CPK is the process capability index; 6σ is the quality methodology. The ONLINE platform builds CPK/OEE/6σ analytics on real line data — full-dimension traceability across item/spec/LOT — upgrading quality management from inspection to process control.
Three Levels of Quality Management: Inspection, Statistics, Prevention
Quality management typically evolves through three stages:
- Inspection gatekeeping (Level 1): measure after production, sort out rejects — costly, and it only catches defects that already happened
- Statistical control (Level 2): monitor the process with SPC/CPK, intervene on abnormal trends — from 'sort after' to 'control during'
- Prevention by design (Level 3): guarantee process capability at the design source — ideal, but needs a large base of historical data
Most factories are stuck at Level 1: they inspect, they have data — but the data sleeps in Excel sheets and paper reports, never becoming 'process control'.
What Is CPK: Process Capability Index
CPK answers one question: how consistently is this process producing good parts?
CPK = min(USL - μ, μ - LSL) / (3σ)
USL/LSL: upper/lower spec limits (customer requirements)
μ: process mean (actual level)
σ: process standard deviation (variation)Plainly: specs are the customer's 'pass line', μ is where you are, σ is how stable you are. The higher the CPK, the farther and more stably you sit from the pass line.
| CPK value | Meaning | Action |
|---|---|---|
| > 1.67 | Capability ample | Consider relaxing inspection |
| 1.33 ~ 1.67 | Capability normal | Maintain monitoring |
| 1.00 ~ 1.33 | Capability weak | Needs improvement |
| < 1.00 | Severely insufficient | Stop and improve now |
Why CPK Analysis Often Becomes a Formality
Many factories do CPK too, but with poor results — almost always for the same reason: unreliable data —
- Measurement values typed by hand; omissions and typos make CPK fake
- Sampling is discontinuous (first/last piece only), missing true process variation
- Data scattered across Excel files, impossible to correlate by item/spec/LOT
- CPK is computed and forgotten — never linked to line state (equipment, takt, anomalies)
ONLINE's 6σ/CPK: Built on Real Line Data
ONLINE's CPK/6σ analytics rests on data auto-collected by MES-LINE from the real line, not manual entry:
- Item/spec dimension: CPK computed independently per item and per spec — one line making different products shows each process capability at a glance
- LOT batch traceability: LOT batches correlate station records, inspection data and anomaly records — when a batch fails, the data chain is fully traceable
- CPK + OEE linkage: CPK answers 'how well', OEE answers 'how fast' — efficiency and quality on the same screen, so output never trumps quality
- 6σ methodology: DMAIC (Define → Measure → Analyze → Improve → Control) organizes the analysis flow — from 'knowing CPK is low' to 'knowing why and how to fix it'
- Private LLM assistance: once history accumulates, an on-premises LLM assists quality analysis — querying trends and locating anomalies in natural language, turning data into decisions
Value: From Inspection Gatekeeping to Process Control
| Aspect | Traditional | After ONLINE |
|---|---|---|
| Data source | Manual entry, unreliable | Auto-collected, real and continuous |
| CPK computation | A few sampled numbers | Full data, fine-grained by item/spec/LOT |
| Anomaly response | Found at batch failure | Early alert on process trend |
| Traceability | Excel digging | One-click LOT chain across all data |
| Improvement | Gut feeling | CPK + OEE + anomaly data-driven |
Relation to the TSingley MES II Industrial IoT Three-Loop Control Model
In the TSingley MES II Industrial IoT three-loop control model, CPK/6σ analytics are the decision support of the Work-Order Loop (position loop): the Device Loop secures health (current), the Efficiency Loop secures takt stability (velocity), and the Work-Order Loop secures progress and quality achievement (position) — and quality achievement is measured and driven by the CPK/6σ quantification system.
Quality is not inspected into products; it is controlled into processes. ONLINE's CPK/6σ analytics give 'process control' a foundation of real data.
Suitable for: automotive electronics, machining, battery, 3C — manufacturers with explicit process-capability requirements (customer audits, TS16949/IATF).
From 'Inspection Gatekeeping' to 'Process Control'
Quality is not inspected into products; it is controlled into processes. When CPK is built on real, continuous, traceable line data, 'process control' stops being a slogan.
Are your CPK numbers trustworthy? Want to see how ONLINE grounds quality analytics on real line data? Contact us for a demo.
