Advanced Control & Project Closure

Beyond Basic Control Charts

In the Yellow Belt course, you learned the fundamentals of Statistical Process Control (SPC): plotting data on control charts, calculating upper and lower control limits (UCL/LCL) at ±3σ from the mean, and identifying out-of-control points. At the Green Belt level, we go deeper — applying advanced detection rules that catch subtle process shifts before they become major problems, conducting capability studies to quantify how well a process meets specifications, and building robust control systems that sustain improvements long after the project team has disbanded.

The Control phase is where improvement projects either deliver lasting value or fade away. Research consistently shows that without rigorous control mechanisms, 60-70% of process improvements degrade within 12-18 months. Your Green Belt project will be judged not by the improvement achieved during the pilot, but by whether that improvement is sustained six months and twelve months after project closure.

Advanced SPC Rules: Western Electric and Nelson Rules

The basic control chart rule — "investigate any point beyond ±3σ" — catches large, sudden shifts. But processes can deteriorate in subtler ways: a gradual drift in the mean, increased variability, or systematic patterns that indicate an assignable cause is at work. The Western Electric Rules and Nelson Rules provide additional detection criteria.

Western Electric Rules

Originally developed at Western Electric (Bell Labs), these four rules detect non-random patterns in control chart data:

Rule 1 — Beyond 3σ: Any single point beyond the upper or lower control limit (±3σ). This is the classic out-of-control signal. Probability of occurring by chance: 0.27%.

Rule 2 — 2 of 3 beyond 2σ: Two out of three consecutive points are beyond 2σ from the mean, on the same side. This suggests the process mean has shifted. While a single point beyond 2σ is not unusual (about 5% probability), two out of three is very unlikely (about 0.15%).

Rule 3 — 4 of 5 beyond 1σ: Four out of five consecutive points are beyond 1σ from the mean, on the same side. This detects a moderate but sustained shift.

Rule 4 — 8 consecutive on one side: Eight consecutive points on the same side of the center line (mean). Even without exceeding any sigma boundary, a run of 8 on one side has a probability of less than 0.4% under normal variation.

Nelson Rules

The Nelson Rules expand on the Western Electric Rules with eight detection criteria. The additional four rules include:

Rule 5 — Trend: Six consecutive points steadily increasing or steadily decreasing. This indicates a systematic drift (e.g., a gradual process degradation, increasing workload, or deteriorating system performance).

Rule 6 — Oscillation: Fourteen consecutive points alternating up and down. This suggests two populations are being mixed or a systematic oscillation in the process (e.g., two shifts with different performance levels producing alternating data points).

Rule 7 — Hugging the center: Fifteen consecutive points within ±1σ of the mean. Counterintuitively, this can indicate a problem — it may mean the subgroup size is too large (averaging out real variation) or the data is being manipulated to look "in control."

Rule 8 — Avoiding the center: Eight consecutive points beyond ±1σ (on either side). This suggests two populations with different means are being mixed.

Applying Advanced SPC Rules in Banking

Not all rules are equally relevant in every context. Select the rules that match the types of shifts you want to detect:

• Monitoring daily STP rates in payments processing: Use Rules 1, 2, 4, and 5. These detect sudden drops (Rule 1), moderate sustained shifts (Rules 2, 4), and gradual degradation trends (Rule 5).
• Monitoring weekly error rates in trade confirmations: Use Rules 1, 4, and 5. Weekly data has fewer points, so rules requiring many consecutive points (Rules 6, 7, 8) may take weeks to trigger.
• Monitoring AML alert disposition times: Use Rules 1, 2, 3, and 6. Rule 6 (oscillation) is useful here because alternating patterns may indicate inconsistency between shifts or analyst groups.

Banking Example: SPC Monitoring for Automated Payments Processing

A bank has implemented an improved payments routing algorithm that increased STP rates from 78% to 91%. The Green Belt needs to establish a control system that detects any degradation before it reaches the customer or the regulator.

Control chart setup:

• Metric: Daily STP rate (%)
• Center line: 91.0% (post-improvement mean, calculated from 25 days of stable post-implementation data)
• UCL: 94.2% (mean + 3σ, where σ_daily = 1.07%)
• LCL: 87.8% (mean - 3σ)
• Monitoring frequency: Daily
• Responsible: Payments Operations Manager

Week 3 observations:

• Day 15: 90.8%
• Day 16: 89.9%
• Day 17: 89.2%
• Day 18: 88.5%
• Day 19: 88.1%
• Day 20: 87.6%

No individual point has breached the LCL (87.8%). Under basic SPC (Rule 1 only), no action would be triggered. But applying the advanced rules:

• Rule 4 triggered: Days 15-20 are 6 consecutive points below the center line (need 8 for Western Electric, but this is concerning)
• Rule 5 triggered: Days 15-20 show 6 consecutive decreasing points — a clear downward trend

The operations manager investigates and discovers that a software patch deployed on Day 14 introduced a subtle formatting change in beneficiary reference fields that causes a subset of domestic payments to fail STP validation. Without the advanced rules, this issue might not have been detected until a point breached the LCL (potentially Day 22 or 23), by which time hundreds of additional payments would have required manual intervention.

Action taken: The software patch is rolled back, a fix is developed and tested, and the STP rate returns to 91%+ within 2 days. The advanced SPC rules saved approximately 3 days of degraded performance.

Process Capability Studies: Cp, Cpk, Pp, Ppk

Capability indices quantify how well a process performs relative to its specification limits — the boundaries of acceptable performance defined by the customer, the regulator, or the organization.

Specification Limits vs. Control Limits

This distinction is critical and frequently confused:

• Specification limits (USL, LSL) are set by the customer or regulator. They define what is acceptable. Example: "Payments must be processed within 4 hours" (USL = 4 hours).
• Control limits (UCL, LCL) are calculated from the process data. They describe what the process actually does. Example: UCL = 3.8 hours, LCL = 0.5 hours.

A process can be:

• In control but not capable: All points are within control limits, but the process variation is too wide for the specification limits. The process is stable but consistently produces some output outside specs.
• Capable but out of control: The process variation fits within spec limits, but special causes are present. The process might be meeting specs today but is unpredictable.
• In control and capable: The ideal state — stable and consistently meeting specifications.

Cp (Process Capability — Potential)

Cp measures whether the process spread fits within the specification width, assuming the process is centered:

Cp = (USL - LSL) / (6σ)

Where 6σ represents the process spread (±3σ from the mean, covering 99.73% of output).

• Cp < 1.0 — The process spread is wider than the specification range. Even if perfectly centered, the process will produce defects.
• Cp = 1.0 — The process spread exactly equals the specification range. No margin for error.
• Cp = 1.33 — A common minimum target. The process spread is 75% of the specification range, providing some margin.
• Cp ≥ 2.0 — Excellent. The process spread is at most 50% of the specification range — very robust.

Limitation: Cp only measures spread, not centering. A process with Cp = 2.0 that is severely off-center can still produce many defects.

Cpk (Process Capability — Actual)

Cpk accounts for both spread and centering by measuring the distance from the process mean to the nearest specification limit:

Cpk = min[(USL - μ) / (3σ), (μ - LSL) / (3σ)]

• Cpk < 1.0 — The process is incapable. Some output falls outside specifications.
• Cpk = 1.0 — Marginal capability. The process mean is exactly 3σ from one spec limit.
• Cpk ≥ 1.33 — Minimum target for established processes.
• Cpk ≥ 1.67 — Target for critical processes (regulatory submissions, high-value transactions).

Key relationship: Cpk ≤ Cp, always. They are equal only when the process is perfectly centered between spec limits. The gap between Cp and Cpk tells you how much improvement is possible simply by re-centering the process.

Pp and Ppk (Process Performance)

While Cp and Cpk use within-subgroup variation (short-term capability under controlled conditions), Pp and Ppk use overall variation (long-term performance including all sources of variation — shifts, drifts, batch-to-batch differences, operator changes).

Pp = (USL - LSL) / (6s_overall)

Ppk = min[(USL - x̄) / (3s_overall), (x̄ - LSL) / (3s_overall)]

The difference matters in banking:

• Cp/Cpk tells you what the process can do under the best conditions
• Pp/Ppk tells you what the process actually does over extended periods

If Cp is high but Pp is significantly lower, the process has significant between-subgroup variation (e.g., performance differs across shifts, days of the week, or analysts). This gap is a clear target for improvement — eliminate the sources of between-subgroup variation.

Banking Capability Study Example

A bank's regulatory reporting team must submit daily transaction reports to the regulator by 07:00 AM. Reports submitted after 07:00 are considered "late" and, if it becomes a pattern, can trigger supervisory attention.

Specification: LSL = not applicable (reports cannot be submitted too early in a meaningful way); USL = 07:00 AM (420 minutes from midnight).

The team measures completion time (in minutes from midnight) for 60 consecutive business days:

• Mean completion time (μ): 385 minutes (06:25 AM)
• Within-subgroup standard deviation (σ_within): 8.5 minutes
• Overall standard deviation (σ_overall): 12.3 minutes

Capability calculations (using one-sided specification — USL only):

• Cpu = (USL - μ) / (3σ_within) = (420 - 385) / (3 × 8.5) = 35 / 25.5 = 1.37
• Ppu = (USL - μ) / (3σ_overall) = (420 - 385) / (3 × 12.3) = 35 / 36.9 = 0.95

Interpretation:

• Cpu = 1.37: Under controlled conditions (short-term), the process has adequate capability — it completes reports with a comfortable margin before the deadline.
• Ppu = 0.95: Over the long term, the process performance is marginal (< 1.0). Some reports are at risk of being late. The gap between Cpu and Ppu (1.37 vs. 0.95) indicates significant between-day variation — some days the process runs much later than others.

Investigation: The Green Belt analyzes the between-day variation and discovers:

• On month-end dates, the report takes 15-20 minutes longer due to higher transaction volumes
• When a particular senior analyst is absent, the backup analyst is 10-12 minutes slower
• System performance degrades noticeably on Mondays (batch processing backlog from weekend)

These findings drive specific improvements: pre-staging month-end data, cross-training the backup analyst, and working with IT to optimize Monday batch scheduling.

Control Plan Maturity

A basic control plan (covered in the Yellow Belt course) lists what to monitor, how often, and what action to take. At the Green Belt level, control plans should mature through three stages:

Stage 1: Project Team Control (Months 1-3)

The Green Belt and project team actively monitor the process, frequently review control charts, and intervene quickly when issues arise. This is the "intensive care" period where the improved process is most vulnerable to regression.

• Monitoring frequency: Daily or more
• Review cadence: Weekly team review of control charts and metrics
• Escalation: Green Belt personally investigates any out-of-control signals
• Documentation: Continue refining SOPs based on real-world experience

Stage 2: Process Owner Control (Months 3-6)

Control transitions from the project team to the process owner and day-to-day management. The Green Belt moves to an advisory role, reviewing performance monthly.

• Monitoring frequency: Daily (automated where possible)
• Review cadence: Monthly review by process owner, with Green Belt present
• Escalation: Process owner investigates out-of-control signals; Green Belt consulted for complex issues
• Documentation: SOPs finalized, training materials complete, new staff onboarded using improved process

Stage 3: Business-as-Usual Control (Month 6+)

The improved process is fully embedded in normal operations. Monitoring continues through standard management reporting and automated alerts.

• Monitoring frequency: Daily or weekly (automated dashboards)
• Review cadence: Monthly operational review, quarterly with senior management
• Escalation: Standard operational escalation procedures
• Documentation: Fully integrated into the team's operational framework

Control Plan Template

Project Closure and Benefits Validation

Project closure is not an administrative formality — it is the final demonstration of your Green Belt capabilities. A rigorous closure process validates that improvements are real, financial benefits are realized, and the organization learns from the experience.

Benefits Validation with Finance

Financial benefits must be validated by the finance business partner using the methodology agreed during the Define phase. Common validation approaches:

Hard savings (direct cost reduction):

• FTE reduction: Validated by headcount reports — were positions actually eliminated or redeployed?
• Vendor cost reduction: Validated by vendor invoices — did spending actually decrease?
• Overtime elimination: Validated by payroll data — did overtime costs decline?

Soft savings (productivity improvement, capacity creation):

• Throughput increase: Validated by production data — are more items being processed with the same resources?
• Cycle time reduction: Validated by process data — are items completing faster?
• Capacity creation: Validated by demonstrating that additional work was absorbed without additional resources

Cost avoidance (preventing future costs):

• Regulatory penalty avoidance: Validated by meeting the regulatory deadline or passing the audit
• Loss prevention: Validated by demonstrating reduced error rates that previously led to financial losses

Important: Benefits should be measured against the baseline established in the Measure phase, not against the worst month or a hypothetical scenario. Use conservative assumptions and let finance sign off on the final number.

Benefits Realization Timeline

Not all benefits are realized immediately at project closure:

Report benefits at closure as "annualized projected benefits based on X months of post-implementation data" and commit to a benefits review at 6 and 12 months.

Lessons Learned Documentation

Every Green Belt project generates insights that should benefit future projects. Document:

What went well:

• Which tools and techniques were most effective?
• Which stakeholder engagement strategies worked?
• What data sources were most valuable?

What was challenging:

• Where did you encounter resistance, and how did you overcome it?
• What data quality issues arose?
• Which assumptions proved wrong?

What you would do differently:

• Scope changes — was the scope right, or did it need adjustment?
• Timeline — was the project plan realistic?
• Team composition — did you have the right people?

Recommendations for the organization:

• Process improvements that should be replicated in other areas
• Systemic issues (data quality, technology limitations, training gaps) that require broader organizational action
• Suggestions for future Green Belt projects in adjacent processes

Green Belt Project Portfolio Management

As you complete your first project and build confidence, you will begin to manage a portfolio of improvement initiatives — some active, some in the pipeline, some completed and in sustain mode. Portfolio management at the Green Belt level involves:

Balancing Active Projects

Most Green Belts can manage 1-2 active DMAIC projects simultaneously (remember, you are doing this part-time alongside your regular role). Attempting to lead 3+ projects typically results in all of them suffering. Be disciplined about finishing one project before starting another.

Pipeline Development

Continuously identify potential improvement projects within your area. Maintain a pipeline of 3-5 candidate projects, each with a preliminary assessment of strategic alignment, financial opportunity, and feasibility. When your current project closes, you can quickly move to the next highest-priority initiative.

Sustain Portfolio

For completed projects, maintain a "sustain portfolio" that tracks:

• Is the control plan being followed?
• Are metrics still within specification?
• Have benefits been realized as projected?
• Are there signs of regression that require intervention?

Review your sustain portfolio monthly. It takes less than 30 minutes per project and prevents the demoralizing experience of watching hard-won improvements fade away.

Reporting to Leadership

Develop a concise portfolio dashboard for your sponsor and leadership that shows:

• Active projects: phase, timeline, key risks
• Pipeline: next projects with estimated start dates
• Completed projects: sustained performance, validated benefits
• Total value delivered: cumulative financial impact of your Green Belt portfolio

This dashboard demonstrates the ROI of the Green Belt program and builds the case for continued investment in process improvement.

Key Takeaways

• Advanced SPC rules (Western Electric, Nelson) detect subtle process shifts before they cause major problems
• Capability indices (Cp, Cpk) measure whether a process can consistently meet specifications; Cp measures potential, Cpk measures actual performance including centering
• Pp/Ppk use overall (long-term) variation while Cp/Cpk use within-subgroup (short-term) variation — the gap reveals between-subgroup variation that is a target for improvement
• Control plan maturity progresses from project team control to process owner control to business-as-usual
• Benefits validation with finance must use the baseline from the Measure phase and conservative assumptions
• Lessons learned documentation ensures organizational learning from every project
• Green Belt portfolio management balances active projects, pipeline development, and sustain monitoring
• The true measure of a Green Belt project is sustained improvement 6-12 months after closure, not the results achieved during the pilot

You have now completed all seven modules of the Six Sigma Green Belt course. In the final exam, you will be tested on project selection, advanced define, measurement system analysis, hypothesis testing, regression analysis, design of experiments, and advanced control — all applied to banking operations scenarios. Good luck.