Quality Management

Quality touches every process: what you receive, what you make, what you ship, and what comes back. In regulated industries it’s the module that keeps the company licensed.

The quality processes

INCOMING          IN-PROCESS         FINAL           CUSTOMER
Goods receipt  →  During production → Before ship  →  Complaints
inspection        checks              inspection      returns
     │                 │                  │              │
     └─────────────────┴──────────────────┴──────────────┘
                              │
                    Non-conformance → Investigation → CAPA
                              │
                    Supplier quality / Audits / Certification

Inspection lots

The core transactional object:

inspection_lot(
  id, lot_number, inspection_type,
  material_id, batch, plant_id, storage_location,
  reference_document,        -- PO, production order, delivery
  quantity_inspected, uom,
  inspection_plan_id, sample_size,
  status,                    -- CREATED | IN_INSPECTION | RESULTS_RECORDED | CLOSED
  usage_decision,            -- ACCEPT | REJECT | ACCEPT_WITH_DEVIATION | REWORK | SCRAP
  decision_by, decision_at,
  created_at, due_date
)
inspection_characteristic_result(
  lot_id, characteristic_id, sample_number,
  measured_value, unit, qualitative_result,
  lower_spec, upper_spec, target,
  evaluation,                -- OK | NOK
  inspector_id, recorded_at, equipment_id
)

Inspection lots are created automatically by: goods receipt, production confirmation, delivery creation, batch expiry re-inspection, or a periodic stock inspection schedule.

Inspection plans

Inspection plan for Fabric RM-300, incoming inspection

Char  Description        Type          Target  LSL    USL   Method     Sample
Visual defects     Qualitative   —       —      —     Visual     AQL 2.5
GSM (weight)       Quantitative  180     175    185   Balance    5 pcs
Colour match       Qualitative   —       —      —     Lightbox   3 pcs
Shrinkage %        Quantitative  3.0     —      4.0   Wash test  2 pcs
Tensile strength   Quantitative  450     400    —     Tensometer 3 pcs

Characteristic types:

Quantitative — a measured value against a specification range
Qualitative — pass/fail against a catalogue of defect codes
Required vs optional — some characteristics are mandatory for the usage decision
Destructive — the sample is consumed and must be scrapped

Sampling

Testing everything is usually impossible or uneconomic. Statistical sampling decides how much to check.

AQL — Acceptable Quality Level (ISO 2859 / ANSI Z1.4)

Lot size 1,200, General Inspection Level II, AQL 2.5

→ Sample size code letter K → sample 125 units
→ Accept if ≤ 7 defects, reject if ≥ 8

Skip-lot and reduced inspection: a supplier with a consistent record earns reduced sampling; a failure moves them to tightened inspection. This is both a cost saving and a supplier incentive.

Sampling procedures

100% inspection — critical characteristics, safety items
Fixed sample — always N units
Percentage — 5% of the lot
AQL-based — statistically derived, the standard for incoming goods
Skip lot — inspect every Nth lot
Dynamic modification rules — the level moves automatically based on recent history

Usage decision

The formal outcome that releases or blocks the stock:

Decision	Stock movement
Accept	Inspection stock → unrestricted
Accept with deviation	Unrestricted, with a documented concession and approval
Rework	To rework order, then re-inspect
Reject / return to vendor	Return delivery, vendor claim
Scrap	Written off
Downgrade	Moved to a lower-grade material or valuation type

The decision must be traceable: who decided, when, on what evidence, and under what authority. In regulated industries only a qualified person (QP) may release, and the system must enforce that.

Statistical Process Control (SPC)

Monitoring a process to detect drift before it produces defects.

Control chart (X-bar and R):

UCL ─────────────────────────────────────
                    ●
              ●          ●
CL  ──────●───────●───────────●──────────
        ●                          ●
                                        ●
LCL ─────────────────────────────────────

Western Electric / Nelson rules flag out-of-control conditions:

Any point beyond 3σ
9 consecutive points on one side of the centre line
6 consecutive increasing or decreasing
14 alternating up/down
2 of 3 beyond 2σ on the same side

Process capability

Cp  = (USL − LSL) / 6σ           potential capability
Cpk = min[(USL − μ), (μ − LSL)] / 3σ    actual, accounting for centring

Cp = 1.33 → capable (4σ)
Cp = 2.00 → six sigma
Cpk < Cp  → the process is capable but off-centre

The Cp vs Cpk gap is diagnostic: if Cp is fine and Cpk is poor, you don’t need a better machine, you need to re-centre the setting. That distinction saves capital expenditure.

Non-conformance and CAPA

Non-conformance detected
      ▼
NC record: what, where, when, how much, severity
      ▼
Immediate containment: quarantine stock, stop the line, trace forward and back
      ▼
Investigation: root cause analysis
      ▼
Correction:  fix this instance
Corrective:  fix the cause so it doesn't recur
Preventive:  fix similar risks elsewhere
      ▼
Effectiveness check (after N days/units — did it actually work?)
      ▼
Close

Root cause techniques: 5 Whys, Ishikawa/fishbone (Man, Machine, Material, Method, Measurement, Environment), 8D (the automotive standard), FMEA, fault tree analysis.

The effectiveness check is the step everyone skips, and it’s the one that distinguishes a real CAPA system from a paperwork exercise. If you build CAPA workflow, make the effectiveness check a mandatory, scheduled, un-skippable task.

Certificates

Certificate	Content
CoA — Certificate of Analysis	Actual measured results for a specific batch
CoC — Certificate of Conformance	A statement that it meets specification
Mill certificate / EN 10204 3.1	Material composition, for metals
Test report	Full inspection results

Customers demand CoAs with every shipment in chemicals, pharma, food, and metals. The CoA is generated from inspection results for that specific batch — which means the traceability chain from delivery → batch → inspection lot → characteristic results must be complete and queryable. Build it as a report early; it’s requested constantly.

Supplier quality

supplier_quality_score(
  vendor_id, material_group, period,
  ppm_defective,              -- parts per million
  lots_accepted, lots_rejected, acceptance_rate,
  on_time_delivery_pct,
  complaint_count, complaint_severity_weighted,
  corrective_action_response_days,
  audit_score, certification_status,
  overall_score, classification    -- A | B | C | BLOCKED
)

Uses: source list restriction (only A and B suppliers for critical materials), quota arrangements weighted by score, inspection level (A suppliers get skip-lot), and commercial negotiation.

Supplier complaints / 8D — a formal claim to the vendor with a required response, tracked to closure, often with a cost recovery claim attached.

Quality costs

The classic framework — and the argument for investing in prevention:

Category	Examples
Prevention	Training, process design, supplier development, FMEA
Appraisal	Inspection, testing, calibration, audits
Internal failure	Scrap, rework, downgrading, re-inspection
External failure	Warranty, returns, recalls, complaints, lost customers, liability

The rule of ten: a defect costs ~10× more to fix at each stage it escapes. Caught at incoming inspection: ₹100. On the line: ₹1,000. At final inspection: ₹10,000. At the customer: ₹100,000. In a recall: unbounded.

A quality cost report by category is a genuinely high-impact thing to build, because it converts quality from a compliance conversation into a financial one.

Standards and regulation

Standard	Domain
ISO 9001	General quality management systems
IATF 16949	Automotive — adds APQP, PPAP, FMEA, MSA, SPC
AS9100	Aerospace
ISO 13485	Medical devices
GMP / cGMP	Pharma and food manufacturing
HACCP / ISO 22000 / FSSC	Food safety
21 CFR Part 11	FDA electronic records and signatures
ISO 17025	Testing laboratory competence

Automotive core tools (IATF 16949)

APQP — Advanced Product Quality Planning, the phased launch process
PPAP — Production Part Approval Process; the customer approves your part and process before you may ship
FMEA — Failure Mode and Effects Analysis, with Risk Priority Numbers
MSA — Measurement System Analysis (gauge R&R): is your measuring device itself reliable?
SPC — as above

If you work on an automotive supplier’s ERP, these five acronyms will appear in every requirement document.

Calibration

Measuring equipment that has drifted invalidates every measurement it took since the last good calibration.

equipment(id, code, type, location, manufacturer, serial,
          calibration_interval_days, last_calibration, next_calibration,
          status, certificate_reference)

Requirements: calibration schedule with alerts, block use of overdue equipment, retain certificates, and — critically — traceability from inspection results back to the equipment used, so that a failed calibration can trigger a review of everything it measured.

Developer notes

Inspection lots are created automatically by other processes. Get the triggers right, or QC never sees the stock.
Stock in inspection is a distinct stock type — not available to sell, but on the balance sheet.
Usage decision drives a stock movement. They are one transaction, not two.
Results are append-only. A correction is a new recorded result with a reason, never an overwrite.
CoA generation must be a first-class report from batch + inspection results.
Specification limits are versioned and dated. A batch is judged against the spec in force at the time.
In regulated environments: audit trail on everything, electronic signature on decisions, and a validated system. Design for it from the start; retrofitting Part 11 compliance is brutal.