Remote kirana underwriting through real-time computer vision — analyzing shelf density, SKU diversity, and inventory velocity from a single smartphone video.
No field visits · No documents · 60-second decisions
The cash flow is real. The data just doesn't exist on paper — yet.
Cash-based transactions and handwritten registers leave zero digital footprint for traditional banks to analyze.
UNVERIFIABLE DATAEach manual audit costs ₹800–₹1,200. At 13 million stores, that's ₹10,400 Crore just to assess everyone once.
NEGATIVE ROI UNIT ECONOMICSSubjective officer judgment causes high NPAs and systemic bias against small-town retailers who deserve credit.
SYSTEMIC RISKA 60-second video walkthrough of shop shelves. Any Android or iPhone. No special hardware.
Edge-AI identifies SKU diversity, shelf density, inventory depth, refill velocity, and brand concentration in real time.
Instant creditworthiness report with confidence bands — backed by visual evidence, not a field officer's gut.
Four visual signals extracted from every frame.
Volume utilization across main racking — proxy for working capital deployed.
Unique product identifiers detected — signals footfall capture across customer types.
Stock-at-hand estimate using average market prices — fast-moving vs slow-moving mix.
Void-pattern detection predicts replenishment rate — partially empty = recent demand.
Population within 500m radius, residential vs commercial mix — demand baseline signal.
Road type, proximity to schools, offices, transport hubs → traffic multiplier for daily sales estimate.
Nearby store count — moderate competition proves demand; excess compresses margins.
Every manipulation has a counter-signal. Every signal has a cross-validator.
AI recognizes borrowed goods — multi-image temporal consistency check catches inventory that wasn't there yesterday.
Cross-referencing GPS metadata with architectural landmarks. High inventory in low-footfall geo = automatic flag.
Visual area estimation vs GPS footprint discrepancy. Differentiates store-owned inventory from distributor transit stock.
Fraud resilience is a first-class feature, not an afterthought.
Every signal has a counter-signal. Every output is auditable.
98.7% reduction in operational overhead.
Break-even at 1 prevented bad loan per 500 assessments.
Three-layer architecture. Production-grade from day one.
KiranaLens turns a smartphone into a branch.
Every store in India becomes a bankable asset — in 60 seconds, from anywhere.
Built for Poonawalla Fincorp · TenzorX 2026 · Problem Statement 4C