Manufacturing Loss Data Visualizer & Analyzer
Drop in an Excel file, get a dashboard you can drill into
INFORMATION
SKILLS: Tauri v2, Rust, calamine, odbc-api, SQL Server (ODBC), Vanilla JavaScript, ApexCharts, Python (openpyxl), OOXML (ZIP/XML)
Link: Private
CREATED: 2026-05-25
PROJECT DESCRIPTION
A desktop app that turns monthly manufacturing-loss Excel files into a self-contained HTML dashboard with drill-down and cross-filtering, just by dropping them in. Rust aggregates several files at once, and the UI switches between Japanese and Thai. It reads the inspection site from Excel cell formatting to filter every chart, and a companion app joins the data with a production database to analyse dyeing defects statistically.
Highlights
Background & Problem
Replacing 'open the monthly Excel and tally by eye' with a single drop
Read more →Key Features
Drill from machine type to outsourced products to period to inspection site with clicks
Read more →Technical Approach & Architecture
No server: read locally, hand over a single HTML file
Read more →Engineering Highlight 1: Reading the inspection site from cell formatting
Getting past a library that only reads values by reading the xlsx directly
Read more →Engineering Highlight 2: No silent failures
One fix for dependency bloat, one for failures that could go unnoticed
Read more →Engineering Highlight 3: Data design for size and speed
Doubling the number of buckets barely grows the row data that dominates the size
Read more →Executive-level Analysis
Summary, control chart, machine Pareto and treemap in a collapsible section
Read more →Bilingual Support
Japanese and Thai switch together across the screen and chart legends
Read more →Companion App: Dyeing defect analysis joined with the production database
Finding out which dyeing machine, which the loss Excel alone cannot tell
Read more →Cross-filtering toward the cause
Stack machine, defect, colour and dyeing count to see a machine-by-defect skew
Read more →Statistics to separate noise from anomalies
Control chart, residual heatmap and relative risk distinguish chance from real signals
Read more →Machine-by-defect skew at a glance
Dark cells mark machines that are prone to a given defect
Read more →Is re-dyeing really more defect-prone?
Comparing re-dyeing, multi-bath and multi-pass risk against a first-pass baseline of 1.0
Read more →Outcome, Current State & What Is Next
Improved incrementally in operation; open issues are tracked as ongoing, not hidden
Read more →