KPI Dashboard (Streamlit BI App)

Interactive Dashboard Demo

This React component embeds real data from the ETL pipeline. Toggle regions, show/hide targets, and see anomaly detection in action — all rendered directly in your browser.

Live Streamlit Version

One-click deploy to Streamlit Cloud for the full interactive experience with CSV export.

The Business Problem

A regional retail company needed a dashboard a non-technical Operations Manager could open every Monday morning to check how each region performed — and get flagged automatically if something was off.

Zero Excel. Zero manual refresh. One command to run the whole thing.

Key Stats (Real Data)

ETL Pipeline Architecture

Raw CSVs (sales + targets)
        ↓
  ETL Pipeline (scripts/etl.py)
  ├── Date parsing & null cleanup
  ├── Weekly aggregation by region
  ├── Merge with targets → vs_target %
  └── Anomaly detection (>2σ flagged)
        ↓
  weekly_summary.csv
        ↓
  Dashboard (Streamlit + React)
  ├── 4 KPI tiles (Revenue, vs Target%, Units, Return Rate)
  ├── Line chart with optional target line + anomaly dots
  ├── Stacked bar chart by region
  ├── Conditional-formatted data table
  └── CSV download button

Dashboard Features

• Region multiselect filter — isolate any region combination (shown above)
• Show targets toggle — overlay target line on the revenue chart
• Anomaly highlighting — weeks with >2σ revenue spike shown as red dots
• Conditional formatting — table rows colored by performance vs target
• CSV export — download filtered data with one click

The South Region Story

From July 2024 onward, South region consistently missed revenue targets by ~14–16%. The dashboard surfaces this automatically — no need to manually look at numbers. A manager viewing this on a Monday morning would immediately know to investigate South region operations.

The demo above shows this pattern clearly: toggle "South" on/off and see how it drags down overall performance in Q3-Q4.

What Hiring Managers See

Technical Depth

• ETL Pipeline: Production-ready Python with logging, error handling, data validation
• Statistical Analysis: Z-score anomaly detection (>2σ threshold)
• Data Modeling: Clean separation of raw → processed → presentation layers
• React + Recharts: Custom interactive components (the dashboard above)

Product Thinking

• Stakeholder-first design: Non-technical users can filter and explore
• Anomaly detection: Automated flagging of unusual patterns
• Performance storytelling: South region narrative emerges from data

Resume Bullet

KPI Dashboard — Streamlit BI App (Python, Pandas, Streamlit, React, Recharts) Built a full ETL + dashboard pipeline for weekly retail KPI monitoring across 4 regions. ETL cleans raw sales data, merges with targets, and flags statistical anomalies (>2σ). Created both a Streamlit dashboard for stakeholders and an embedded React component for portfolio demonstration. Includes region filtering, target comparison, anomaly highlighting, and CSV export.