Rebuild the record
- Owner
- The Tier-1's engineers, with our analyst
- Scope
- One table, part by part, from every production log.
- Closed when
- Every part traces to its date, its settings and its material lot.
Case study · Manufacturing · Problem Solving & Statistical Investigation
Second Opinion: Testing a Root-Cause Analysis Against the Data
A Tier-1 supplier had to answer its customer's root-cause analysis point by point. We tested each point against the production record, process and material.
Anonymised project. The client, its customer and the product are not named, and the case carries no figures.
- Client
- A Tier-1 supplier
- Its customer
- An OEM
- Also involved
- The Tier-1's material supplier
- Product
- A component with a recurring defect
- Our part
- A second opinion, working in the background
Situation
A defect came back on a component after a long run without it. Then it showed up in a large share of daily production.
The customer wrote its own root-cause analysis and asked the Tier-1 to answer it point by point. The Tier-1 and its material supplier did not agree on where the cause sat.
How it started
Before it answered, the Tier-1 asked us for a second opinion. Not a new 8D. A review of the analysis that already existed, tested against the production data.
A second opinion from the supplier's side can look like an argument to avoid the cost. So every point had to stand on records both sides could check. The Tier-1 kept all communication with its customer and its supplier. We worked in the background.
What we looked at
We started by rebuilding the record.
- The production record for the whole period, part by part, by production date, not by claim date.
- The onset: the point where the defect started, and what changed at that point.
- The process settings and the material lots in one model, so each effect was read with the others held steady.
- Material lots compared within the same weeks, to keep a drift over time out of the comparison.
- Every point of the customer's analysis, read against that record.
Work packages
We tested the customer's analysis against the records the plant already had.
Test the explanations
- Owner
- Our analyst
- Scope
- The process settings and the material lots in one model.
- Closed when
- An effect counts only if it holds with the other factors held steady.
Review the analysis
- Owner
- Our analyst, with the Tier-1's project lead
- Scope
- A written comment on every point of the customer's analysis.
- Closed when
- Each comment names the record it rests on.
Agree how to test further
- Owner
- The Tier-1's project lead
- Scope
- No new trial until the plan was agreed.
- Closed when
- Plan, sample size and decision rule agreed in writing before any sampling.
How the work ran
- Documents first
- Written comments, each tied to the record it rests on.
- Then a call
- Only where a point needed discussion.
- All along
- The Tier-1 spoke to its customer and its supplier. We prepared the evidence.
What the data showed
- The process settings had an influence.
- The largest share of the variation followed the material supplier's lots.
- The onset lined up with a change in the material, not with a change in the process.
What the data could not show
- The exact mechanism. The data showed an association that held with the other factors held steady. It did not prove a single cause.
- How much the part's own design adds. The analysis kept it as a second factor, not the trigger.
Outcome
The Tier-1 answered its customer with this evidence. The customer revised its analysis: one process factor was removed from its conclusions, and the improvement was tied to the material.
Exit criteria
- Every point of the customer's analysis had a written answer with its evidence.
- Any further trial runs on an agreed plan, with the decision rule set before sampling.
The Tier-1 decided what to send. The customer owned its analysis.
What stayed with the client
- The analysis report the Tier-1 used with its customer and its supplier.
- The rebuilt production record, part by part.
- A written comment on every point of the customer's analysis.
- A rule for any further trial: plan, sample size and decision rule agreed in writing before sampling.
Related services and resources
Talk first
Working through a similar quality issue?
Tell us what is happening, what your customer requires and what your team has tried. We can review the support needed.