According to the late Dr. W. Edwards Deming, the father of Japan's industrial revolution, quality is the most important success factor for sustained competitive advantage. Even one small glitch—think of the 2010 Toyota brake problems—can damage a company’s hard-earned reputation. To judge quality, you will need to know the percentage of your output that is defective. This is estimated using statistical sampling, which looks at a portion of your output to estimate the overall quality.

Determine the population characteristics. This is the universe from which your sample is drawn. If you are in the tools business, then each kind of tool could represent a separate sample population. If you run a transcription business, then your population consists of the transcribed documents.

Define the sample size. If you are making tools, then you might look at random batches of a thousand on the assembly line. If you are in transcription, then you could look at a random sampling of 10-minute audio segments.

Define what constitutes a defect. For a tool, that might be faulty part. For a transcription, it could be a misspelled word that changes the context of a sentence.

Count the number of defects in your sample. In most cases, this means an audio/visual inspection. In some assembly lines, equipment can be programmed to automatically detect and track certain kinds of defects.

Calculate the percent defective. It is the number of defects divided by the sample size, multiplied by 100. So, if one tool is defective out of a sample size of 1,000, your percent defective is 0.1 percent. You then have to determine, as part of your overall quality management program, whether this defect rate meets the acceptable quality level (AQL) of your organization.


According to Dr. Deming, businesses that are successful in today’s global marketplace build quality into their development process right from the start. Measuring and improving your product and service quality should be an everyday process, not something you do once or twice a year.


Statistical sampling introduces errors, known as sampling errors, because you are estimating characteristics, such as quality, by looking at a slice rather than the whole population. You can reduce these errors by increasing the sample size, but that will also increase costs.