# Cropnest: An Experiment in Rural Trade Digitization

##Goal

To validate whether a simplified digital ledger can act as a single source of truth for farmers and traders, effectively replacing the "Kachha" (informal) paper trail that currently dominates the market.

##Context

The hypothesis is built on the observation that rural trade in India is broken in three fundamental ways:

• Information Asymmetry: Farmers rarely know the real-time price of commodities in neighboring districts.
• Credit Invisibility: Without formal transaction records, farmers cannot access institutional credit, forcing reliance on predatory lending.
• Settlement Delays: Payments are often deferred by weeks with no legally binding record of the debt.

##Current State

We have deployed an MVP in two districts in Madhya Pradesh. The stack is intentionally boring: a Django monolith with Postgres, served via a PWA. We are serving ~500 DAU, mostly traders who use it for inventory tracking.

##What Worked

1. WhatsApp Integration: Using the WhatsApp Business API for transaction receipts saw a 98% open rate compared to SMS. It serves as a trusted "digital receipt".

2. Localized UI: The interface uses Hindi transliteration rather than translation. Terms like "Udhaar" (Credit) and "Jama" (Deposit) were instantly understood, whereas formal banking terminology failed in user testing.

##What Failed

1. Real-time Sync: Our initial assumption of 4G availability was wrong. The "online-first" React approach resulted in data loss during network flakes.

2. OCR for Receipts: We attempted to use Tesseract for scanning handwritten paper slips. The variety in handwriting styles and paper quality made this feature unusable (less than 15% accuracy).

##What I'm Unsure About

The regulatory landscape for digital mandi payments is shifting. It is unclear if we can legally facilitate payments without a full banking license, or if we must remain purely a ledger system.

##What I'd Do Differently

I would start with an Offline-First architecture using SQLite on the client (via ElectricSQL or similar) rather than trying to patch a REST API for low connectivity later. The sync logic is currently the biggest source of technical debt.