QMS Insights — Computerized System Health Dashboard

Deviation & CAPA Trending — Fleet I-MR Control Chart

Monthly deviation volume across the QC computerized-systems fleet · Shewhart Individuals with Nelson rules 1–4 applied · CL ± 3σ control limits derived from baseline (Dec 2024 – Jun 2025)

Signal — automatically detected A Nelson Rule 1 violation in April 2025 coincides with the IW7 → Windows 10 fleet revalidation cutover. A subsequent Nelson Rule 2 violation (9 consecutive points above CL) develops from Nov 2025 – Feb 2026, indicating a sustained upward shift rather than a sporadic event. Recommend retrospective process review of post-cutover audit-trail review cadence.

Segregation-of-Duty Exceptions

Active SoD exceptions flagged in the latest quarterly user-access review

Audit-Trail Review Compliance — Fleet × Month Heatmap

Percentage of required audit-trail reviews completed within calendar month · Drill-through to underlying review records by clicking any cell

User Access Review Aging

Days since last quarterly access review per validated system

Validated-State Inventory — QC Computerized Systems

28 systems across the QC estate · Periodic review status, owner, open changes, and CAPAs · Sortable, searchable, filterable

System ID	Description	Type	Criticality	State	Last PR	Next PR Due	Status	Open CRs	Open CAPAs	Owner

Change-Control Throughput & Cycle Time

Monthly change-record volume stacked by risk tier with overlaid mean cycle time (days from initiation to closure)

Deviation Root-Cause Mix

Distribution of confirmed root causes across closed deviations · Last 12 months

Methodology · How this dashboard would be built

This is a portfolio mockup with synthetic data. The architecture below mirrors how I would deploy this for an MSAT team in production. The data covers 28 QC computerized systems (TOC, FTIR, KF, Maldi-ToF, Zetasizers, microplate readers, KQCL, LabX-controlled balances and pH meters, plus EM/water-system/document interfaces) across an 18-month window.

Data Pipeline

Bronze layer ingests raw audit-trail exports (CSV/XML) from each computerized system, change-control records from the QMS, and user-access logs from Active Directory. Python utilities (pandas, lxml) normalize the dozen audit-trail formats into a unified schema. Silver layer applies ALCOA+ rule packs and tags every record with system, owner, criticality, and risk tier context from the validated-system master. Gold layer materializes the analytics tables this dashboard reads against.

I-MR Control Chart

Center line is the mean of the baseline period (Dec 2024 – Jun 2025, before the IW7 → Windows 10 cutover). UCL/LCL are CL ± 3σ where σ is estimated from the average moving range divided by d₂ (1.128 for n=2). Color-coded points show Nelson Rule 1 (single point beyond 3σ) and Nelson Rule 2 (9 consecutive points on the same side of CL) violations.

Heatmap

Audit-trail review compliance = (# reviews completed within calendar month) / (# reviews required). Cell color thresholds: ≥98% green, 90–97% amber, <90% sienna alert. Hover surfaces system metadata and underlying review counts.

DAX Measures

Each KPI is a versioned DAX measure. The capability and CL/UCL calculations live in measures (not pre-calculated tables) so they recompute when filters change. Nelson rule logic is implemented as inline DAX with EARLIER patterns for the consecutive-point checks.

GAMP 5 Posture

Dashboard treated as Category 5 (custom configured application). URS, FS, DS, IQ/OQ/PQ versioned in Git. Power BI tenant audit logs capture every report edit. Source of truth: the gold layer in Databricks; the dashboard never edits data.

Synthetic Data Disclosure

All system names, system IDs, deviation counts, user identifiers, dates, and review outcomes shown above are fabricated for portfolio demonstration purposes. The data is generated by a seeded random process designed to produce a coherent, realistic narrative across the fleet. No real GSK or Alphanumeric Systems data is present anywhere in this artifact.