The FDA has increased focus on the effectiveness of failure investigations during recent inspections. A review of recent 483 observations and warning letters clearly demonstrate the agency’s current concern about the adequacy of failure investigations to really identify root cause. Effective failure investigation to true root cause is the heart of the CAPA (corrective action preventive action) system.
In an effort to simplify the analysis of data, failures are often grouped into generic categories (human error, inadequate training, procedure not followed, etc.) Unfortunately investigations often stop at this point until the actual cause gets lost in this attempt at simplification. Too often a “band aid” fix is put in place (re-train, change the procedure, etc.)
Most text books on failure investigation identify key steps in the process:
- Identify the general problem - the problem statement should be as clear and concise as possible.
- Gather data or samples as necessary.
- Investigate the failure through experiment and/or examination as necessary - can you re-create the failure? Is it repeatable?
- Brainstorm the causal factors - this is a great place to use tools (fishbone, affinity diagram, 5 whys, etc.).
- Define the root cause based on the data and analysis.
- Develop the hypothesis - based on the data and analysis, what do you think is the real problem? What do you need to do to prevent it from happening?
- Test the hypothesis - design the experiment to demonstrate you have eliminated the source of failure.
How do you choose which tools work best when starting a failure investigation? Some organizations try to force-fit all investigations into the same tools. We all know that there is no "one size fits all." A co-worker recently shared this chart from a Minitab training session:
 |
Probability Plot
This chart was used to demonstrate a process that is out of control at both ends of the process. But is it really? As I look at this chart there are many unanswered questions that come to mind:- What are the specifications for the test being performed?
- How were the acceptable limits determined?
- Was the process validated?
- How much variation can be due to employees and technique?
- Was there any characterization of the material and process as part of the design development activity?
We could assume these are the only issues and start jumping in to answer them immediately but there are many unanswered questions that proper failure investigation techniques could help address.
In this case a fishbone diagram might be the most appropriate place to start. A fishbone diagram – also known as Ishikawa diagrams or cause-and-effect diagrams, can help identify the causes related to a certain event. The diagram got its name because it looks like a fishbone when drawn.
 |
Fishbone diagram |
Here's how to interpret the diagram:
- People: identify the individuals in the process. Person-to-person variation is another great area to investigate. What is the actual process for performing the pull test? Is it something that can have great variability between the individuals performing the test? A gauge R&R should be performed to understand the variation involved. Another suggestion would be to certify the employees performing the procedure to be sure it is done properly.
- Methods: describe how the process is performed. What are specific requirements for the process? Are any critical-to-quality (CTQ) steps involved? When considering the various methods involved, don't forget to consider the design development activities related to the item being manufactured. Is it possible that something was modified in the manufacturing process that is actually crucial to the final output? For example, if the item was properly characterized and validated, the specifications should reflect acceptable tolerances of the pull test. Maybe he data points at both ends of the analysis are actually out of spec because they were outside the original design.
- Machines: identify any special equipment, computers, tools, etc. required to accomplish the job. Be sure to look at calibration and maintenance records for the equipment and/or tools used in the process.
- Materials: what raw materials, consumables, etc. were used to produce the final product?
- Measurements: what data are generated from the process that are used to evaluate its quality.
- Environment: identify the conditions for operations. Consider location, time, temperature and culture in which the process operates.
The real key to a good failure investigation is to understand the various tools available and what you might get from using the tool selected. Other tools you might consider in the failure investigation include:
- 5 whys – This is a question asking method used to explore the cause and effect relationships of a particular issue. The goal of the 5 whys is to determine root cause of the defect or problem. Unfortunately there is a tendency for investigators to stop at symptoms rather than getting to the lower level causes. The investigator may not have enough technical knowledge to get to the real answer. Additionally, the results are not always repeatable. Different individuals performing the analysis may come to different conclusions. When using this tool it is best to work as a team of subject matter experts rather than having one individual try to get to the answer.
- Affinity diagrams - a business tool used to organize ideas and data. It is one of the Seven basic quality tools. This is often used with project management allowing a lot of ideas to be generated as a result of brainstorming sessions. The ideas are then sorted into groups for further review and analysis.
- Is/Is not analysis – this is another very subjective tool and is best completed with a team of subject matter experts. This tool generates the documentation of the circumstances leading to the problem as well as those not associated with the issue at all. It may be helpful with the use of the fishbone analysis diagram.
- Use common sense
- Don't overlook the obvious. Look for the most likely causes first
- The problem should dictate the too, not the other way around
- Avoid analysis paralysis. Don't make a problem more complicated than it is.
- Most problems can have more than one cause. Remember, this is an integrated system. There are many touch points within processes. If you have identified those critical to quality, you can be proactive in preventing failures.
- Tools are generic. You may need to modify the tool to fit the situation.
- Don't generalize the investigation. Ask more "why" questions. Be as specific as possible
