DMAIC Project Board

Lean Six Sigma Green Belt project: First Pass Yield at Stage 8, caravan assembly line

Richard Milburn, Erwin Hymer Group UK (Elddis), Cell Leader

Objective: raise first pass yield at the end of the line (Stage 8 PDI)

1Define

Problem statement

Stage 8 First Pass Yield was inconsistent and frequently below target, driving rework, delayed sign-off, and Damaged-in-Production scrap cost above budget.

Project charter

Project: Improve FPY at Stage 8
Method: DMAIC
Scope: Caravan assembly, Stage 8 Pre-Delivery Inspection (PDI) checkpoint
Team: Production, Quality, Engineering (Design Change Requests, DCR)

SIPOC (summary)

Suppliers: Goods-in, upstream stages
Inputs: Torque wrench cal., operator competency, jig cal., DCR status
Process: Stage 8 assembly & PDI check
Outputs: Pass/fail status, fault category
Customers: PDI, downstream stages, end customer

Voice of the customer / critical to quality

First-time pass rate
Consistent, auditable measurement
Reduced DiP scrap cost

2Measure

First pass yield data

Year to Aug 2024, average71.8%

Sep 2024 to Mar 2025, average57.7%

Lowest month (Sep 2024)43.2%

Jul 2025 (to the 15th, 12 days)92.4%

Weeks at 100%2 of 97

Share of vans out of PDI that passed first time (A vans out divided by all vans out), added up across the days in each period. Source: Caravan Combined Performance Tracker, daily records from 1 Aug 2023 to 15 Jul 2025. The two 100% weeks are 2025 week 17 and week 30, each from 3 recorded days.

Data collection plan

2 Y variables, 4 X variables tracked:

Y: Pass/fail status (Avante system)
Y: Fault category
X: Torque wrench calibration
X: Operator competency stage
X: Jig/fixture calibration
X: DCR status

Measurement system

All faults logged on Avante, including on-the-spot corrections. Vans cannot receive final sign-off until every fault clears, which gives an auditable, gap-free measure of first-time pass performance.

Monthly first pass yield (tracker data)

Real monthly figures from the performance tracker. Months with fewer recorded days or vans are less reliable: Aug 2024 has 7 days, Jul 2025 has 12. One day in Aug 2024 was left out because it recorded more A vans than vans out.

3Analyse

C&E matrix: top inputs

Operator competency543

Torque wrench calibration433

SOP / work instruction clarity411

Jig / fixture calibration385

Goods-in component quality381

PFMEA: highest risk priority number (RPN)

Both top failure modes trace to torque discipline:

Torque wrench calibrationRPN 192

Operator torque disciplineRPN 192

Both flagged with weak or no current detection control, so the action is to collect data.

Root cause (5 Why style)

Fault: torque-related rework at Stage 8

Why? No continuous torque reading recorded

Why? Wrench calibration tracked as pass/fail only

Why? No formal continuous calibration log existed

Run your own analysis with the 5 Whys root cause tool.

Regression check

Weekly FPY vs time: quadratic fit R² = 0.275. Time alone is a weak predictor, which suggests the process inputs above, not the passage of time, are the drivers.

4Improve

Interventions

Shop-floor teams given direct ownership of quality at Stage 8
Torque wrench & jig calibration upgraded from attribute (in/out) to continuous logging
DCR escalation used to route recurring faults to Engineering for design-level fix
Cross-functional 3x3 training raised operator capability from about 70% to 100% across 3 roles in 5 months

Result

57.7% to 92.4%

Average Sep 2024 to Mar 2025, then July 2025 to the 15th. Individual weeks reached 100%. The tracker ends on 15 Jul 2025, so it does not show whether this held.

Linked wins

Vehicle defect reduction40%

DiP scrap vs budget, Oct 2023 to Mar 202417.7% under

Team skills competency36% to 86.6%

5Control

Control strategy

Sustain gains through daily SQCDP performance tracking
Continuous torque & jig calibration logs kept live, not retrospective
Root-cause escalation via DCR remains standing process

Control plan

SOP / quality visual aid (QVA) critical coverage97.7%

Close rate on my DCRs93%

DCRs I raised, of 38 site-wide29

Cross-functionality44.2%

Lessons learned

Clear direction and genuine team ownership drove results, not top-down enforcement alone
Continuous data beats attribute data even when it costs more upfront to set up
Data-driven root cause outperforms inspection/rework as the fix

Note on scope

C&E matrix and PFMEA scoring shown here reflect the Green Belt project workup built from this data. The first pass yield figures on this page come from the Caravan Combined Performance Tracker and measure the whole line at PDI. Stage 8 end-of-line figures on the SQCDP dashboard come from a different source and have not been reconciled with these.