Part A: Foundations
Introduction to quality improvement
Key takeaways
- Control holds a process at its existing capability; improvement changes the capability itself, so correcting a defect only returns the process to where it already was.
- Common cause variation is inherent in the process and is reduced only by changing the process; special cause variation comes from something identifiable and is removed by investigation.
- Reacting to common cause variation as though it were special is tampering: the adjustments add variation of their own and the process gets worse.
- PDCA, DMAIC, DMADV, kaizen, lean and Juran's breakthrough sequence all establish the current state, find causes, change something, verify it and hold the result; they differ in rigour, scale and who does the work.
- The seven basic tools can be used by people who are not statisticians, which is what lets those who do the work carry out improvement; the control chart is the tool that tells an operator whether to act.
Quality improvement is the part of quality management directed at increasing the ability to fulfil quality requirements. Control holds a process where it is; improvement moves it. The distinction matters because organisations routinely mistake heroic firefighting for improvement: correcting a defect returns the process to where it already was, and nothing has improved.
Fundamental issues
Tampering: reacting to common cause as though it were special
Deming's most practically useful warning. When a process varies only from common causes and a manager adjusts it in response to each fluctuation, the adjustments add variation of their own and the process gets worse. Reacting to every result outside yesterday's figure is not vigilance; it is tampering. This is why control charts exist: they distinguish the signal that warrants action from the noise that does not.
The improvement method
The principal improvement methods
| Method | Steps | Best suited to |
|---|---|---|
| PDCA / Deming cycle | Plan, Do, Check, Act | General-purpose improvement of any scale; the underlying logic of every other method here |
| DMAIC (Six Sigma) | Define, Measure, Analyse, Improve, Control | Reducing variation and defects in an existing process where data is available |
| DMADV / DFSS | Define, Measure, Analyse, Design, Verify | Designing a new process or product to a required capability, where DMAIC has nothing to improve |
| Kaizen | Continuous small improvements by those who do the work | Sustained incremental gain; building an improvement culture |
| Lean | Identify value, map the value stream, create flow, establish pull, pursue perfection | Removing waste and shortening lead time |
| Juran's breakthrough sequence | Prove the need, identify projects, organise, diagnose, remedy, hold the gains | Project-by-project improvement, with explicit attention to holding the gain |
| Quality circle / QC story | Select theme, grasp status, set target, plan, analyse causes, act, confirm, standardise | Employee-led improvement of local problems; the chapter on problem solving approaches develops this |
The methods differ less than their names suggest. Every one of them establishes the current state, identifies causes, changes something, verifies the change worked and holds the result. What differs is the rigour of the measurement, the scale of the problem addressed and who does the work. Choosing between them is a question of the problem's size, the availability of data and the maturity of the organisation, not of allegiance to a school.
Tools and techniques
The seven basic tools of quality, associated with Ishikawa, are the foundation. Their significance is that they are usable by people who are not statisticians, which is what allows improvement to be performed by those who do the work rather than by a specialist function.
Beyond the seven basic tools sit the seven management and planning tools, used for less structured problems: the affinity diagram, the relations diagram, the tree diagram, the matrix diagram, the matrix data analysis chart, the process decision programme chart and the arrow diagram. Other techniques recur throughout this paper: the five whys for root cause, failure mode and effects analysis for anticipating failure, benchmarking for establishing what is achievable, and value stream mapping for seeing waste across a whole process rather than within one step.
Choosing the right tool for the question
Where does the problem occur? Use a check sheet to gather data, then stratification to see whether it concentrates by shift, machine or operator.
Which problem should we work on first? Use a Pareto chart, ranked by cost rather than frequency where the two differ.
Why does it happen? Use a cause and effect diagram to generate candidate causes, then five whys to pursue the promising ones to root.
Is this cause actually related to the effect? Use a scatter diagram, then establish a mechanism before believing it.
Is the process stable, and did our change work? Use a control chart, which is the only tool here that answers both.
Is the process capable of meeting specification? Use a histogram against the specification limits, then a capability index.
Self-check
Distinguish common cause from special cause variation, and explain the practical consequence of confusing them. (8 marks)
Describe the seven basic tools of quality, stating for each what question it answers. (12 marks)
Show worked answer
Common and special cause variation (8 marks).
Common cause variation, also called chance or random variation, is inherent in the process as it is currently designed and operated. It arises from the many small factors always present, affects every outcome the process produces, and is predictable within limits although no individual result can be predicted. A process exhibiting only common cause variation is described as stable or in statistical control, which means its future behaviour can be predicted, not that its performance is acceptable.
Special cause variation, also called assignable variation, arises from a specific identifiable factor that is not always present: a new batch of material, an untrained operator, a tool that has worn, an equipment fault. It is not part of the process as designed, it affects outcomes intermittently, and it makes the process unpredictable.
The distinction determines who must act and what action is appropriate. A special cause is removed by local investigation, and the operator or supervisor can usually identify and eliminate it. Common cause variation cannot be removed locally at all, because it is a property of the process itself; reducing it requires changing the process, which is a management responsibility requiring authority over design, equipment, materials or method.
The practical consequences of confusion run in both directions and both are damaging. Treating common cause as special cause is tampering: the manager reacts to each ordinary fluctuation with an adjustment, and because the fluctuation was not caused by anything identifiable, the adjustment is essentially random. The adjustments add variation of their own, so the process becomes less stable than if it had been left alone, and Deming demonstrated this experimentally. It also generates blame, since each fluctuation is attributed to whoever was on duty, which damages the reporting culture on which improvement depends. Treating special cause as common cause is the opposite error: a genuine and removable problem is accepted as an inherent characteristic of the process, so it persists indefinitely and the process never reaches its own designed capability. The control chart exists precisely to distinguish the two, which is why it is the tool that tells an operator whether to act.
The seven basic tools (12 marks).
The check sheet is a structured form for collecting data at the point it arises, designed so that recording is quick and classification consistent. It answers the question: what is actually happening, and how often? It is normally the first tool used because it converts impression into evidence.
The histogram displays the distribution of a variable as a set of frequency bars, revealing the centre, the spread and the shape of the distribution. It answers: how is this characteristic distributed, and does its spread fit within the specification?
The Pareto chart is a bar chart of categories ordered by frequency or by cost, usually with a cumulative percentage line. It answers: which few causes account for most of the effect, and therefore where should improvement effort be directed first? Ordering by cost rather than by frequency is often the more useful analysis, since the most frequent problem is not necessarily the most expensive.
The cause and effect diagram, also called the fishbone or Ishikawa diagram, structures the search for causes of a stated effect under headings, commonly machine, method, material, manpower, measurement and environment. It answers: what could be causing this? Its particular value is that the headings force consideration of categories that the group would not otherwise have raised.
The scatter diagram plots paired observations of two variables to reveal whether a relationship exists between them and of what form. It answers: is this suspected cause actually associated with the effect? Its standing limitation is that association is not causation, so a plausible mechanism must be established before acting on an observed relationship.
The control chart plots a process characteristic over time against a centre line and statistically derived control limits. It answers two questions no other tool answers: is this process stable, that is, exhibiting only common cause variation, and did the change we made actually alter its behaviour? It is the tool that distinguishes the signal warranting action from the noise that does not.
The seventh tool is given differently by different authorities. Stratification separates data by source, such as shift, machine, operator or supplier, to reveal patterns concealed within the aggregate; it answers: does this problem concentrate anywhere? The process flow chart documents the actual sequence of steps in a process; it answers: what actually happens, as distinct from what is supposed to happen, and where in the sequence does the problem arise? Both are legitimate and both are used at the start of most investigations.