SPC Training: Statistical Process Control, Control Charts and Capability

Opens with a self-rating, then two recalls: a Ford Ranger airbag scoring setting changed at a supplier's laser process, and lower bolt torque on Hyundai IONIQ 5 vehicles after a wheel alignment error. Turns them into three questions about process change, data and reaction. Introduces XYZ Electronics in Shanghai, a connector and circuit board supplier, and compares sorting at final inspection with charting five pins every hour. Covers Shewhart's first control chart in 1924 and the NIST/SEMATECH e-Handbook as the main source, four myths about SPC, the twelve modules in three parts, calculating on screen with a calculator, the final exam and the certificate. It ends with a recap and three practice questions.

What you will be able to do

• Explain why a control chart shows change while you can still act
• Turn a recall into three questions about process control
• Describe how the course, practice and final exam work

Why it matters

Knowing a process has changed within the hour means checking one hour of parts, not a whole shift of suspect product.

Practice: Practice: Welcome, and why SPC matters · 3 questions

Starts at a plastics moulder in Toronto, where one operator adjusts after every reading and the other reacts only to control chart signals, and the first operator's parts spread wider. From the NIST/SEMATECH e-Handbook, covers Shewhart's chance and assignable causes, Deming's special causes, what 'in control' means, and why limits sit three standard deviations from the centre line, with a false alarm risk of about 0.27%. Explains control limits from process data against specification limits from the drawing, and why specifications never go on a chart of averages. Shows how common causes are reduced by changing the system and special causes removed at the process, and how to compare an adjustments log with the chart. It ends with a recap and three practice questions.

What you will be able to do

• Tell a common cause from a special cause on a control chart
• Explain why over-adjustment adds variation
• Distinguish control limits from specification limits

Why it matters

Leaving normal variation alone and acting only on real signals keeps parts tighter and stops wasted adjustments.

Practice: Practice: Variation: common and special causes · 3 questions

Compares two control plans for the same connector pin at XYZ Electronics: one says check regularly, the other names the gauge, a sample of five pins every hour, the chart and the reaction plan, so every shift charts the same way. Covers SPC as one of the automotive core tools, how the FMEA and special characteristics decide what goes into the control plan, what each control plan row must name, when a reaction is needed because a process is unstable or not capable, and the harmonised AIAG & VDA SPC manual released in July 2026. Makes clear that the customer's own requirements and forms come first, and how to plan SPC for a characteristic before production starts. It ends with a recap and three practice questions.

What you will be able to do

• Explain how FMEA and special characteristics feed the control plan
• Check that a control plan row names method, sample, chart and reaction
• Plan SPC so that every shift charts and reacts the same way

Why it matters

Decisions made once in the control plan mean every shift measures and reacts in the same way, without arguments at the machine.

Practice: Practice: SPC in the core tools and the control plan · 3 questions

Opens at XYZ Electronics, where a new engineer mixes one pin from each of five presses into one subgroup and hides a drift on press three, while a colleague takes five pins in a row from each press and keeps one chart per press. Explains variable data that are measured and attribute data that are counted, the difference between a defect and a defective unit, rational subgroups taken close together under the same conditions, and sampling often enough for how fast the process can change. Shows how measurement variation adds to part-to-part variation, using Minitab's published formulas, and why the measurement system is studied before the chart is trusted. Stresses honest, traceable data. It ends with a recap and three practice questions.

What you will be able to do

• Tell variable data from attribute data, and a defect from a defective
• Form rational subgroups that show real process changes
• Explain why the gauge is studied before the chart is trusted

Why it matters

Well-formed subgroups and a studied gauge give charts that show real drift, instead of wide limits that hide it.

Practice: Practice: Data, subgroups and sampling · 3 questions

Starts with a week on press two at XYZ Electronics, where subgroup averages stay near 5.01 millimetres while the spread inside each subgroup grows, until a range chart shows the ranges climbing and maintenance finds a loose guide. Teaches the method from section 6.3.2.1 of the NIST/SEMATECH e-Handbook: the centre lines, limits of X-double-bar plus or minus A2 times R-bar, and D3 and D4 times R-bar for the ranges, with the factors from the NIST table for subgroups of two to five. Works the limits on screen with XYZ Electronics' pin data, shows how to read the range chart before the averages chart, and covers the average and standard deviation chart with the constant c4 for larger subgroups. It ends with a recap and three practice questions.

What you will be able to do

• Calculate control limits for an average and range chart
• Read the range chart before the averages chart
• Choose the average and standard deviation chart for larger subgroups

Why it matters

Watching both centre and spread catches a press problem before any part falls outside the specification.

Practice: Practice: Average and range charts · 3 questions

Opens at XYZ Electronics, where conformal coating viscosity is measured once per batch with a flow cup, and an engineer wrongly charts two repeat tests of the same batch as a subgroup, so almost every batch looks out of control. Explains why repeat tests are not a rational subgroup and how the individuals and moving range chart uses the differences between neighbouring results. Works the limits step by step: the average plus or minus three times MR-bar divided by 1.128, and an upper moving range limit of 3.267 times MR-bar. Covers the chart's limits as a tool and gives a summary table of the NIST guidance for choosing between individuals, small subgroups, larger subgroups or one chart per stream. It ends with a recap and three practice questions.

What you will be able to do

• Calculate limits for an individuals and moving range chart
• Explain why repeat tests are not a rational subgroup
• Choose the right chart for a set of variable data

Why it matters

Charting one real result per batch shows genuine batch-to-batch changes instead of false alarms from narrow limits.

Practice: Practice: Individuals and moving range charts · 3 questions

Starts on press five at XYZ Electronics, where a wearing guide bush drifts for nine hours before a point leaves the control limits, while press six catches the same drift in the third hour because the team also reacts to patterns. Shows how to divide the averages chart into zones at one, two and three sigma from the centre line, then teaches the Western Electric rules: a point beyond three sigma, two of three beyond two sigma, four of five beyond one sigma, and eight in a row on one side, plus trend patterns and the causes behind them. Covers Nelson's rules and his use of four rules on moving range charts, and the price of more rules in false alarms, so rules are chosen, written down and aligned with the customer. It ends with a recap and three practice questions.

What you will be able to do

• Divide a chart into zones and apply the Western Electric rules
• Recognise runs, trends and other patterns that signal change
• Choose written rules that balance speed against false alarms

Why it matters

Reading patterns, not only points, can catch a drift hours earlier and cut the number of parts that need checking.

Practice: Practice: Reading the chart: out-of-control rules · 3 questions

Opens in the daily meeting at XYZ Electronics, where seven failed boards out of two hundred at final electrical test trigger a twenty-minute search, until a p chart shows that 3.5% sits inside limits running from zero to about 5%. Explains the four attribute charts in two pairs: the p chart for the fraction defective and the np chart for the number defective, both on the binomial model, and the c chart for defects per constant unit and the u chart for defects per unit when the amount inspected varies, both on the Poisson model. Works the calculation for each on screen, and covers when an attribute chart is the right choice, defining defects clearly, and recording how many units were inspected. It ends with a recap and three practice questions.

What you will be able to do

• Choose between p, np, c and u charts for count data
• Calculate limits for an attribute chart
• Define defects clearly and record the amount inspected

Why it matters

Charting counts before reacting stops meetings chasing normal day-to-day variation and frees time for improvement.

Practice: Practice: Attribute charts: p, np, c and u · 3 questions

Starts with two capability studies of one press at XYZ Electronics: a rushed morning of fifty pins from one coil gives a Cpk of 2.1 that fails in production, while a two-week study of twenty-five subgroups on a stable chart gives 1.40 that holds up. Covers the conditions before any index: a stable control chart, a studied gauge, a look at the data's shape, and enough data over a real span of production. Then calculates step by step Cp and Cpk from the within-subgroup sigma, showing what the process can do and how Cpk adds centring, and Pp and Ppk from the overall standard deviation, showing what the customer actually received. Ends the method with what a gap between Cpk and Ppk points to. It ends with a recap and three practice questions.

What you will be able to do

• Check the conditions that must be met before any capability index
• Calculate Cp, Cpk, Pp and Ppk step by step
• Explain what a gap between Cpk and Ppk points to

Why it matters

A capability figure built on a stable process holds up in production, instead of describing one quiet morning.

Practice: Practice: Capability and performance indices · 3 questions

Opens with pin coplanarity at XYZ Electronics, a skewed characteristic with a 0.10 millimetre upper limit, where treating the data as normal gives an index of 1.83 but percentiles of the real data give 1.40. Covers checking the shape before any index, with a histogram, a normal probability plot and the physics of the characteristic, then two remedies the NIST handbook describes: the Box-Cox power transformation and percentile-based indices. Explains how the customer's part approval and specific requirements set the minimum index, the type of study and the report format, and how clause 8.3.3.3 of IATF 16949 handles the customer's symbols for special characteristics through a conversion table. It ends with a recap and three practice questions.

What you will be able to do

• Check the shape of data before calculating a capability index
• Apply a transformation or percentile-based index to skewed data
• Identify the customer requirements that set capability targets

Why it matters

Checking the shape first stops a reassuring but wrong capability number from going to the customer.

Practice: Practice: Non-normal data and customer requirements · 3 questions

Starts on the night shift at XYZ Electronics when the pin-height chart on press three signals at two in the morning: one reaction plan only says inform the supervisor and the press keeps running, the other lists the steps, the press stops within minutes and six hundred connectors are held and sorted. Explains what IATF 16949 asks of reaction plans, then the steps after a signal: stop, contain everything made since the last good subgroup, check, find the cause, fix and record. Covers escalation with clear levels, times and authority, telling the customer at once if suspect parts were shipped, notes on the chart, a recommended reaction plan table, and leaving the process alone when there is no signal. It ends with a recap and three practice questions.

What you will be able to do

• Write a reaction plan with clear steps after a chart signal
• Contain suspect parts back to the last good subgroup
• Set escalation levels, times and authority

Why it matters

A written reaction plan means the press is stopped within minutes, whoever is on shift, instead of running on while people search.

Practice: Practice: Reaction plans and escalation · 3 questions

Opens with two plants that both use SPC well, but only one reduces its spread, scrap and checking costs each year. Reads clause 10.3.1 of IATF 16949 on continual improvement and its note that improvement follows once processes are capable and stable. Teaches a five-step cycle: stabilise, centre, reduce the spread, reset the limits and documents, and repeat. Shows how XYZ Electronics adds up three months of signal notes, twenty signals with nine traced to coil lot changes, to choose its next target, as clause 9.1.3.1 asks, and when a designed experiment helps. Gives a road map for your own plant, a checklist for any chart on the shop floor, the key takeaways, and then the final exam.

What you will be able to do

• Apply the stabilise, centre, reduce, reset and repeat cycle
• Use signal notes on charts to choose the next improvement
• Explain when to recalculate limits after a process change

Why it matters

Using the same charts to pick the next source of variation lowers scrap and checking costs year after year.

Practice: Practice: Putting it together: improving the process · 3 questions