/convening-experts

Example: Manufacturing Yield Drop (Single-Round with DMAIC)

User Query: "Our plate coating yield dropped from 96% to 88% after we upgraded to the new coating robot. Same reagents, same protocols."

Expert Panel: Coating Process Analysis

Panel Members:

• Jennifer Kim, Manufacturing Engineer
• Marcus Williams, Six Sigma Black Belt
• Dr. Lisa Wang, Systems Thinker

Jennifer Kim, Manufacturing Engineer

Process Context:

The coating robot upgrade changed three things simultaneously:

1. Dispense mechanism: Peristaltic pump → precision syringe pump
2. Plate handling: Manual loading → automated stacker
3. Environmental control: Open bench → enclosed chamber (better humidity control)

Failure Mode Analysis:

Looking at the 8% yield loss (96% → 88%), that's ~8 plates per 100. Current reject criteria:

• CV% > 15% across wells (indicates coating uniformity issue)
• Binding capacity < 80% of spec (insufficient coating density)

Data from last month (100 plate run):

• 5 plates failed CV% spec (wells at plate edges)
• 3 plates failed binding capacity spec (consistent pattern: rows G-H)

Root Cause Hypotheses:

1. Dispense precision: Syringe pump may have air bubbles → inconsistent volumes (explains CV% failures)
2. Plate handling: Automated stacker impacts → coating solution redistributes before incubation (explains edge effects)
3. Drying kinetics: Enclosed chamber changes evaporation rate → coating dries unevenly (explains row patterns)

Engineering Assessment: The irony is we upgraded for better precision but introduced new failure modes. Need to isolate which of the three changes is the culprit.

Marcus Williams, Six Sigma Black Belt

DMAIC Framework Application:

Define Phase (already done):

• Problem: Yield dropped from 96% to 88% post-equipment upgrade
• CTQ (Critical to Quality): Coating CV% and binding capacity
• Goal: Return to ≥95% yield within 4 weeks

Measure Phase - Current State:

Let's establish measurement system capability first:

Gage R&R Assessment Needed:

• Are CV% measurements reproducible across operators/instruments?
• Is binding capacity assay sensitive enough to detect process variation?

Baseline Data Collection (need 30+ plates):

• Map failures by position on robot stacker (top/middle/bottom)
• Map failures by position in incubator
• Track environmental variables (temp, humidity, time-of-day)

Key Metric: Defects Per Million Opportunities (DPMO)

• Current: 88% yield = 120,000 DPMO (3.0 sigma level)
• Target: 95% yield = 50,000 DPMO (3.3 sigma level)

Analyze Phase - Root Cause:

Jennifer identified three hypotheses. Let's quantify each:

Fishbone Diagram Categories:

Defects (8% yield loss)
├─ Man: Operator training on new equipment?
├─ Machine: 
│  ├─ Syringe pump air bubbles (Jennifer's hypothesis 1)
│  ├─ Stacker impact forces (hypothesis 2)
│  └─ Chamber airflow patterns (hypothesis 3)
├─ Method: Protocol adapted for new equipment?
├─ Material: Coating reagent lot change?
├─ Measurement: CV% calculation method consistent?
└─ Environment: Chamber humidity/temp stability?

Statistical Analysis Plan:

1. Multi-vari study: Separate within-plate variation (wells) from plate-to-plate variation (equipment)
2. Hypothesis testing: Run plates with old method (manual) in parallel for 1 week (5 plates/day × 5 days = 25 plates per method)
   • Statistical power: Can detect 5% yield difference with 80% confidence
3. DOE (Design of Experiments): If parallel testing is inconclusive, run 2³ factorial:
   • Factor A: Pump type (syringe vs peristaltic)
   • Factor B: Loading (manual vs stacker)
   • Factor C: Chamber (open vs enclosed)

Dr. Lisa Wang, Systems Thinker

System-Level Patterns:

This is a "Shifting the Burden" archetype—we upgraded equipment to solve one problem (precision) but created new problems (complexity).

Interconnections Map:

Precision Syringe Pump
├─ (+) More accurate dispensing
├─ (-) More sensitive to air bubbles
├─ (-) Requires different priming procedure
└─ (-) Operator learning curve

Automated Stacker
├─ (+) Higher throughput
├─ (-) Mechanical impact on liquid
├─ (-) Less flexibility in plate positioning
└─ (-) New failure mode (stacker jams)

Enclosed Chamber
├─ (+) Better humidity control
├─ (-) Restricted airflow observation
├─ (-) Harder to troubleshoot in real-time
└─ (-) Creates microenvironment gradients

Emergent Behavior: The three changes interact non-linearly:

• Syringe pump precision helps IF air bubbles are purged (operator-dependent)
• Stacker benefits require gentle plate handling (machine calibration-dependent)
• Chamber control helps IF airflow is uniform (chamber design-dependent)

Feedback Loop Identification:

Reinforcing (Bad):

Equipment complexity → Operator uncertainty → Procedural variations → 
Inconsistent results → More troubleshooting → Less production time → 
Pressure to run faster → Skip steps → More failures

Balancing (Good):

Yield drops → Engineering investigation → Process optimization → 
Better procedures → Yield improves → Production stabilizes

Leverage Points:

1. Highest: Operator training on new equipment (affects all three changes)
2. Medium: Syringe pump priming SOP (Jennifer's hypothesis 1)
3. Lower: Chamber airflow modification (expensive, time-consuming)

Systems Recommendation: Focus on operators as the integration point. New equipment requires new muscle memory. Don't assume protocol transfer is straightforward.

Synthesis (DMAIC Phase: Improve & Control)

Integrated Root Cause Assessment:

All three experts converge on operator adaptation as the highest-leverage intervention:

• Jennifer: "New equipment introduced new failure modes"
• Marcus: "Man" category in Fishbone + need for training validation
• Lisa: "Operator uncertainty" in reinforcing loop

Immediate Actions (Week 1-2):

1. Validate Measurement System (Marcus):

   • Run Gage R&R on CV% and binding capacity assays
   • Ensure defects are real, not measurement artifacts

2. Operator Competency Assessment (Jennifer + Lisa):

   • Observe 3 operators running coating protocol
   • Document differences in technique (pump priming, plate loading, timing)
   • Identify "best practices" from operator with highest yield

3. Parallel Testing (Marcus):

   • Run 25 plates with new equipment (strict SOP adherence)
   • Compare to historical baseline (96% yield on old equipment)
   • Confirm yield gap is reproducible

Optimization Actions (Week 3-4):

4. Standardize Technique (Jennifer):

   • Create detailed work instructions with photos/videos
   • Focus on:
     • Syringe pump priming procedure (bubble elimination)
     • Plate placement in stacker (minimize impact)
     • Chamber loading sequence (environmental equilibration)
   • Train all operators to best-practice standard

5. Equipment Calibration (Marcus + Jennifer):

   • If operator training doesn't close gap, run DOE to isolate equipment variables
   • Adjust pump priming cycles, stacker speed, chamber airflow systematically

Control Phase (Ongoing):

6. Statistical Process Control (Marcus):

   • Track yield weekly with control charts (UCL/LCL at ±2 sigma)
   • Implement reaction plan for out-of-control signals
   • Quarterly capability studies (Cpk monitoring)

7. System Resilience (Lisa):

   • Monitor for "work-arounds" (operators reverting to old habits)
   • Build in redundancy: cross-train operators on troubleshooting
   • Create feedback loop: operators report issues weekly, engineering responds monthly

Success Criteria:

• Week 4: Yield ≥93% (5% improvement)
• Week 8: Yield ≥95% (return to baseline)
• Week 12: Cpk ≥1.33 (process capability sustained)

Risk Mitigation:

• If yield doesn't improve after operator training (Week 2), escalate to equipment vendor for factory calibration
• If DOE shows fundamental equipment limitation, document findings for future capital equipment decisions