Variation
- Common (natural) variation
- The ordinary random variation built into a stable process. It is reduced only by changing the process itself.
- Assignable (special) variation
- Variation with a findable cause — a worn tool, a new material lot, an untrained operator. It is removed by fixing that cause.
- In control
- Only common variation is present; the process is predictable — which is not the same as acceptable.
- Out of control
- A point beyond a control limit, or a non-random pattern such as a run, trend, or cycle.
Control charts plot sample statistics over time against a centerline and control limits set three standard deviations either side. Variables charts measure something (length, weight, minutes); attributes charts count something (defective or not, number of defects).
Charts for variables: x̄ and R
Formula
x̄ chart: centerline = x̄̄, UCL = x̄̄ + A₂R̄, LCL = x̄̄ − A₂R̄R chart: centerline = R̄, UCL = D₄R̄, LCL = D₃R̄| n | A₂ | D₃ | D₄ |
|---|---|---|---|
| 2 | 1.880 | 0 | 3.267 |
| 3 | 1.023 | 0 | 2.575 |
| 4 | 0.729 | 0 | 2.282 |
| 5 | 0.577 | 0 | 2.115 |
| 6 | 0.483 | 0 | 2.004 |
| 7 | 0.419 | 0.076 | 1.924 |
| 8 | 0.373 | 0.136 | 1.864 |
| 9 | 0.337 | 0.184 | 1.816 |
| 10 | 0.308 | 0.223 | 1.777 |
| 12 | 0.266 | 0.283 | 1.717 |
| 15 | 0.223 | 0.347 | 1.653 |
Worked example
Worked example: mean and range charts
A filling line is sampled five times, four bottles per sample. Target fill is 10.0 ounces.
| Sample | Obs 1 | Obs 2 | Obs 3 | Obs 4 | Mean x̄ | Range R |
|---|---|---|---|---|---|---|
| 1 | 10.2 | 10.0 | 9.8 | 10.0 | 10.00 | 0.4 |
| 2 | 10.1 | 10.3 | 9.9 | 10.1 | 10.10 | 0.4 |
| 3 | 9.9 | 10.0 | 10.2 | 9.9 | 10.00 | 0.3 |
| 4 | 10.3 | 10.2 | 10.0 | 10.3 | 10.20 | 0.3 |
| 5 | 9.8 | 10.0 | 9.9 | 10.3 | 10.00 | 0.5 |
Formula
x̄̄ = 50.30 ÷ 5 = 10.06 oz R̄ = 1.9 ÷ 5 = 0.38 ozn = 4, so A₂ = 0.729, D₃ = 0, D₄ = 2.282x̄ chart: UCL = 10.06 + 0.729(0.38) = 10.34 LCL = 10.06 − 0.277 = 9.78R chart: UCL = 2.282(0.38) = 0.87 LCL = 0(0.38) = 0Every sample mean falls between 9.78 and 10.34, and every range falls below 0.87, so the process is in control. Note that the process is centered at 10.06 rather than 10.0 — being in control does not mean being on target.
Charts for attributes: the p-chart
Formula
p̄ = total defectives ÷ total units inspectedσp = √[ p̄(1 − p̄) ÷ n ]UCL = p̄ + 3σp LCL = p̄ − 3σp (use 0 if negative)Worked example
Worked example: proportion defective
Ten samples of 100 finished orders are audited for errors.
| Sample | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 |
|---|---|---|---|---|---|---|---|---|---|---|
| Defectives | 3 | 2 | 4 | 1 | 5 | 2 | 3 | 4 | 2 | 4 |
Formula
Total defectives = 30, total inspected = 1,000, so p̄ = 0.03σp = √[0.03(0.97) ÷ 100] = √0.000291 = 0.0171UCL = 0.03 + 3(0.0171) = 0.081 LCL = 0.03 − 0.051 → 0The highest sample proportion is 0.05, inside the upper limit, so the error rate is stable at about 3%. Reducing it below 3% requires changing the process, not chasing individual samples.
Pareto analysis
Pareto analysis ranks causes by frequency or cost so effort goes where the payoff is. Typically a small share of causes produces most of the problem.
| Complaint | Count | Percent | Cumulative |
|---|---|---|---|
| Late delivery | 80 | 50% | 50% |
| Wrong item shipped | 40 | 25% | 75% |
| Damaged packaging | 20 | 12.5% | 87.5% |
| Billing error | 12 | 7.5% | 95% |
| Other | 8 | 5% | 100% |
| Total | 160 | 100% |
Fixing late delivery and mis-shipments addresses three-quarters of all complaints.
Process capability
Control limits describe what the process does; specification limits describe what the customer requires. Capability compares the two.
Formula
Cp = (Upper spec − Lower spec) ÷ 6σCpk = minimum of [ (Upper spec − μ) ÷ 3σ , (μ − Lower spec) ÷ 3σ ]Cp or Cpk of 1.00 means the process just fills the tolerance; 1.33 or more is the common target.