Frequently Asked Questions About Quality

Please contact Media Sciences if your questions are not answered on these pages.

When I talk to knowledgeable quality engineers, they talk of AQL methods that require sample sizes involving hundreds of units. Media Sciences seems to use much smaller sample sizes. Does this compromise your testing?

Obsolete quality methods die hard. AQL methods involving complex tables, many different levels, sample sizes determined by lot sizes, and second sampling options are remnants of outdated methods that neither improved nor maintained quality. Some adherents to this method just do not know that there is a better way. Others love the complexity and job security of the old way. Devotion to obsolete quality methods such as AQL almost destroyed the U.S. automotive, semiconductor, and other industries.

Users often have false confidence in obsolete attribute methods such as AQL. Such methods do not guarantee that good lots will be accepted and that bad lots will be rejected. Instead, there is a significant probability inherent to statistical sampling that a bad lot will be accepted and that a good lot will be rejected. Another problem is that large sample sizes associated with obsolete methods either generate immense amounts of data that no one ever looks at or employ only a few, limited tests that offer little protection to the buyer. Hundreds of diskettes may only be certified for "clip level" or a comparable number of compact discs for only BLER. Lot acceptance or rejection is then based on the wrong tests using obsolete attribute methods and an excessive number of samples. Superior results can be obtained using modern methods employing smaller sample sizes and comprehensive test plans that include proper data analysis and interpretation.

Modern quality methods discard the old attribute approach of classifying each sample as "good" or "bad." No longer are products simply accepted or rejected on a lot basis. Modern variables methods, such as statistical process control, supports predictability of quality, continuous improvement, and preventive action. Manufacturers using modern methods seek to consistently ship products that are in 100% conformance with customer requirements.

Statistical process control, or SPC, methods require continual access to a process using a regular sampling plan that is capable of detecting changes over time. Sample sizes may be as small as three or five, but disciplined data analysis provides valuable information about the process that is not available from older methods. Xbar-R charts and control limits provide real time process and product information that is employed by operators, supervisors, and managers to control the inevitable process variations. Problems are identified and corrected before product quality is adversely affected. Manufacturers solve inevitable quality problems instead of shipping them to the customer who attempts to sort out "good" from "bad" lots.

Independent test laboratories do not have constant access to the production process. Such access is unnecessary if the manufacturer uses SPC methods, and would be ineffective if the manufacturer could not utilize the methods and benefits of SPC. Instead, a hybrid approach is used by the laboratory that provides media users with a clear picture of vendor quality without the expense associated with testing large samples for every lot.

Independent laboratories use uncomplicated LTPD sampling methods with sample sizes that are independent of lot sizes, use only one level of sampling, and do not involve second sampling. Comprehensive tests on relatively small sample sizes then provide excellent protection to the buyer against accepting poor product. Tests are conducted for every quality characteristic that can be measured in the laboratory. Costly manual tests may be conducted on only a few units, while fast, computerized tests probe the full sample. Testing concludes with a careful analysis and review of all test results, often with annotations of either favorable or unfavorable variances. This hybrid of careful analysis and statistical sampling has a proven record of accurate, cost effective vendor qualification.

Although qualification test results enable the customer to select vendors capable of consistently supplying high quality, one-time qualification can generate a false sense of security. Periodic requalification is necessary to detect shifts in baseline quality or out-of-control processes. Regular, detailed feedback to manufacturers supports continuous improvement, and encourages a supportive attitude between buyer and seller. Obsolete quality methods created an advisarial relationship. Modern methods featuring comprehensive tests on modest sample sizes benefit the customer with a supply of high quality products and appropriately reward manufacturers of those products. Use these methods, not obsolete methods that require large sample sizes.