Project SENTINEL: Lessons Learned from a Label Damage Reduction DMAIC Project

Report Context

The Story Starts with a Problem That Looked Smaller Than It Was

One of the important lessons from this project is that apparently cosmetic defects can be systemically serious. The damaged-label issue was not just about appearance. It affected regulatory legibility, product traceability, rework labor, replacement-label consumption, on-time delivery, and customer trust. External complaint volume tied to cosmetic label defects had grown 23% year over year. Internally, the defect was responsible for 38% of all cosmetic nonconformances at final inspection.

The charter problem statement was explicit: between January 1 and December 31, 2025, 17.6% of MC-4400 controllers arriving at final inspection exhibited one or more damaged labels, defined as scratched, torn, cut, abraded, or illegible. That translated to 176,000 DPMO at the unit level, far below the plant’s 4.0 sigma aspiration. The goal was to cut the defect rate to 2.0% or less by June 30, 2026, achieve at least 3.5 sigma and Cpk of 1.33 or better, and do it without increasing assembly cycle time by more than 3%.

What the Team Did Right in Define

The Define phase did not stop at a complaint summary. It translated the voice of the customer into measurable CTQs. Customers wanted labels with no visible damage, compliance markings that remained readable, barcode and serial labels that scanned on the first try, and labels that stayed intact through the product life. That moved the conversation from “people are scratching labels” to “the process is failing a defined set of customer and regulatory requirements.”

The team also mapped the process and confirmed that labels were applied at Step 40 and then passed through six downstream handling opportunities before final inspection. That process mapping mattered because it reframed the problem from an inspection issue to a handling-exposure issue. A key lesson here is simple: if the product receives a fragile feature early in the flow, every downstream touchpoint becomes part of the defect system.

Measure Phase: The Team Earned the Right to Trust Its Data

Another strong lesson from the report is that baseline data only matters if the measurement system is credible. The team did not rush straight into Pareto charts. It first created an operational definition for label damage, built a defect-classification guide with photo standards, and ran an attribute agreement study before collecting the main baseline.

That discipline paid off. The report explicitly calls out that catching a Kappa problem with one inspector early prevented weeks of bad baseline data. After corrective action, the measurement system was released with inspectors performing at 0.87 Kappa or better against the standard and between-appraiser agreement at 0.841. The lesson is not subtle: if you skip MSA in an attribute project, you can spend weeks optimizing noise.

The baseline Pareto was instructive. Scratch and abrasion made up 69.5% of all label damage, and tear and cut added another 24.2%. Defects were also concentrated heavily in four positions: nameplate, rating label, serial/barcode, and UL/agency marks. That helped the team avoid the common mistake of “analyzing all defects equally” when the physical evidence was already pointing to a smaller set of dominant failure zones.

Analyze Phase: The Breakthrough Came from Separating Suspected Causes from Validated Causes

The report’s strongest technical lesson is that brainstorming was only the starting point. The fishbone workshop generated many plausible causes across people, machine, method, material, measurement, and environment. But the team did not confuse plausibility with proof. It pushed through a cause-and-effect matrix, 5-Why analysis, PFMEA, multi-vari analysis, and targeted hypothesis tests until it could distinguish high-confidence causes from convenient opinions.

One of the most elegant moments in the report is the bypass trial. Instead of starting with a complicated model, the team temporarily bypassed the conveyor using padded-cart transfer and immediately isolated the conveyor rail contribution. That is a useful lesson for real-world engineers: not every breakthrough requires statistical heavy lift. A good simple experiment can answer the most important question faster than a sophisticated dashboard.

The two 5-Why chains in the report are especially important because they surface management-system causes, not just mechanical symptoms. The conveyor issue traced back to the absence of a handling-risk assessment in NPI. The substrate issue traced back to a gap in the material-spec lifecycle process that failed to trigger re-qualification when a substrate was reused on a new platform. That is exactly what mature lessons-learned work should do: convert a local defect into a systemic insight.

Improve Phase: The Team Fixed the Process, Not Just the Symptom

The report shows a good Improve phase because the team did not rely on one countermeasure and hope for the best. It matched each confirmed cause with a specific intervention and then validated those interventions before full release.

• Conveyor rails were modified with UHMW polyethylene low-friction strips and centering guidance to reduce abrasive contact and lateral drift.
• The label substrate was upgraded to 3M 7876 over-laminated polyester after a 2³ full-factorial DOE confirmed laminate as the dominant factor in cycles-to-failure performance.
• FT-1 and FT-2 clamp pads were replaced with softer closed-cell silicone foam, and soft-touch overlays were added at key fixture contact points.
• Standard work was rewritten for handling, including grip-zone graphics, no-stack supermarket practice for vulnerable surfaces, and explicit load/unload rules at functional test.

A particularly good lesson from the project is the disciplined use of DOE before capital or recurring-material commitment. The sponsor required proof before approving the substrate change. The team responded with a compact 2³ experiment that cost very little, took four days, and de-risked the material decision. That is the type of rigor that makes future spending easier to justify.

Pilot Validation Was Decisive

The pilot ran for 8 production days and covered 1,142 units after the main solutions were installed. This is where the story becomes hard to ignore. The defect rate dropped to 0.96%, and a two-proportion z-test returned a z-value of 14.12 with p < 0.00001. That is not just a visible improvement. It is statistically overwhelming evidence that the process changed in a meaningful way.

Control Phase: Sustaining the Win Was Treated as Part of the Project, Not a Postscript

Another important lesson is that the project did not end when the pilot looked good. The team built the control plan, updated standard work, re-scored the PFMEA, defined out-of-control actions, transferred ownership to the process owner, and then validated the economics with Finance.

The 90-day post-closure audit is one of the strongest features in the report because it tests whether the improvement was real or just pilot theater. The audit confirmed that the gains held. That matters because many DMAIC stories look strong at tollgate closeout and quietly decay once the project team moves on.

What Went Well

• Early and sustained sponsor engagement kept the project evidence-driven rather than approval-driven.
• Running MSA before baseline collection prevented the team from building analysis on questionable inspection judgments.
• The simple padded-cart bypass trial isolated conveyor contribution quickly without unnecessary analytical overhead.
• Including a line operator on the core team from day one surfaced fixture variation before the data fully proved it.
• Running DOE before substrate commitment made the eventual material change technically defensible and financially easier to approve.

What the Team Would Do Differently

• Bring the label supplier into the analysis earlier, ideally during the cause-and-effect matrix stage rather than waiting until Improve.
• Allow more schedule buffer for ECN approval cycles. Two of the four ECNs took longer than expected and nearly created a project-timing problem.
• Include the Training function formally as a stakeholder instead of treating rollout training as a downstream coordination task.
• Capture defect-location body-map evidence earlier as image data rather than relying on post-hoc mapping after the baseline begins.

This section is valuable because it shows maturity. Good case studies do not only celebrate wins. They also identify where execution was tighter than it looked and where the next team can shorten the path.

Systemic Gaps the Project Exposed

The strongest lessons-learned reports do not stop at local corrective action. They ask what business process or management-system weakness allowed the same risk to exist in the first place. Project SENTINEL surfaced two important systemic gaps:

1. No formal handling-risk assessment was required in the NPI process when an existing product or design was deployed onto an existing line.
2. The material-spec lifecycle process, QP-014, did not force abrasion or environmental re-qualification when a label substrate was reused on a new platform.

Both were escalated outside the project itself. That is the right move. A project should not claim to have solved the business if it only fixes one local symptom and leaves the policy gaps untouched.

Replication Plan and Why It Matters

The replication section is another reason this report reads like a strong lessons-learned document rather than a simple project closeout. The team did not treat the improvement as an isolated win. It identified clear extension opportunities across Plant 2 and the broader product portfolio.

• Extend UHMW conveyor treatment to the MC-3200 line, which shares a similar architecture.
• Upgrade the MC-5600 family to the same 3M 7876 label substrate and capture scale benefits.
• Close the dead-plate gap issue that remained in PFMEA as a carried-over item.
• Expand supermarket capacity to permanently eliminate stacking pressure.
• Roll out grip-zone marking standards to all Plant 2 products.
• Formalize the NPI handling-risk gate and QP-014 substrate re-qualification trigger as QMS changes.

This is the real difference between local improvement and operational excellence. A local fix improves one line. A replication plan converts hard-won learning into institutional gain.

Main Lessons for Quality and Operations Teams

Why This Report Is Worth Reading in Full

This case study is useful because it shows what a complete DMAIC story looks like when the team does not skip the hard parts. It contains sponsor discipline, CTQ translation, measurement-system correction, Pareto logic, 5-Why escalation, PFMEA use, multi-vari thinking, statistical confirmation, DOE-based decision making, pilot validation, control-plan transfer, financial validation, and replication planning.

That makes it more than a success story. It is a useful reference for engineers, Black Belts, quality managers, and operations leaders who want an example of how to move from an expensive chronic defect to a sustained and transferable process improvement result.