For nearly 100 years, the central focus of quality efforts in the U.S. has been on the control of variation. It has proven to be a valuable and effective approach. However, the view that “variation is the enemy” [1] is so deeply ingrained in our thinking that we may frequently overlook some simpler and dramatically effective ways to solve quality problems.
Working with Professor Phil Barkan at Stanford in the early 1990s, we set out to identify what could be done during product concept development to achieve the greatest improvement in quality [2]. Based on the variation quality model, we sought data on the relationship between product level defect or non-conformance rates and the control of variation measured by process capability indices (Cpk).
We were surprised to learn that most organizations had never tried to estimate product quality in this manner, and even more surprised when we could not find any correlation between system level defects and the control of variation. This suggested that variation is not the dominant quality problem in modern production!
We discovered that system defect rates were consistently correlated only with product and process complexity, pointing to the key role of mistakes in product quality [3]. Subsequent studies substantiated our initial conclusions. Although this relationship does not in any way suggest that controlling variation is unimportant, it shows that control of variation has advanced to the point where mistakes are the dominant quality problem today.
In contrast to companies that focus on Six Sigma and rarely achieve defect rates below 1000 to 5000 parts per million (ppm), companies like Toyota maintain defect rates below 30 ppm at a fraction of the cost of traditional quality control methods [3]. Most importantly, the link between complexity and defects points to the two most effective methods for improving quality as: a) mistake-proofing processes, and b) simplifying products and processes.
Mistake-proofing in Pharmaceutical Applications
In a brief review of On-the- Job Training (OJT) documents and Standard Operating Procedures (SOPs) for one pharmaceutical batch process during April 2008, less than five mistake-proofing controls were identified while over 130 mistake-proofing opportunities were found.
Mistake-proofing was virtually non-existent in the process. A three-hour tour of the facility identified yet another potential 20 to 30 mistake-proofing opportunities. Perhaps the best way to illustrate the differences between traditional quality control and mistake-proofing is to provide some specific examples of what we saw during this plant visit for this batch manufacturing process.
Case 1: Guess Which Scale is Which?
In one set up, controls for two digital scales were positioned one above the other at the center of the wall, when the associated scales were located at the left and right corners of the room. As a result, the relationship between the scales and the controls was ambiguous.
Regardless of how much operator training takes place in this environment, errors will occur and time will be wasted in selecting the correct control. After the fact, arrows could be placed by the controllers showing the link between the scale and the controllers. However, if the controls had been placed side-by-side with the left control operating the left scale and the right control operating the right scale,the use of the controls would be more efficient and errors in selecting the controller would have been virtually eliminated.
Case 2: Identify the Water Source
Another case in point. At this pharma facility, three different sources of water at various temperatures are available in the processing room. The controls for each are identical and adjacent to each other. This situation will inevitably lead to errors and waste, since a batch can be scrapped if the wrong water source is selected. Solutions would be simple.
For example, by using a unique tip shape on each nozzle and a matching nozzle restriction in the tank, much as our automobiles do to prevent the use of leaded bas, errors in using the wrong water source could be eliminated. As an alternative, features that hold the hose in the tank during filling could be modified so that only the correct hose could be held in place for each specific operation.
Case 3 : Guess When to Turn Off the Water?
Currently, at this same facility, operators are required to turn off the water manually when the correct amount has been put in a tank. The operators are often performing other tasks while the tank is slowly filling. Getting the correct weight depends on operator diligence.
Because operators may be distracted at the time the water should be shut off, errors can occur. Operators may misread the scale, or even when reading gauges correctly, they may not shut off the water at the correct time, resulting in costly corrections or scrapped batches. Every problem can be solved in more than one way.
One approach that could prevent filling errors would be to use an alarm that starts beeping as the correct water weight is approached. The frequency of the beeping could increase until the correct weight is reached. When the correct weight is reached a steady tone would sound. Thus, the operator would have a warning in advance, allowing time to move to the water control knobs, and would know when the tank is full without even looking at the gauge.
A different approach may be to interlock the scale with a shutoff valve, so the water is shut off when the correct weight is added. A third approach may be to adapt a nozzle like those used to refuel vehicles that shuts down when a specified fluid level is reached. Superior solutions are developed by generating a variety of ideas and comparing them to find the best one.
For example, the nozzle shutoff might not be the right solution for this application if there is a risk of contaminating the nozzle tip, or generating variability in the fill quantity.
Jidoka – Automation with a Human Touch
A powerful concept associated with mistake-proofing is Jidoka, which has been translated from Japanese as “automation with a human touch [4].” A key principle is that the machine should work, while the human works. This requires that machines be capable of self correction or shutting down when processes are completed or when continued operation would damage the product. This principle is not commonly observed in many pharmaceutical plants.
Case 4: Chilling the Tank (During a Force Majeure)
Consider the drug manufacturing facility that we toured in April. In the observed pharmaceutical process, the contents of the tank are supposed to be chilled for roughly 90 minutes. During this time, however, operators must work in a different part of the building. Shutting down the cooling process requires that the operators return to the tank on time.
If they’re distracted, they won’t shut down the cooling process in time and the batch may be ruined. During a recent extreme weather event, operators had to move to a safe location and remain there, resulting is just this type of problem. A simple timer could have been used to shut down the process in case operators were delayed. Better still would be an interlock with a temperature sensor which would shut down the process once it had completed.
The simplest and least expensive approach may be for the worker to wear an alarm that warns him or her that it is time to return for the shut down regardless of where they are within the facility. As mentioned previously, simplifying processes is a key strategy in reducing product defects and waste. Remarkably, our most common response to a quality problem is the implementation of administrative controls.
dministrative controls always make the process more complex, assuring that more errors occur! To perform this pharmaceutical batch process, operators estimated that nearly 100 verification steps were required, in which two individuals would sign off that a task or function has been completed. In a typical verification, for example, one operator signs a checklist that the correct chemical has been selected, and this action is verified by a second person as indicated by initialing a worksheet.
When the material is added to the tank, the operator signs the worksheet, and a second person signs to verify he or she saw the material loaded into the tank. Such verification inspections may reduce product errors, but they themselves are error prone. As a result, product defects are reduced but not eliminated.
Unfortunately, the manual verification dramatically increases paperwork, and paperwork errors. When the sign-off line is found to be blank after production is completed, did the operator simply forget to sign the document or was the step not verified? Huge resources are wasted addressing documentation errors while significant uncertainty remains regarding the true product state or quality because of these problems.
Lessons from Grenade Manufacturing
Mistake-proofing is always a superior solution to manual verification. During the assembly of hand grenades during World War II, if the safety pin were omitted, the grenade would arm and detonate when dropped [3].
To prevent safety pin omission errors two inspectors stood behind the assembler to verify the installation of the pin, one on the right and one on the left. Despite this double verification, at one facility, a grenade was passed down the line without a safety pin, rolled off the table, detonated, and killed a number of workers. The process was changed so that the safety pin had to be in place to hold the grenade on the transportation rack, making it impossible to transport the grenades downstream without correct assembly.
This eliminated the need for verification and assured that every grenade contained a safety pin. In nearly every case the traditional verification can be converted to more effective mistake-proofing. The weigh room in the pharmaceutical plant we visited currently has a digital scale and computer, which must be linked together to achieve mistake-proof material loading.
To ensure that the correct material is selected it should first be staged in correct order of use in the batch process. Just before adding a material to a batch, a barcode scan, or optical character recognition (OCR) on the material container could upload the material identification to a computer, which would then compare the information to the required material specification.
If the wrong material were selected, an alarm would sound, the tank weight would not be uploaded to the computer, and the computer could not advance to the next step until a correct material for the step were scanned. Even better than an alarm would be the automatic deployment of a barrier across the tank opening that would prevent material insertion until the correct material for the process is scanned.
When the correct material is scanned, the current tank weight would be automatically uploaded to the computer, and a sound or tone would indicate that loading may proceed. Rather than manually verifying that material has been loaded, when the addition of the pre-weighed material to the batch is completed, the operator would then have to hit a key on the keyboard to advance to the next step.
As soon as the key is hit, the new current weight would be uploaded to the computer. If the difference between the current weight and previous weight did not match the specification for the step, the program could not advance until the correct amount is added or corrective action is taken.
Unlike manual verification, it would be impossible to proceed unless the right material in the right quantity is actually added to the batch in the correct sequence in the process. Since the weights to be added for each material are generally different, determination that the correct weight has been added provides a better check than a manual verification.
Even more importantly, it is easy to detect and correct additional errors such as the failure to completely empty a barrel of material into the tank. This approach completely eliminates the need for manual verification, and the non-value-adding activity of signing forms. Virtually all problems are caught while executing the process step, rather than attempting to recreate what happened days or weeks after the execution.
The documentation with mistake-proofing is superior, virtually error free, and more succinct with little or no handwritten paperwork. Naturally, for such a system to work correctly we must mistake-proof the labeling of the material, the pre-weighing of the materials, and the loading of the correct materials into each container. In addition empty containers should always been placed in specified locations distinctly different from the full containers.
Quality through Mistake-Proofing Versus Automation
Automation is often justified on the basis of quality improvement. However, many manufacturers have found that non-conformances or defects with automated equipment are not significantly better than what they experienced with manual operations [5].
Mistake-proofing can generally achieve results comparable or better than automation at a fraction of the cost. Certainly, automation may be justified based on product volumes, productivity, and the performance of tasks which are beyond the skill of human operators or where essential processing environments would be hazardous to humans. However, basing decisions to automate on projections for improved quality often yield disappointing results.
Mistakes do occur with the most sophisticated automation, often resulting in large scrap events that offset benefits of sustained periods of controlled production. Problems frequently occur after equipment restarts following power failures. Even when the automated equipment functions exactly as intended, operator errors at the man-machine interface may lead to quality problems.
At the pharmaceutical plant we toured, we briefly visited a highly automated process. In the control room, a single control panel was used to direct the filling of four large tanks. At times, two or more operators are in the room, taking actions on separate tanks. The first operator may take some action on Tank A. Another operator comes to the screen, believing it is displaying controls for Tank C, and direct the loading of the tank with water, ruining product because the operator thought he or she was adding water to Tank C rather than Tank A.
This problem occurs about once every six months at the facility we visited, according to an operator we interviewed, often when new operators are being trained on the equipment. This automated process could be mistake-proofed with four separate controls panels, each physically associated with the controlled tank.
A Cost Effective Solution
The traditional problem solving approach of implementing administrative controls to prevent errors is only marginally effective, while causing a dramatic increase in paperwork and paperwork errors. The extensive use of manual verification in pharmaceutical environments points to mistakes as key quality issue in this industry. Mistake-proofing can achieve nearly defect free production in every environment, including pharmaceutical manufacturing, while significantly improving productivity.
Every manual verification should be replaced with mistake-proofing where possible. Naturally, resources are required to implement sound mistake-proofing. However, the benefit of saving even a single batch of product can be much greater than the cost of implementation of many devices, particularly when focusing on low cost mistake-proofing solutions.
About the Author
C. Martin Hinckley, Ph.D. is President of Assured Quality, Inc. based in Perry, Utah. He can be reached at (888) 599-2100
References
- Juran, J. M. editor-in-chief; Frank M. Gryna, associate editor. Juran’s Quality Control Handbook, 4th Edition, New York, McGraw-Hill, 1988.
- Hinckley, C. Martin, A Global Conformance Quality Model-A New Strategic Tool For Minimizing Defects Caused by Variation, Error, and Complexity, Dissertation, Stanford University, December 1993, pp. 68-70, 78-79, 84, 160.
- Hinckley, C. Martin, Make No Mistake- An Outcome-Based Approach to Mistake-Proofing, Productivity Press, Portland, Oregon, 2001, pp. 3, 18, 260.
- Hirano, Hiroyuki, Editor in Chief, Factory Revolution – A Pictorial Guide to Factory Design of the Future, Productivity Press, Portland, Oregon, 1988, p. 134.
- Hinckley, C. Martin, The Quality Question, Assembly Magazine, November 1997, pp. 44-47.
