Manufacturing Analytics Platform

Engineered an end-to-end ETL pipeline and interactive BI suite that transformed disparate data sources into strategic insights for production optimization.

Python Pandas & NumPy Plotly ETL Pipeline Data Engineering Business Intelligence

Project Overview

A plastics manufacturer was performing daily material substitutions to maintain production flow, but had no systematic way to measure financial impact or identify operational patterns. Leadership needed to move from assumptions to data-driven decision-making.

I built a complete analytics solution from raw Excel files to interactive dashboards, revealing critical cost drivers and enabling strategic optimization.

The Challenge

  • Multiple disparate Excel files with complex, inconsistent formatting
  • Merged-cell headers breaking standard parsers
  • Thousands of non-standardized text entries
  • No system to track financial impact of substitutions
  • Critical business questions unanswered

The Solution: Four-Phase Engineering Approach

1

Data Engineering

Built custom header parser and alignment engine to solve critical extraction bugs

2

Normalization

Developed regex-based cleaning system with validation checkpoints

3

Analysis & Strategy

Created categorization logic and aggregated insights across dimensions

4

BI & Deployment

Built interactive dashboards with drill-down capabilities

Technical Deep Dive

Key engineering challenges and solutions

Header Parsing

Problem: Multi-level merged-cell headers caused standard parsers to misalign data, creating silent corruption.

Solution: Engineered a custom parser that dynamically detects header rows, reads three independent tables, and programmatically enforces perfect alignment. This single innovation ensured data integrity across the entire pipeline.

Surgical Normalization

Challenge: Thousands of text entries with inconsistent whitespace, casing, and punctuation.

Solution: Built a regex-based cleaning engine that standardizes entries without altering business terminology. Included mid-pipeline validation checkpoints exporting cleaned values to .txt files for manual verification before proceeding.

Strategic Categorization

Innovation: Transformed raw blend changes into meaningful business strategies using logical hierarchies.

Examples: "Removed All Virgin," "Recycled to Virgin Dominant," "Pure Virgin Swap." This enabled leadership to analyze profitability by strategy type, not just by material codes.

Interactive Business Intelligence Suite

Developed seven specialized dashboards using Plotly with multi-level drill-down capabilities:

  • Executive Summary: High-level financial overview and KPIs
  • Machine 360°: Complete performance profile by machine with month-over-month analysis
  • Part Number Profitability Analyzer: Identify which parts drive costs vs. savings
  • Strategy Performance: Compare financial outcomes across substitution strategies
  • Blend Analysis: Material composition impact on costs

Business Impact

Key Findings

  • Quantified financial impact, proving substitution process was a net positive
  • Achieved 80% "good" substitution rate across operations
  • Identified optimal strategy: "Remove All Virgin" material for maximum savings
  • Revealed 22.5% visibility gap due to missing cost data

Surgical Optimization

80/20 Analysis Revealed: Over 92% of machine-related cost increases came from just 3 machines.

This finding enabled the client to focus engineering resources on specific high-impact areas rather than broad, unfocused process improvements.

Result: Transformed reactive operations into strategic, data-driven optimization with clear ROI priorities.

Tools & Technologies

Core Technologies

Python, Pandas, NumPy, Plotly

Data Engineering

ETL Pipelines, Regex, Openpyxl

Deployment

GitHub Pages, Git, LaTeX

Development

Jupyter Notebooks, VS Code