Production Quality & Process Optimization
End-to-end analytics consulting project for Perry’s Ice Cream to enhance production quality, reduce fill-weight variance, and improve line stability using data-driven insights.
Situation
Perry’s Ice Cream wanted to investigate production-line variability and product weight inconsistencies across its manufacturing network. Each WorkCenter produces different package types (pints, tubs, bars), and maintaining fill-weight control is essential for compliance and yield efficiency.
Task
As project lead, I designed and executed a multi-phase analytics framework to quantify sources of variance, assess temporal stability, and develop actionable insights to improve process control and quality performance across lines.
Action
Business & Strategy
Led a team-based consulting engagement with Perry’s operations and IT teams. Defined the analytical scope, validated production data with the client, and presented findings to senior management. Developed a statistical control framework to identify stable vs. unstable SKUs, benchmark line capability, and translate findings into process improvement recommendations.
Technical Implementation
Engineered a 14M+ record dataset by merging run and checkweight data using temporal joins. Conducted multi-dimensional mean–variance analysis (WorkCenter × SKU × Shift × DayOfWeek), variance decomposition via weighted ANOVA, and process capability analysis (Cp, Cpk). Applied time-series decomposition (STL), autocorrelation, and change-point detection to detect drift and instability. Built Python pipelines for EDA, SPC classification, and automated report generation with over 20+ visual diagnostics (heatmaps, control charts, ridge plots, and trend decompositions).
Results
Business Impact
Revealed that over 65% of total variability originated from SKU-level factors, with specific lines (e.g., WC 1201) exhibiting recurrent mean drift and higher instability. Provided Perry’s with a ranked stability report and visual control benchmarks that now guide continuous-improvement discussions and line calibration priorities.
Technical Achievement
Delivered 30+ analytical exports, 20+ publication-quality visualizations, and an integrated temporal stability scoring model. Automated data aggregation, reducing manual QA time by 80%. Established a replicable analytical pipeline for future factory performance monitoring.