⚡ Enterprise Platform Blueprint · v1.0 · June 2026

SREE Delivery — one super-app for everything hyperlocal

An enterprise-grade, Cloudflare-native blueprint to launch a Swiggy/Zomato & UrbanClap-class platform across five verticals: restaurant & hotel food, grocery & goods home delivery, in-stay hotel service, errands & courier, and on-demand home professionals — starting in Bhubaneswar & Odisha, designed to scale nationally.

5 verticals, 1 platform 4 stakeholder apps Cloudflare-native · Workers + DO + D1 Live at delivery.samparana.com Payments: Razorpay / UPI
⬇ Stakeholder Deck — Odia (PDF) ⬇ PowerPoint (.pptx)
Executive Summary

What we are building, and why now

SREE already operates a multi-sector business with live hyperlocal infrastructure (Saptahik Bazar, vendor network, battery-vehicle delivery, an agent earning program). SREE Delivery unifies these into a single consumer super-app that monetises the same delivery fleet and merchant relationships across five demand streams instead of one — maximising orders per rider, per kilometre, per day.

🎯

The opportunity

Tier-2/3 Odisha is underserved by national aggregators. Local trust, Odia-first UX, lower commissions and an existing fleet are a defensible moat.

🧩

The strategy

One logistics backbone, five revenue verticals. Shared identity, wallet, address book, and rider pool — each new vertical adds margin without new fixed cost.

🏗️

The build

Cloudflare-native: Pages frontends, Workers APIs, Durable Objects for live order/GPS state, D1 for transactional data, Queues for dispatch. Low ops, global edge, pay-per-use.

Guiding principle — ship a thin but real slice end-to-end (one vertical, one zone, real money, real riders) before broadening. Every phase below is independently shippable and revenue-generating.
The Product

Five verticals on one logistics backbone

Each vertical reuses the same accounts, payments, dispatch and tracking core. They differ only in catalog shape, fulfilment SLA and merchant type.

Food

🍽️ Restaurant & Hotel Food

Browse nearby restaurants & hotels, dish-level menus, customisations, live prep + delivery tracking. The classic Swiggy/Zomato flow. SLA: 30–45 min.

Goods

🛒 Grocery & Goods Home Delivery

Kirana, daily essentials and Saptahik Bazar produce delivered home. Slotted & express. Reuses the existing 207-item catalog & vendor portal.

Hospitality

🏨 Hotel In-Stay Service

B2B2C: partner hotels offer room-service, amenities & local errands to checked-in guests via QR menu. Hotel-branded, commission per order.

Logistics

📦 Errands & Courier

On-demand pickup-and-drop, parcel, document & "buy-me-this" tasks. Distance-priced. Monetises idle rider capacity between food peaks.

Services

🛠️ Home Professionals (UrbanClap-style)

Plumber, electrician, AC repair, cleaning, salon-at-home, appliance service. Scheduled bookings, vetted pros, fixed/quoted pricing, ratings.

Extensible by design

Pharmacy, meat/fish express, B2B restaurant supply and ticketing slot in as new "catalog types" — no re-architecture.

Marketplace Model

A four-sided platform

Every vertical is served by the same four actors and four product surfaces. Build the surfaces once; configure per vertical.

1

Customer

Installable PWA + web. Discover → cart/booking → pay → live-track → rate. Wallet, saved addresses, SREE Card discounts, Odia/English/Hindi.

2

Merchant / Partner

Restaurants, hotels, kirana, Bazar vendors. Menu & stock management, accept/reject, prep status, payouts, hours & availability. Extends existing vendor-portal.

3

Rider / Professional

Delivery riders & service pros. Go-online, receive assignment, navigate, OTP-verified handoff, earnings & payout ledger, ratings.

4

Operations / Admin

Onboarding & KYC, live ops console (map of orders/riders), commission & pricing config, disputes & refunds, analytics, zone management.

Technical Architecture

Cloudflare-native, edge-first

Chosen to match SREE's existing stack (Pages + Workers + KV + R2, as used by samparana.com and bazar.samparana.com). No servers to manage, global low-latency edge, and pay-per-use economics that suit a hyperlocal launch.

CUSTOMERS / MERCHANTS / RIDERS / ADMIN  (installable PWAs + web)
        │
        ▼
┌──────────────────────────────────────────────────────────┐
│  Cloudflare Pages  — 4 frontends, one design system        │
│   customer · merchant · rider · admin (sub-paths/sub-apps) │
└──────────────────────────────────────────────────────────┘
        │  HTTPS / WebSocket
        ▼
┌──────────────────────────────────────────────────────────┐
│  Cloudflare Workers  — API gateway + business logic       │
│   /api/auth  /api/catalog  /api/orders  /api/dispatch      │
│   /api/payments  /api/track  /api/admin                    │
└──────────────────────────────────────────────────────────┘
   │            │            │            │            │
   ▼            ▼            ▼            ▼            ▼
Durable      D1          KV          R2          Queues
Objects   (SQL: orders, (sessions,  (menu/dish   (dispatch +
live order   merchants,   rate-limit,  images,     notify +
state +      menus,       config,      KYC docs,   settlement
rider GPS    payments,    geo-cache)   invoices)   jobs)
+ WS hub     riders…)
        │
        ▼   external integrations
Razorpay (UPI/cards/payouts) · Google Maps/Ola Maps (geo, ETA, routing)
WhatsApp Cloud API + SMS (MSG91) · Workers AI (support bot, reuse from samparana.com)

Why each Cloudflare primitive

PrimitiveRoleWhy it fits
Pages4 frontend apps + this siteAlready the SREE deploy pattern; per-app projects + subdomains.
WorkersStateless API + auth + routingEdge-fast, scales to zero, same model as existing functions/api/*.
Durable ObjectsOne object per live order; rider GPS hub; WebSocket fan-outStrongly-consistent, single-threaded state — perfect for an order state machine + real-time tracking without Redis/Socket.io.
D1 (SQLite)Transactional system-of-recordRelational orders/menus/payments with SQL; cheap; edge-replicated reads.
KVSessions, rate limits, config, geo cacheAlready used for visitor counter + rate limiting on samparana.com.
R2Images, KYC docs, invoicesZero-egress object storage; same pattern as planned VENDOR_DOCS bucket.
QueuesAsync dispatch, notifications, settlementDecouple slow work (assign rider, send WhatsApp, compute payouts) from the request.
Workers AISupport chatbot, menu taggingLlama 3.3 already wired on samparana.com — reuse.
Data Model

Core D1 schema (system-of-record)

Polymorphic merchant_type + vertical columns let one schema serve food, goods, hotel, errands and services. Live, fast-changing state lives in Durable Objects; D1 holds the durable truth.

TableKey columnsNotes
usersid, phone, name, role, wallet_balance, localeOne identity across all verticals & roles.
merchantsid, vertical, type, name, geo(lat,lng), zone_id, status, commission_pct, open_hoursRestaurant, hotel, kirana, vendor, service-firm.
catalog_itemsid, merchant_id, name(en/od), price, category, veg_flag, prep_time, stockDishes / goods / services / errand types.
ordersid, user_id, merchant_id, vertical, status, total, payment_id, rider_id, addr, slot, otpDO mirrors the live status; D1 is the ledger.
order_itemsorder_id, item_id, qty, unit_price, customisationsSnapshot of price at order time.
ridersid, user_id, vehicle, status, current_zone, kyc_status, ratingRiders + service professionals.
service_bookingsid, user_id, pro_id, service_id, scheduled_at, quote, statusUrbanClap-style scheduled jobs.
paymentsid, order_id, gateway, amount, status, settled_atRazorpay txn + reconciliation.
payoutsid, party_id, period, gross, commission, net, statusMerchant + rider settlements.
zonesid, name, polygon, surge_pct, sla_minPer-area pricing, surge, serviceability.
reviewsid, order_id, target, stars, textTwo-way ratings.
Core Engines

The five systems that make it work

① Order state machine

One Durable Object per active order owns the transitions; every change is broadcast over WebSocket and persisted to D1.

PLACEDACCEPTEDPREPARINGREADYPICKED UPON THE WAYDELIVERED

Side-paths: REJECTED, CANCELLED, REFUNDED. Service jobs use a parallel set (BOOKED → ASSIGNED → EN ROUTE → IN PROGRESS → COMPLETED).

② Dispatch & assignment

  • Geo-query riders within zone radius (KV-cached H3/geohash buckets)
  • Score by distance, current load, rating, acceptance rate
  • Offer → accept window (Queue-driven, auto-reassign on timeout)
  • Batch nearby orders for the same rider (multi-pickup)
  • Express vs. slotted routing (reuses Bazar's veg/non-veg track logic)

③ Payments & settlement

  • Razorpay Orders API — UPI, cards, netbanking, wallet, COD
  • Server recomputes total from live catalog (never trust client — same rule as Bazar orders)
  • Razorpay Route / RazorpayX for split payouts to merchants & riders
  • SREE Wallet + SREE Card discounts applied pre-gateway
  • Daily settlement Queue job → payouts ledger + invoice PDFs in R2

④ Live tracking · ⑤ Notifications

  • Rider app streams GPS → order DO → WebSocket to customer map
  • Google/Ola Maps for ETA, distance matrix & turn-by-turn
  • WhatsApp Cloud API for order updates (deep-link, like current Bazar flow)
  • SMS (MSG91) fallback + OTP; Web Push for PWA
  • Workers AI support bot for "where's my order?" deflection
Security, Compliance & Trust

Enterprise-grade from day one

🔐

AuthN/Z

Phone-OTP login, JWT sessions in KV, role-based access (customer/merchant/rider/admin), signed Razorpay webhooks.

🛡️

Platform hardening

Cloudflare WAF + Bot Fight Mode + Turnstile (already on samparana.com), per-IP rate limits, strict CSP, HSTS — extend existing _headers.

📋

Compliance

IT Act 2000 & DPDP Act 2023 privacy, GST-compliant invoicing, FSSAI display for food merchants, KYC for riders/merchants in R2.

💳

Payment safety

PCI handled by Razorpay (no card data touches us), server-side amount recompute, idempotency keys, refund audit trail.

🧯

Reliability

Edge-distributed, scales to zero, DO single-writer consistency, Queue retries with DLQ, structured logs + alerts.

🔍

Trust & safety

Two-way ratings, OTP-verified handoff, SOS button, fraud heuristics on refunds/cancellations, merchant & rider vetting.

Delivery Roadmap

Phased, each phase independently shippable

Timeline assumes a small focused team (see resourcing). Dates are indicative from a Phase 0 kickoff.

Phase 0
Weeks 1–2

Foundations

  • Provision samparana-delivery Pages project + delivery.samparana.com (this site)
  • D1 schema, KV/R2/Queues bindings, Razorpay + Maps + WhatsApp sandbox keys
  • Design system, auth (phone-OTP), one identity service
Phase 1
Weeks 3–8

MVP — Food, one zone

  • Customer PWA: discover → cart → pay → live-track → rate
  • Merchant console + Rider app + order state machine in Durable Objects
  • Manual-assist dispatch, Razorpay live, WhatsApp updates — real orders, real money
Phase 2
Weeks 9–14

Logistics maturity + Grocery

  • Automated dispatch & assignment engine, batching, surge/zone pricing
  • Fold in Grocery/Bazar vertical (reuse 207-item catalog + vendor portal)
  • Admin ops console, settlement & payout ledger, ratings
Phase 3
Weeks 15–22

Services + Hotel

  • UrbanClap-style home professionals: scheduled bookings, quotes, pro vetting
  • Hotel in-stay QR ordering (B2B2C), errands/courier vertical
  • SREE Wallet, referral & agent program tie-in, multi-language
Phase 4
Weeks 23+

Scale & optimise

  • Multi-zone & multi-city expansion, demand forecasting
  • Advanced fraud/trust, loyalty, ad/promoted listings revenue
  • Native app wrappers, analytics warehouse, A/B experimentation
Resourcing & Economics

Team, infra cost & unit economics

Lean launch team

RolePhase 1
Full-stack lead (Workers/DO/D1)1
Frontend (PWA × 4 surfaces)1–2
Ops / city launcher (merchants, riders)1
Designer / QA (shared)0.5
Product owner (SREE)1

Grows with verticals; Cloudflare's serverless model keeps DevOps near-zero.

Infra cost (pay-per-use)

ServiceIndicative
Cloudflare Workers Paid + DO + D1 + R2 + Queues~$5–$30/mo at launch scale
Razorpay~2% per txn (no fixed)
Google/Ola Mapsusage-based, free tier covers MVP
WhatsApp Cloud API + SMSper-message (paise)

Infra scales with revenue — no upfront server spend. Major cost is people + rider incentives, not technology.

Revenue streams

Merchant commission (15–25%) Delivery fee (distance + surge) Service booking margin Hotel B2B2C commission Errand/courier distance fee Promoted listings / ads Subscription (free delivery tier) SREE Card cross-sell
Success Metrics

What we measure

<45m
Avg food delivery time
>95%
Order fulfilment rate
>2.5
Orders / rider / active hr
>4.5★
Avg CSAT rating
<2%
Order cancellation rate
>40%
Repeat-customer rate (M2)
>30%
Cross-vertical adoption
+ve
Contribution margin / order
Risks & Mitigations

Known risks, deliberate answers

Rider supply & reliability

  • Leverage existing SREE fleet + agent program; incentivise early; auto-reassign on timeout

Merchant onboarding velocity

  • Reuse existing vendor network & portal; self-serve menu tools; dedicated city launcher

Real-time scale on DO

  • One DO per active order (not global); hibernation; load-test before multi-zone

Unit economics / discount burn

  • Margin-first: server-recomputed totals, surge pricing, no blanket subsidies, batch deliveries

Competition (Swiggy/Zomato/UC)

  • Win on local trust, Odia-first UX, lower commission, multi-vertical bundle they don't offer locally

Compliance (FSSAI/GST/DPDP)

  • Built into onboarding & invoicing from Phase 0; legal review before live food sales
Immediate Next Steps

From blueprint to Phase 0

  • Approve scope & phasing (this document)
  • Pick the launch zone (recommend: one Bhubaneswar pincode cluster)
  • Choose pilot vertical for MVP (recommend: Food)
  • Open Razorpay merchant + Google/Ola Maps + WhatsApp Business accounts
  • Provision samparana-delivery Pages project + DB bindings
  • Stand up auth + identity + D1 schema (Phase 0 scaffold)
  • Recruit/assign the lean team
  • Sign 10–15 pilot merchants & 8–10 riders for the launch zone

📂 Stakeholder presentation — Odia (ଓଡ଼ିଆ)

Non-technical, business-focused deck for SREE stakeholders. 12 slides. Download as PDF (opens anywhere) or editable PowerPoint.

⬇ Download PDF ⬇ PowerPoint (.pptx)

Ready to build SREE Delivery

This blueprint is approved-architecture and ready to execute. The next command starts Phase 0: provisioning the platform and scaffolding the MVP.

← Back to Samparana Discuss on WhatsApp