Keemti OS

One Key · One Truth · One Experience: the retail operating ecosystem for a 7 → 80 store scale-up.

Interactive product prototype built for La Conquista 2026: The Responsible Business Conquest (SDA Bocconi Asia Center × Keemti), Product Management · Enterprise Platforms track. Prepared by team ISBians, Indian School of Business.

🔗 Live prototype: keemtios.vercel.app · 📄 Submission deck (PDF)

---

The problem

Keemti, a lab-grown diamond jewellery brand, runs 7 stores and plans to grow to 80+. Its systems were not built for that scale:

Priority	Fault line	Today
P0	Disconnected backbone: manufacturing and retail run on separate platforms	2 siloed systems, 0% real-time sync
P0	Inventory without truth: about 1,200 SKUs per store pass through 7 unlinked stages	14% dead stock, 2.3 inventory turns a year
P1	Checkout is the bottleneck: billing takes 15–20 minutes, with no clienteling	about 5.5 staff-hours lost per store per day
P1	Leadership is flying blind: there is no unified dashboard	reporting is manual only

The solution: seven modules on one integration spine

Every action publishes one event to a shared integration spine, and every module subscribes to it. The prototype shows each module as a screen:

Module	What it shows
Executive Cockpit	Six live KPIs, weekly sales vs plan, store leaderboard, live exception feed, SKU and stone health, the factory and supply line, and the "Responsible Business" layer
Traceability & Inventory	A serial-level Stone Twin registry and chain of custody (factory → SGL certification → warehouse → store → customer). Also a lifetime buyback flow: verify the stone, value it at 87% of its last certified price, pay out by voucher or NEFT within 48 hours.
Store Operations	Store health scores, the AI replenishment engine, warehouse → store transfers, and an AI decision log where each decision is marked auto, review or override
Checkout · CX Demo	A clickable 3-step checkout simulator (scan & sell → price + certificate → confirm & invoice) with a running timer, plus clienteling suggestions drawn from purchase history
Customer Journey · App	The customer's app: the stone's journey, QR verification, instant buyback quote, concierge booking and a working DPDP consent toggle
Customer 360	Repeat rate, average order value, cross-sell, personas with triggered moments, and the NPS breakdown
Scale Command · 7→80	A network-size slider that recomputes run-rate, live stones, SKU-store cells, auto-replenishment, inventory turns and NPS from the pilot's economics

Target impact

Metric	Today (7 stores)	Target (80 stores)
Run-rate	₹35 Cr	₹480 Cr
Average checkout time	15–20 min	4:42
Dead stock	14%	7.5%
Inventory turns per year	2.3×	3.4×
NPS	57	71
Stone traceability	–	100%

Roadmap

Phase	Stores	Months	Ships
0 · Stabilize	7 → 20	1–4	Stone Twin, trace ledger, Integration Spine MVP, checkout pilot at 3 flagship stores
1 · Connect	20 → 50	5–9	Spine at every store, AI replenishment, buyback at scale, customer app
2 · Optimize	50 → 80	10–15	Predictive demand, dynamic pricing, company-wide executive MIS, DPDP audit

Platform strategy: Keep the factory ERP and connect it through an API bridge. Buy store POS and hardware. Build the differentiated core: the Stone Twin, the integration spine, the cockpit and the app.

The full reasoning is in the submission deck.

About the data

All stores, stones, customers, orders and metrics are illustrative mock data written for the case. The "live" feeds and timers are simulated in the browser. There is no backend and no real customer data.

Tech

• One self-contained index.html (~67 KB): vanilla JavaScript, inline CSS and SVG charts
• No framework, no build step and no external requests

Run locally

open index.html

Deployment

This repo is connected to the Vercel project keemtios. Every push to main deploys to production at keemtios.vercel.app.

vercel.json turns on clean URLs. It also redirects the old paths (/KeemtiOS-Prototype-LaConquista2026, /R1_ISBians_ISB.pdf) to their new locations, so links shared earlier still work.

.
├── index.html                                # the whole prototype
├── vercel.json                               # clean URLs + legacy redirects
└── docs/
    ├── KeemtiOS-LaConquista2026-Deck.pdf     # 5-page submission deck
    └── screenshots/                          # images used in this README

Disclaimer

This is a case-competition prototype. It is not an official Keemti product, and all figures are illustrative.