnaire using questions that focus on these aspects. Avoid technical terms when phrasing the questions. It is also a good idea to request that the customer provide a print sample made with the defective cartridge.

Internal Data Analysis Forms

The next step is to collect information about the actual returned cartridge. Every company has its own procedure for this sort of activity. No matter how it is done, in every case, data will be generated. Choosing which information is needed must be established on a case-by-case basis. If the evaluation was being conducted by an OEM, the inspection processes would likely be driven by intimate design and manufacturing knowledge. The inspection would evaluate any deviations from design specifications.

All information collected must be captured. Taking the time to develop a checklist for warranty returns will ensure that all data is consistent, which will facilitate future analysis. For example, if circumstances warrant a print test on the returned cartridge, your form might include spaces to enter observations, such as “blank print” or “black lines” or “excessive background.” These results would then lead to further procedural steps to investigate the particular observations.

Developing a form to analyze warranty returns can be either a very simple process or a more rigorous one, depending upon the potential return on investment. If your company places a premium on knowledge, a more detailed procedure would allow specific data to be captured. On the other hand, a lot of time evaluating warranty returns in detail can be spent without closing the loop with corrective action (paralysis by analysis). The key is to collect enough information to understand the nature of the problem with the objective of correcting it.

Take, for example, a fictitious remanufacturing company that has created a warranty return analysis procedure using a print test and a data collection form. The data collection form is used in conjunction with the print test to evaluate image quality. The form has a checklist for the observed print defects:

• No image
• Excessive background
• Streaking
• Image deletions
• Other

This form was developed using an ongoing process so that it can be continually updated by those employees conducting the analysis. Again, if an OEM were developing this procedure, it would be created by the design and manufacturing engineers. In the case of a remanufacturer, it must be generated by the production staff, using their experience and common sense.

Graphical Data Presentation

Once all this information is collected, an analysis must be done. A qualified person must review it, draw conclusions and then make decisions for action. The most popular way to look at information is using graphical techniques. Determining the best graphical technique depends upon the type of information that is being evaluated.

For data collected with the data form described above, a Pareto chart would likely be the best method. A Pareto chart (at right) is simply a bar chart that plots the frequency of each category or “bin.” The Pareto chart is different from other bar graphs in that the categories are sorted from the greatest to least frequency. This is an excellent way to evaluate quickly the “big hitters” or “vital few.” There are many graphical techniques including radar plots, histograms, scatter plots, and control charts that can also be used to organize and communicate data.

A Sample Analysis

Using our previous example, the data collected over the past four weeks is shown in the Pareto chart. Note that “Background” is the most frequently observed defect after “No Problem Found.” If rate of returns due to background problems constitute reason for future analysis, a quality improvement initiative could be started to reduce or eliminate warranty returns caused by this defect.

“No Problem Found” is a special case. It is perhaps the number one problem in warranty returns in terms of both frequency and frustration. The types of issues included in the “No Problem Found” bin may include problems that leave no trace and may never be resolved. Unfortunately, a surprising number of random factors can come together to create a problem that may never occur again. In these cases, it may not be worth the effort to address each of these limited, random problems.

However, when a particular problem occurs in sufficient numbers in the field but cannot be recreated in the lab, the resolution of the issue will require greater resources. The only way to define the “No Problem Found” cause is to go to the next step and collect information directly from the user. For example, imagine a case where a particular cartridge model is suddenly being returned in relatively significant numbers. In the majority of the cases, the cartridge works perfectly when installed in the test printer. Through further analysis, it becomes clear that just two customers are responsible for the majority of the returns (the further analysis may be angry phone calls in this case!). By discussing the issue with these customers and looking into the configuration of their machines, it is discovered that a printer software glitch has been causing them to believe that the cartridge was defective.

Depending on the type of customer, it may be worth the effort to clear up the problem even though it is beyond the cartridge remanufacturer’s direct responsibility.

Some other questions to explore with customers when faced with no obvious cause for a cartridge’s defect might be: Was the cartridge empty and therefore not performing? Was the seal not pulled? Was the cartridge removed from the printer and shaken, then not replaced properly?

Root Cause Analysis Tools

Root Cause Analysis Tools refer to methods that individuals or teams can use to organize their thoughts and focus their actions. Successful application of these tools will define a root cause of the problem under analysis and hopefully lead to corrective action.

One important tool in this area is the Fishbone diagram. A quality engineer named Kaoru Ishikawa developed the Fishbone diagram. It is sometimes referred to as an Ishikawa or a Cause-and-Effect diagram. The purpose of the Fishbone diagram is to schematically organize all the possible factors that may create a particular effect. A sample diagram (at left, on previous page) is shown for our example of a background development defect.
The information used to illustrate the potential causes comes from knowledge of the process. The Fishbone diagram is particularly well suited to capturing information generated by groups of people who are brainstorming potential causes. Everyone’s ideas are captured, and the group or person responsible for evaluating the problem can focus on the most likely sources.

Another important tool for root cause analysis is Failure Mode and Effects Analysis (FMEA). Product designers and manufacturing engineers use FMEA to formally document the primary failures that can be anticipated with a design or process. The failures are tied to the effects that the failure will cause in the product or process. If a problem-solving team has access to this information, it can be invaluable in understanding root causes based on observed failures.

Test procedures

At this point, the problem-solving team likely will form one or more hypotheses to investigate. It is important to explicitly specify which hypothesis is under consideration in order to design tests that address that issue and no other. This sort of activity takes vigilance and practice because seemingly definitive results can sometimes be misleading. For example, if removing a component makes the problem go away, does putting it back make the problem reappear? If the malfunctioning part is replaced, is something else contaminating that component so that the same problem will reappear over time?

Prevention Tools

Suppose that the background problem outlined earlier led to the discovery that the PCR was not charging the OPC drum sufficiently. More specifically, the problem was tracked to the electrical contact between the PCR and the cartridge. In this case, it was determined that the lubricant on the PCR contact assembly was actually insulative and was creating charging failures. Corrective actions can now be taken. The first is to start using the proper lubricant on the contact assembly. Hopefully, that will completely address the issue, but without monitoring the solution the problem may resurface.

“Failsafing” is the process of applying countermeasures at each point of the process where problems may occur. In our example, failsafing may include a specific grease dispenser that is clearly marked. In addition, operator training, which is another prevention tool, may be included as part of the remedial actions. A more quantitative failsafing technique, however, would be to implement an electrical resistance check through the contact mechanism. Several issues would need to be examined, such as whether the measurement would be performed while the unit is rotating and also whether the resistance must be monitored while the operating voltage is applied.

After the test process is designed and verifiably tied to the failure mode, the data that is collected can be used to monitor quality. The resistance of the contacts can be charted using a process called Statistical Process Control (SPC). SPC provides the techniques to calculate what the process normally delivers, as well as techniques to identify when the process output changes from that normal state. An example of the electrical resistance is shown at the left. Note the marked change in resistance when the old grease finds its way back into the process.

Monitoring quality with SPC is the best safeguard against changes in processes causing product failures. By deciding what to measure, and monitoring and taking action when things go out of control, problems can be stopped before they reach the customer.
Solving problems quickly and effectively when they occur is very important to reducing the total cost of quality. Yet, stopping problems before they occur is still the most effective way to reduce overall quality costs. The prevention tools discussed in this article are only some of the ways problems can be avoided.

This article is a sequel to “Total Cost of Quality,” which appeared in the August 2002 issue of Imaging Spectrum magazine. That article explored the concept of quality in a company’s products and services and the costs involved, whether quality is emphasized or ignored completely. The article demonstrated that the cost of ignoring quality issues is actually higher than the costs associated with the prudent use of methods to ensure quality. We will refer to these methods as “quality tools.” We also discussed in detail the concepts of prevention and appraisal costs in quality control.

This month’s article will focus on specific quality tools that are commonly used in the search for quality. The tools one would use to analyze product quality (design) and process quality (manufacturing) are the same in many cases, but the applications differ slightly. Often, the root cause of a product returned under warranty turns out to be a production process that is too variable. Therefore, this article will focus on root cause analysis techniques and manufacturing process improvement.

Defining and Measuring Quality

Quality has been defined, at least in part, as inversely proportional to variation. Everything in the world has some variation associated with it. No two of the same model cars from the same factory will perform exactly the same. No two toner cartridges from the same production line will have exactly the same yield. The purpose of applying quality improvement tools is to reduce variation of attributes that affect performance. Therefore, it is useful to consider the application of quality tools to processes or products in order to reduce variation.

A fundamental consideration in improving quality is determining what to measure. It is much easier for a customer to define quality than it is for a manufacturing or design engineer to do so. Customers simply state, “I want this product to work all the time.” Design and manufacturing engineers must translate customer requirements into measurable quantities. Kodak calls these product or process attributes key characteristics (KCs). Xerox refers to them as critical parameters (CPs). Whatever they are called, these attributes of a process or product are intimately tied to quality.

To select meaningful attributes for a manufacturing process, one must consider the function of the process. What is it that this step is trying to accomplish? What is a good output? What is a bad output? These questions are good starting points for determining what about the process should be monitored.

Appraisal Tools

The first step in solving any issue is to understand the actual problem. This may sound like common sense, yet many people are prone to “solution-jumping” before the problem has been fully understood. On the other hand, some people are disposed toward “making a career” out of every issue that comes up. Using established quality tools helps strike a balance between quick fixes and paralysis by analysis.

Customer Warranty Questionnaires

Suppose a toner cartridge remanufacturer is interested in understanding, analyzing and reducing the defective units returned under warranty. When evaluating a cartridge returned under warranty, the first source of information may be a questionnaire that should be returned as part of the warranty claim. Typically, the questionnaire asks for information based on the customer’s observations. This information can help direct the in-house evaluation effort. For example, if a particular cartridge were returned with a questionnaire that specified “wrong cartridge type,” it would not make sense to diagnose the internal components of the cartridge.

Customer feedback is sometimes difficult to interpret because customers lack in-depth knowledge about cartridges. What they typically look for is limited to printer function and image quality. It is useful, therefore, to develop a questionnaire using questions that focus on these aspects. Avoid technical terms when phrasing the questions. It is also a good idea to request that the customer provide a print sample made with the defective cartridge.

Internal Data Analysis Forms

The next step is to collect information about the actual returned cartridge. Every company has its own procedure for this sort of activity. No matter how it is done, in every case, data will be generated. Choosing which information is needed must be established on a case-by-case basis. If the evaluation was being conducted by an OEM, the inspection processes would likely be driven by intimate design and manufacturing knowledge. The inspection would evaluate any deviations from design specifications.

All information collected must be captured. Taking the time to develop a checklist for warranty returns will ensure that all data is consistent, which will facilitate future analysis. For example, if circumstances warrant a print test on the returned cartridge, your form might include spaces to enter observations, such as “blank print” or “black lines” or “excessive background.” These results would then lead to further procedural steps to investigate the particular observations.

Developing a form to analyze warranty returns can be either a very simple process or a more rigorous one, depending upon the potential return on investment. If your company places a premium on knowledge, a more detailed procedure would allow specific data to be captured. On the other hand, a lot of time evaluating warranty returns in detail can be spent without closing the loop with corrective action (paralysis by analysis). The key is to collect enough information to understand the nature of the problem with the objective of correcting it.

Take, for example, a fictitious remanufacturing company that has created a warranty return analysis procedure using a print test and a data collection form. The data collection form is used in conjunction with the print test to evaluate image quality. The form has a checklist for the observed print defects:

• No image
• Excessive background
• Streaking
• Image deletions
• Other

This form was developed using an ongoing process so that it can be continually updated by those employees conducting the analysis. Again, if an OEM were developing this procedure, it would be created by the design and manufacturing engineers. In the case of a remanufacturer, it must be generated by the production staff, using their experience and common sense.

Graphical Data Presentation

Once all this information is collected, an analysis must be done. A qualified person must review it, draw conclusions and then make decisions for action. The most popular way to look at information is using graphical techniques. Determining the best graphical technique depends upon the type of information that is being evaluated.

For data collected with the data form described above, a Pareto chart would likely be the best method. A Pareto chart (at right) is simply a bar chart that plots the frequency of each category or “bin.” The Pareto chart is different from other bar graphs in that the categories are sorted from the greatest to least frequency. This is an excellent way to evaluate quickly the “big hitters” or “vital few.” There are many graphical techniques including radar plots, histograms, scatter plots, and control charts that can also be used to organize and communicate data.

A Sample Analysis

Using our previous example, the data collected over the past four weeks is shown in the Pareto chart. Note that “Background” is the most frequently observed defect after “No Problem Found.” If rate of returns due to background problems constitute reason for future analysis, a quality improvement initiative could be started to reduce or eliminate warranty returns caused by this defect.

“No Problem Found” is a special case. It is perhaps the number one problem in warranty returns in terms of both frequency and frustration. The types of issues included in the “No Problem Found” bin may include problems that leave no trace and may never be resolved. Unfortunately, a surprising number of random factors can come together to create a problem that may never occur again. In these cases, it may not be worth the effort to address each of these limited, random problems.

However, when a particular problem occurs in sufficient numbers in the field but cannot be recreated in the lab, the resolution of the issue will require greater resources. The only way to define the “No Problem Found” cause is to go to the next step and collect information directly from the user. For example, imagine a case where a particular cartridge model is suddenly being returned in relatively significant numbers. In the majority of the cases, the cartridge works perfectly when installed in the test printer. Through further analysis, it becomes clear that just two customers are responsible for the majority of the returns (the further analysis may be angry phone calls in this case!). By discussing the issue with these customers and looking into the configuration of their machines, it is discovered that a printer software glitch has been causing them to believe that the cartridge was defective.

Depending on the type of customer, it may be worth the effort to clear up the problem even though it is beyond the cartridge remanufacturer’s direct responsibility.

Some other questions to explore with customers when faced with no obvious cause for a cartridge’s defect might be: Was the cartridge empty and therefore not performing? Was the seal not pulled? Was the cartridge removed from the printer and shaken, then not replaced properly?

Root Cause Analysis Tools

Root Cause Analysis Tools refer to methods that individuals or teams can use to organize their thoughts and focus their actions. Successful application of these tools will define a root cause of the problem under analysis and hopefully lead to corrective action.

One important tool in this area is the Fishbone diagram. A quality engineer named Kaoru Ishikawa developed the Fishbone diagram. It is sometimes referred to as an Ishikawa or a Cause-and-Effect diagram. The purpose of the Fishbone diagram is to schematically organize all the possible factors that may create a particular effect. A sample diagram (at left, on previous page) is shown for our example of a background development defect.
The information used to illustrate the potential causes comes from knowledge of the process. The Fishbone diagram is particularly well suited to capturing information generated by groups of people who are brainstorming potential causes. Everyone’s ideas are captured, and the group or person responsible for evaluating the problem can focus on the most likely sources.

Another important tool for root cause analysis is Failure Mode and Effects Analysis (FMEA). Product designers and manufacturing engineers use FMEA to formally document the primary failures that can be anticipated with a design or process. The failures are tied to the effects that the failure will cause in the product or process. If a problem-solving team has access to this information, it can be invaluable in understanding root causes based on observed failures.

Test procedures

At this point, the problem-solving team likely will form one or more hypotheses to investigate. It is important to explicitly specify which hypothesis is under consideration in order to design tests that address that issue and no other. This sort of activity takes vigilance and practice because seemingly definitive results can sometimes be misleading. For example, if removing a component makes the problem go away, does putting it back make the problem reappear? If the malfunctioning part is replaced, is something else contaminating that component so that the same problem will reappear over time?

Prevention Tools

Suppose that the background problem outlined earlier led to the discovery that the PCR was not charging the OPC drum sufficiently. More specifically, the problem was tracked to the electrical contact between the PCR and the cartridge. In this case, it was determined that the lubricant on the PCR contact assembly was actually insulative and was creating charging failures. Corrective actions can now be taken. The first is to start using the proper lubricant on the contact assembly. Hopefully, that will completely address the issue, but without monitoring the solution the problem may resurface.

“Failsafing” is the process of applying countermeasures at each point of the process where problems may occur. In our example, failsafing may include a specific grease dispenser that is clearly marked. In addition, operator training, which is another prevention tool, may be included as part of the remedial actions. A more quantitative failsafing technique, however, would be to implement an electrical resistance check through the contact mechanism. Several issues would need to be examined, such as whether the measurement would be performed while the unit is rotating and also whether the resistance must be monitored while the operating voltage is applied.

After the test process is designed and verifiably tied to the failure mode, the data that is collected can be used to monitor quality. The resistance of the contacts can be charted using a process called Statistical Process Control (SPC). SPC provides the techniques to calculate what the process normally delivers, as well as techniques to identify when the process output changes from that normal state. An example of the electrical resistance is shown at the left. Note the marked change in resistance when the old grease finds its way back into the process.

Monitoring quality with SPC is the best safeguard against changes in processes causing product failures. By deciding what to measure, and monitoring and taking action when things go out of control, problems can be stopped before they reach the customer.
Solving problems quickly and effectively when they occur is very important to reducing the total cost of quality. Yet, stopping problems before they occur is still the most effective way to reduce overall quality costs. The prevention tools discussed in this article are only some of the ways problems can be avoided.