Customer Case Study · Quality Root Cause Assistant
On the cable manufacturer's insulation lines, the same defects kept coming back: pinholes, thin insulation, off-centre conductors. HMC built a root cause assistant for the quality team that joins rejection logs with line data, ranks the defects that cost most, scores every possible cause against evidence, and drafts the 8D report.
Figures indicative; validated metrics available on request.
Every insulation line produces thousands of kilometres of wire a month. A single pinhole means cutting out a length; thin insulation can scrap a whole coil.
Rejection registers, spark-tester counts and extruder settings lived in different sheets and logbooks.
Teams chased the defect seen most recently, not the one costing the most metres and rupees.
Cause lists were brainstormed in meetings; few causes were ever tested against data.
Customer complaints waited around three weeks for a root cause and corrective action report.
Built with the quality and production teams on the house wire and power cable insulation lines.
Select a step to see what happened.
Six months of rejection data walked through with QA and production.
Deliverable: Defect dictionary and data map.
Every fishbone cause written as a test on line data.
Deliverable: Cause library and scoring rules.
One workflow from alarm to report.
Deliverable: Root cause assistant.
The Pareto opens every morning meeting.
Deliverable: Huddle routine and results on two lines.
Same cause library, local data.
Deliverable: Roll-out plan.
Thirty days of synthetic data from four lines. Inject a fault on the live line, find the costliest defect on the Pareto, see which causes the data supports, and get the 8D report.
Optional. Without a key, the 8D and 5 Whys come from the cause library and the data. Add your OpenAI key to write the 8D from your own shop-floor notes and to classify a photo of a defect.
Copper conductor runs through the extruder, cooling trough, laser diameter gauge and spark tester. Inject a fault and watch what the gauges see: diameter drifting out of spec, the conductor going off-centre, the spark tester firing.
Simulated line; values are illustrative.
Thirty days of rejections ranked by rupees, metres or count. The bars left of the 80% line are where most of the money goes. Click a defect to analyse it.
| Defect | Count | Metres | ₹ scrap |
|---|
Synthetic rejection log, September 2026, lines EXT-1, EXT-2, EXT-3 and SHE-1. Scrap valued at product cost per metre.
Every cause on the fishbone is tested against 360 production runs. The bar shows how much more often the defect occurs when the cause is present. Click a cause for the evidence and the 5 Whys.
The eight disciplines filled from the data: problem, containment, root cause with evidence, corrective and preventive actions. Add your own notes and GenAI rewrites it; or upload a defect photo to classify it.
The assistant drafts; the quality engineer confirms the root cause and signs the 8D.
With moisture checked at the hopper, die changes on hours and speed ramps capped, insulation thickness now varies far less and stays well inside the limits.
Control chart (X-bar) with control limits and minimum spec
Capability curves against spec limits, with Cpk
Figures indicative; validated metrics available on request.
The Pareto in money terms tells the team which single defect to fix first.
A fishbone cause that can't be checked against data is an opinion. Write each one as a test.
A root cause is confirmed only when the chart shows the process staying in control afterwards.