For devices in production, the manufacturer should be analyzing its quality problem information (e.g. in-process failures, complaints, field service reports, etc.) with appropriate statistical methodology where necessary to detect recurring quality problems. Examples of acceptable statistical methodology include statistical process control (SPC), periodic frequency distribution analysis and other methods of data trend analysis. Reference Juran "Quality Control Handbook" for examples of acceptable statistical quality control methods.
Service records are a significant source of possible failure data for serviceable devices to identify possible (or potentially) defective devices. Caution should be taken when reviewing service data to separate out routine maintenance (sometimes referred to as preventive maintenance) from servicing requested by the customer to a correct a problem, or that required by a "service bulletin" issued by the manufacturer to correct a problem. Such problems may indicate that possible or potentially defective devices were produced as the result of design or manufacturing problems.
Some examples of routine and non-routine service are:
- Servicing performed at time of a scheduled maintenance per contractual agreements or per device labeling would be considered routine. Although this servicing is considered routine, the records may contain unscheduled and non-routine maintenance such as a convenient time to perform a product recall or market correction.
- Non-scheduled, under warranty, performing product recalls or market withdrawals are some examples of non-routine servicing.
The records reviewed should start from the most current and work back toward a 24 month maximum time period. The selection of the number of service records for review is governed by the form of record availability (electronic or hard copy) and the experience and skill of the Investigator in reviewing a large volume of hard copy or electronic records.
Determine if the firm has analyzed the service records to identify existing and potential causes of nonconforming product or other quality problems. Before reviewing service records, review any trending information that the firm has performed and review any corrective actions the firm may have already taken.
The review of service records should be done by grouping "field of information" to determine any potential trends for possible defective devices. The grouping of data is performed by sorting the data by different fields and combination of fields to determine any relationship of product failures to product design and/or manufacturing processes.
Suggestions for grouping would be by model, serial number (lot or batch number), or component failures to focus the inspection to possible product design or manufacturing problems. The serial numbers and corresponding device history records could point to manufacturing problems with out-of-specification components, inadequate employee training, acceptance testing problems, etc.
Serial number grouping can also reveal repeat failure modes indicating an inadequate corrective action. A common repeat service action is to repeatedly replace computer processing units (CPU) with the same failure mode recurring. This is a common ineffective fix when the problem is software related but the symptom is always computer related.
Grouping by service site - area of country, type of user and time of failure - could point to product design defects. For example, if the failures appear to be related to dry climate areas, the device may not have adequate protection from Electrostatic Discharge (ESD) damage. If the user type is emergency medical service (EMS), the device may not be robust enough for vibration, shock, electromagnetic compatibility (cell phones, radio transmissions, etc.) and other environmental stresses encountered with EMS services.