Process Capability, Control Charts & Variation
A bottle-filling line looks healthy when the average fill is exactly 500 ml. Then the quality engineer plots every few minutes of output and sees the real story - the line swings from underfill to overfill, quietly creating rework, giveaways and customer risk.
That is the power of process capability and control charts: they stop you from managing averages and force you to manage variation.
- Variation is the enemy of reliable operations; the average can look fine while customers still receive defects.
- Control charts answer: “Is the process stable, or is something unusual happening?”
- Process capability answers: “If the process is stable, can it meet customer specifications consistently?”
- Common-cause variation is built into the system; improve it by redesigning the process.
- Special-cause variation is unusual; investigate and remove the specific trigger before changing the whole process.
- Cp measures potential capability; Cpk measures actual capability after accounting for off-centering.
- The correct sequence is always: first stabilise the process, then calculate capability, then improve the process.
Big Picture - Stability Comes Before Capability
Think of quality in three layers. First, measure the output. Second, check whether the process is statistically stable. Third, compare that stable process with customer specification limits. A process can be stable and still incapable; it can also appear capable today but be unstable tomorrow.
Core Explanation - How Variation Becomes a Quality Problem
Variation means the output of a process is not identical every time. In operations, the important question is not “Do we have variation?” because every process does. The question is: “Is the variation predictable, and is it small enough for the customer?”
For any quality characteristic, called a CTQ - critical to quality, there are usually two boundaries:
- USL - upper specification limit: the highest acceptable value from the customer, design or regulation.
- LSL - lower specification limit: the lowest acceptable value.
Control limits and specification limits are not the same. Control limits come from process data. Specification limits come from the customer or design requirement.
The Two Types of Variation You Must Diagnose
A strong answer separates variation into two buckets before suggesting action. This is where many candidates sound managerial instead of mechanical.
The practical rule is simple: do not calculate process capability on an unstable process. If the process is moving because of special causes, Cp and Cpk become misleading averages of a changing reality.
Control Charts - The Stability Test
A control chart is a time-ordered plot of process data with a center line and statistically calculated upper and lower control limits.
It usually has three visual elements:
- Center line: the process average.
- UCL and LCL: upper and lower control limits calculated from process variation.
- Time-ordered points: readings collected in sequence, not rearranged by size.
Common chart choices:
A control chart is telling you: “Is this process behaving like one process, or have multiple hidden processes entered the data?” If you need the broader project sequence around this, revise the DMAIC improvement cycle.
Signals on a Control Chart
Do not stare only at points outside the control limits. Special causes can also appear as patterns inside the limits.
The discipline is not just statistical; it is behavioural. When teams react to every normal wiggle, they create tampering - unnecessary adjustment that increases variation.
Process Capability - The Specification Test
Process capability is the ability of a stable process to produce output within specification limits.
Capability asks a different question from control charts:
- Control chart: “Is the process stable over time?”
- Capability: “Is the stable process good enough for the specification?”
Key Metrics and Formulas
These are the numbers interviewers expect you to know. Do not memorise them as isolated formulas; know what each one decides.
Cp assumes the process is centered. Cpk punishes the process if the mean drifts toward one specification limit. In interviews, say: “Cp tells me potential; Cpk tells me actual delivered capability.”
Worked Example - Calculating Cp and Cpk
A filling process has a target of 500 ml. The customer allows 490 ml to 510 ml. A stable sample gives mean = 503 ml and standard deviation = 2 ml.
The managerial insight: do not buy a new machine yet. First re-center the filling setting closer to 500 ml. If Cpk improves, the issue was centering, not fundamental variation.
Definitions You Can Say in One Breath
- Variation: difference between process outputs over time, even when the same method is followed.
- Common-cause variation: routine variation created by the current process system.
- Special-cause variation: unusual variation caused by a specific event, change or disturbance.
- Control chart: a time-ordered chart with center line and control limits to detect non-random variation.
- Process capability: a stable process’s ability to meet specification limits consistently.
- Cp: potential capability based on process spread versus tolerance width.
- Cpk: actual capability after considering both spread and centering.
Case Study - Asian Paints and the Battle Against Shade Variation
Asian Paints shows why variation control is strategic in a category where customers can visually detect even small differences in shade, texture and finish.
Paint is an unforgiving product. A customer may not understand viscosity, pigment dispersion or batch control, but they immediately notice when two walls meant to be the same shade look different. For a company like Asian Paints, variation is not just a factory issue; it directly affects brand trust at the consumer’s home.
Situation: Paint manufacturing has several CTQs - shade accuracy, viscosity, opacity, drying behaviour, fill quantity and packaging integrity. Variation can enter through raw material lots, mixing conditions, tinting, temperature, equipment wear and handling.
The move: The operating logic is to define CTQs clearly, measure them repeatedly, separate normal variation from abnormal signals, and keep formulations, equipment settings and inspection routines standardised. The primary driver is disciplined measurement and process control around shade and formulation consistency. Supporting drivers include standard recipes, automated production steps, supplier quality control, trained operators and fast feedback from dealers and customers.
Lesson: The company does not win quality by “checking harder at the end.” It wins by reducing upstream variation so fewer defects are created in the first place. That is the exact logic behind control charts and capability: detect instability early, then improve the system so customer-visible variation shrinks.
So what: A control chart is not a statistical decoration. In a consumer-facing manufacturing business, it protects the promise the brand makes every time a customer opens the product.
How AI Changes Process Capability, Control Charts & Variation
AI does not replace SPC logic; it makes variation visible earlier and across more data streams.
- Real-time defect detection: computer vision can inspect surfaces, labels, welds or packaging continuously, turning visual defects into measurable data for control charts.
- Pattern detection before rule breach: machine-learning models can flag drift in machine temperature, vibration, cycle time or defect mix before a point crosses a control limit.
- Root-cause prioritisation: AI can connect defect spikes with shift, supplier lot, operator, machine setting or weather data, helping teams narrow the investigation faster.
The caution: AI can find correlations, but the team must still validate the physical cause. A model may say defects rise on Line 3 during night shift; only process observation proves whether the cause is lighting, fatigue, maintenance timing, material handling or measurement error.
Use ChatGPT or Claude to practise: paste a small process dataset with USL, LSL, mean and standard deviation, then ask it to calculate Cp/Cpk, interpret stability versus capability, and draft an interview-style recommendation. For AI use cases in quality, revise AI in defect detection and root cause analysis.
Interview Relevance
“A process has an average within specification, but customers are still complaining about defects. How would you diagnose the problem using control charts and process capability?”
Use this sentence in interviews: “Averages tell me where the process is centered; control charts tell me whether it is stable; capability tells me whether the stable process satisfies the customer.”
If the interviewer connects this to Six Sigma, remember that capability metrics usually sit inside a larger improvement system. For the distinction between defect reduction and waste reduction, revise what Six Sigma is and how it differs from Lean. For the statistical toolkit around these charts, revise statistical tools every improvement project uses.
Common Mistake
The mistake: calculating Cp/Cpk on an unstable process and declaring it capable. Why it costs you: it shows you do not understand the sequence of SPC. Fix: always say, “First check control chart stability, then calculate capability.”