Quality Advisor

Walter Shewhart published the first control chart at Bell Labs in 1924. A century later, statistical process control is still the foundation of plant-floor quality monitoring — because the question it answers ("is this process in a state of statistical control right now?") is the question every manufacturer needs to answer continuously, regardless of how sophisticated the surrounding software stack becomes.

This guide is written for quality engineers, plant managers, and continuous-improvement practitioners who need to either start an SPC program, modernize one inherited from the 1980s, or evaluate which SPC platform fits their plant. It draws on practitioner content originally developed across three legacy Advantive quality brands — InfinityQS (cloud and on-premise SPC for enterprise manufacturers), SQCpack and GAGEpack (PQ Systems' SPC and gage management products), and WinSPC (DataNet's real-time plant-floor SPC) — consolidated under the Quality Advisor banner. The 43 spoke articles linked from this page hold the full technical detail; the pillar gives you the map.

What is Statistical Process Control?

Statistical process control is a quality methodology that uses statistical methods to monitor a manufacturing process and detect when it has drifted out of control. It separates routine "common-cause" variation that is inherent to the process from "special-cause" variation that signals something has changed — so operators can investigate and correct before defects propagate downstream.

The methodology rests on two ideas. The first is that every process has natural variation: no two parts are identical, even from the same machine on the same shift. The second is that this natural variation is statistically describable — and as long as the variation stays within predicted bounds, the process is "in control." When it breaks those bounds, something has changed (a tool wore out, a setting drifted, an operator made a substitution), and that change is detectable in the data before it becomes a defect on the loading dock.

The mechanism is the control chart: a time-series plot of the process output with three lines on it — a center line at the historical average, and upper and lower control limits set at ±3 standard deviations. Points inside the limits are routine. Points outside the limits, or non-random patterns within the limits (seven points trending in one direction, two of three points beyond two sigma, etc.), are signals to investigate.

For the full beginner's primer including process behavior, distributions, sampling, and the difference between control limits and specification limits, see the dedicated SPC 101 guide. For why SPC matters specifically in manufacturing operations and how it ties to cost of quality and shop-floor decisions, see SPC in Manufacturing. For a shorter framing on what SPC is and what tools it includes, see What is SPC?.

Which control chart should you use?

Choose by data type and subgroup size. For continuous (variables) data with subgroups of 2-10, use X-bar and R. For continuous data measured individually, use X-MR. For attribute defectives data with constant subgroup size, use np-chart. For attribute defects data per unit, use c-chart or u-chart. For rare-event data — time or quantity between failures — use g-chart or t-chart.

The choice of control chart is governed entirely by what kind of data you have. Variables data is measured on a continuous scale: a diameter in millimeters, a fill weight in grams, a tensile strength in newtons. Attribute data is counted: number of defective units in a lot, number of scratches per panel, time between equipment failures. The wrong chart for the data type produces statistically invalid limits — points appear "out of control" when the process is fine, or stay "in control" while defects accumulate.

Within variables data, the choice between X-bar and R, X-bar and Sigma, and X-MR (individuals) depends on subgroup size. X-bar/R is the workhorse for subgroups of 2 to about 10, X-bar/Sigma is preferred for subgroups of 10 or more, and X-MR is the only valid choice for individual measurements (continuous-process monitoring, expensive or destructive tests, low-volume production).

Within attribute data, the choice splits on what you're counting and whether subgroup size is constant. P-chart and np-chart track proportion or number of defective units (each unit is pass/fail). C-chart and u-chart track count of defects (a unit can have multiple defects). The "p" and "u" variants handle variable subgroup sizes; "np" and "c" assume constant subgroup size.

For rare events — where defects are so uncommon that traditional charts would show mostly zeros — use the g-chart (counts between rare events) or t-chart (time between rare events). For exploratory analysis of time-series data before applying a formal chart, the run chart and median chart are useful first looks. For the complete decision logic and edge cases, see the control chart selection guide. For how to read a control chart once you've built it — the Western Electric rules, pattern recognition, and the difference between special- and common-cause patterns — see Interpreting Control Charts.

How do you measure process capability?

Process capability compares what a process produces to what the specification requires. Cp measures whether a process could fit within spec if perfectly centered (spread only). Cpk measures whether it actually fits, accounting for centering. Cpm penalizes deviations from a target value. Ppk uses long-term variation rather than short-term capability variation. A Cpk of 1.33 or higher is the typical industry threshold for an acceptable process.

A process being "in control" (stable) is not the same as the process being "capable" (producing parts that meet spec). A process can be in perfect statistical control and still produce 30% scrap if its natural variation is wider than the engineering tolerance — control just means the variation is predictable, not that it's small enough. Capability indices measure the relationship between the predictable variation and the spec window.

Cp is the simplest: (Upper Spec Limit − Lower Spec Limit) / (6σ). A Cp of 1.0 means the natural process spread equals the spec window exactly — any drift in centering pushes parts out of spec. A Cp of 1.33 (the common industry floor) gives a 25% buffer. A Cp of 2.0 (the Six Sigma target) gives a 50% buffer. Cpk is Cp adjusted for centering — specifically, the minimum of (USL − mean)/3σ and (mean − LSL)/3σ. A Cp of 2.0 with a Cpk of 0.8 means a process that could be very capable but is badly off-target.

Cpm adds a penalty for deviating from a stated target value, not just for the centering relative to the spec limits — useful when the target itself matters (a label fill weight, a torque setting) rather than only fitting between min and max. Pp and Ppk use the same formulas as Cp and Cpk but with long-term ("performance") standard deviation rather than the within-subgroup ("capability") standard deviation. Pp/Ppk are typically lower than Cp/Cpk because long-term variation includes shift-to-shift and day-to-day drift. Use Cpk for ongoing process monitoring; use Ppk for long-run validation audits.

For the step-by-step walkthrough of calculating capability from scratch — sketching the distribution, calculating sigma, drawing spec limits, computing the indices — see Process Capability: Step-by-Step Walkthrough. For the formula references and worked numerical examples, see Capability Formulas. For FAQs on which index to use when, see Process Capability FAQ.

What is Measurement Systems Analysis?

Measurement Systems Analysis (MSA) evaluates whether the measurement system producing your data is trustworthy enough to act on. The core technique, Gage R&R, decomposes total measurement variation into the gage itself (repeatability) and the operators using it (reproducibility). A measurement system that contributes more than 30% of total observed variation is generally considered unfit for SPC decisions.

The output of every SPC system is only as good as the data going in, and the data going in is only as good as the measurement system producing it. If two operators measuring the same part with the same gage get different results, the variation you see on your control chart is partly process and partly measurement — and you can't separate them without an MSA study. Industries that mandate MSA include automotive (IATF 16949), aerospace (AS9100), and medical devices (ISO 13485).

The standard MSA technique is the Gage R&R study: a designed experiment in which multiple operators measure multiple parts multiple times, and ANOVA decomposes the resulting variance into part-to-part, operator-to-operator (reproducibility), and trial-to-trial (repeatability) components. The total Gage R&R variance is compared to total observed variance — under 10% is excellent, 10-30% is conditionally acceptable, over 30% means the measurement system is not fit for purpose for the characteristic in question. See Measurement Systems Analysis: Complete Practitioner Guide for the full study design including bias studies, linearity, stability, and how to interpret the results.

MSA's prerequisite — and a topic that drives more measurement disputes than any technical issue — is the operational definition: a written, unambiguous specification of exactly what is being measured, how, with what gage, under what conditions, by an operator with what training. Two operators measuring "the diameter" of a part can produce wildly different results if one measures at the lip and one at the midpoint. See Operational Definitions in Quality for the practitioner-level guide to writing operational definitions that hold up in audit.

How does SPC fit into a quality management system?

SPC is the real-time monitoring engine inside a broader quality management system (QMS). The QMS defines what gets measured, how often, with what specification, and what corrective action happens when a process drifts. ISO 9000 and the seven quality management principles provide the framework; SPC provides the data that drives decisions within that framework.

A QMS is a set of policies, processes, and documented procedures that an organization uses to ensure consistent quality output and continuous improvement. The most widely adopted QMS framework is the ISO 9000 family, anchored by ISO 9001 (the certification standard) and built around seven quality management principles: customer focus, leadership, engagement of people, process approach, improvement, evidence-based decision making, and relationship management. SPC is one of the primary techniques the "evidence-based decision making" principle relies on.

The relationship between QMS and SPC is directional. The QMS defines a quality management plan: which characteristics on which parts are critical to quality, what their specifications are, how often they should be sampled, and what to do when they drift. SPC executes that plan continuously — recording the data, computing the charts, flagging out-of-control signals. When SPC raises a signal, the QMS defines the corrective and preventive action (CAPA) workflow: investigate the special cause, contain affected product, fix the root cause, validate the fix. Quality metrics — first-pass yield, scrap rate, cost of poor quality, customer reject rate — are reported up from SPC into the QMS dashboards. See Quality Metrics That Matter in Manufacturing for the metrics that matter.

The classic dichotomy is quality management (the system) versus quality control (the inspection-based catch-and-reject approach that QMS replaced as the dominant model in the 1980s). See Quality Management vs. Quality Control in Manufacturing for the historical evolution. For organizations evaluating modernization — replacing paper-based QMS or on-premise legacy systems with cloud-based quality intelligence platforms — see Digital Transformation in Manufacturing Quality and How to Choose a Manufacturing Quality Intelligence Platform.

Frequently asked questions

Continue learning

The full Quality Advisor library — 43 deep-dive articles and a 281-term glossary — is organized into seven topic clusters below. The most-referenced spokes by historical traffic:

• Process Capability Index (Cp): The Complete Guide
• Operational Definitions in Quality: Practitioner Guide
• Process Capability: Step-by-Step Walkthrough
• What is Statistical Process Control (SPC)?
• X-bar and R Chart: Complete Guide
• Cpk: Process Capability Index Explained
• G-Chart: Rare Event Control Chart
• Histogram: Shape, Statistics, and Interpretation