DocQment, newsletter of the STC Quality SIG, June 2002.
Musing on Metrics column
Quality Programs: Six Sigma
Six Sigma is a quality program whose goal is to reduce defects to 3.4 per million opportunities for error (six standard deviations below the norm), reducing costs by reducing waste. Considering the millions of transactions, from burgers to bank checks to gigabytes, handled by corporations every day, such a quality level is a requirement to stay in business. Pioneered by Motorola in the 1980s, Six Sigma has been adopted by a Who's Who list of companies to improve their quality.
There are five steps to achieving Six Sigma quality. Readers of my previous columns will recognize the buzzwords:
- Define critical-to-quality characteristics, those items that are essential for customer satisfaction in the product and business process. (This is another definition for critical success factors.)
- Measure those critical processes that influence the critical-to-quality characteristics and measure the defects generated.
- Analyze the reasons defects are generated by identifying the variable that causes the defect. (This is causal analysis.)
- Improve the process so that the variable stays within acceptable ranges as determined in the analysis phase. (This is continuous improvement.)
- Control the variable by measuring if the process changes have had the desired effects. (The measurements are quality metrics.)
If you think of documentation as a manufacturing process, you can apply Six Sigma principles. I offer these documentation-specific definitions:
- Token - a written word, formatting instruction, XML tag, or graphical element. This definition exposes all aspects of the document, from text to formatting to illustration, to scrutiny.
- Defect - a valid, unique customer problem, in units of defects per million tokens (megatokens), presumably reported via Reader's Comment Forms. I suggest that there are probably ten possible errors for each token.
3.4 errors per million tokens is a well-nigh perfect manual. How can we get there?
- Gather customer-critical success factors.
- Take process measurements at milestones.
- Use document inspection to classify all problems by type.
- Adopt best practices.
- Measure process steps and chart step improvements. (Editors are natural Six Sigma people.)
Can you launch a Six Sigma program for your company's documentation? Consulting firms offer tools to check text. But individual doc groups may not produce the volume of materials to make Six Sigma level measurements. (Another way to put it is that if your doc group is having any measurable quality problems, you're nowhere near the Six Sigma quality level.) If there's no push to reduce the cost of documentation waste, your management may not be interested in pursuing Six Sigma. As with other quality programs, your role may not be to improve your documentation quality but rather to document and help improve the processes of your business as a whole.
Six Sigma has adherents – GE estimates it saves $10 billion a year – and critics as well. It assumes that defects follow a normal distribution (a bell curve), which quality guru Dr. W. Edwards Deming said was nonsense, and that all errors are equal (so that, for instance, a typo equals a missing step), which I say is nonsense. We control the numerator by how we solicit comments, and the denominator by our process definitions, which set the opportunities for error. (If you let me control both the numerator and the denominator, I can give you any number you want.) Finally, an unreported problem is by definition not a defect, which to say the least is counterintuitive. Is Six Sigma the Holy Grail of documentation quality, or "voodoo quality"? I haven't made up my mind yet.
Next time: A measured look at the STC Publications Competition.
Steven Jong, a Senior Member of the STC Boston Chapter, is a technical writer and writing manager. He can be reached at SteveFJong@AOL.com