How I Built a QR-Powered ERP for Garment Factories Using React, Firebase, and a Raspberry Pi 🧵⚡

The Problem: The "Spreadsheet Chaos" of the Sewing Floor
Most people think garment manufacturing is just "sewing." In reality, it’s a logistics nightmare. A single factory floor might have 60+ machines, 100+ operators, and thousands of fabric bundles moving at once.

Before Scan ERP, tracking was done on paper or messy Excel sheets. This led to:

Payment Disputes: "I sewed 50 sleeves, why am I paid for 40?"

The Black Hole: A bundle goes missing, and production stops for hours.

Delayed Decisions: Managers only see "yesterday's" data, never "now."

The Solution: A "Zero-Install" System of Action
I didn’t want to build a "System of Record" where people just type in data. I wanted a System of Action.

We built Scan ERP as a Progressive Web App (PWA) because factory environments hate App Stores. No downloads, no updates—just a QR scan and you're in.

1. The Multi-Engine QR Scanner (99% Accuracy) Scanning on cheap, greasy $100 Android phones is hard. To hit 99% accuracy, we didn't rely on one library. We built a hybrid engine combining:

BarcodeDetector API (Native & Fast)

jsQR (Fallback)
ZXing (Robustness)

This ensures that even a smudged QR code on a moving bundle gets registered in <200ms.

1. Architecture: From Monolith to Modular Initially, our firebase-services.js was a 3,000-line beast. We recently performed a major "Architectural Cleanup":

Modular Services: Split by domain (Inventory, Piece-rate, Quality Control).

Base Service Pattern: A standardized FirebaseBaseService for CRUD, automated timestamps (createdAt, updatedAt), and batch transactions.
The Result: A 75% reduction in core service file size and way faster bundle loading. 3. Handling "Off-Grid" Factories with Raspberry PiInternet in factories can be spotty. We deployed a Raspberry Pi as an edge collector. It handles local data sync and ensures that even if the Wi-Fi drops, the sewing lines don't. Once the connection is back, Firebase Background Sync handles the heavy lifting. The Impact: Real Data, Real Fast
In our primary deployment (60+ machines in Nepal), we’ve already tracked 115,370+ pieces.

Bottlenecks Detected in <30 Seconds: If a line slows down, the dashboard turns red instantly.

Zero Payment Disputes: Every scan creates a tamper-proof entry in the operator's digital diary.

Lessons Learned
Building for the "real world" is different from building for "the browser."

UX > Beauty: Big buttons, high contrast, and voice notifications matter more than fancy animations on a dusty factory floor.

PWA is the Future of Enterprise: In 2026, the "install barrier" is the biggest killer of B2B apps. PWAs solve this.

Check out the live project here: https://scanerp.pro/

Are you building niche ERPs or experimenting with PWA scanning? Let's chat in the comments!