Project quality, evidence and process variation
This chapter develops the distinction between quality and grade, between managing quality and controlling quality, and the correct use of the seven basic quality tools. Items also probe the cost of quality and control-chart interpretation, including the rule of seven.
Core concepts
- Quality is the degree to which the product meets requirements; grade is a category of technical features. Low grade is acceptable, low quality is not.
- Prevention over inspection; quality is planned in, not inspected in.
- Managing quality (quality assurance) is process-oriented and performed throughout; controlling quality is product-oriented and inspects deliverables.
- Cost of quality: conformance costs (prevention and appraisal) versus non-conformance costs (internal and external failure). External failures can have substantial consequences, but no cost category is necessarily the largest in every project.
Tools
- Seven basic quality tools: cause-and-effect diagram, flowchart, check sheet, Pareto chart, histogram, control chart, scatter diagram.
- A Pareto chart ranks observed categories and cumulative contributions;80/20 is a heuristic, not a required ratio in every dataset.
- A control chart plots process data against control limits, usually set at ±3 sigma from process data - not specification limits. A chosen run-rule procedure may flag seven consecutive points on one side or a trend for investigation. The rule, chart design and false-alarm behavior must be specified; a signal is not proof of a particular cause.
- Managing-quality tools include affinity diagrams, process decision program charts, interrelationship digraphs, tree diagrams, prioritisation matrices, activity network diagrams, and matrix diagrams.
- Sampling: attribute sampling gives pass/fail; variable sampling measures degree of conformity.
Worked example
A defect log records 120 defects: interface errors 70, data-migration errors 25, UI issues 15, performance 10. Cumulative shares: 58%, 79%, 92%. A Pareto chart shows interface and migration defects are the vital few, so the improvement effort goes there first. A histogram would only show the distribution shape; a cause-and-effect diagram would explore why the interface defects occur.
Traps that cost marks
- Confusing managing quality with controlling quality.
- Treating specification limits as control limits.
- Using a Pareto chart to find root causes; it prioritises, the fishbone analyses causes.
- Assuming sampling proves every unit conforms.
- Claiming quality is solely the QA department's responsibility.