Home/RetailVista

RetailVista.

Reliance Retail's enterprise geospatial intelligence layer. Predicts what a catchment can become — never just reflects its past. Three tracks running in parallel. UCP, Customer Listening, JioGIS, store/online data, government and third-party sources unified onto a shared H3 hexagonal index.

Powered by the Retail Vista Agent → · 4-stage opportunity engine · 10-dimension scoring · conversational workspace · /agents/ catalog.

Live · Internal Track · Opportunity Explorer on grocery NSO leads · JioMart data Pilot · Google Track · MVP demoed end-to-end on Mumbai catchments Building · JioGIS data unlock · Foundation model · Adoption shadow-mode Roadmap · 1000-city scale · Pricing · Customer Listening · SCM · Land Parcel · TAM
Status · Internal Track live · 01 - May - 2026
Tracks running
3
Internal · Google · JioGIS
Live opportunities
68
scored on the Internal Track
LI dimensions
10
composite 0–100 score per hex
Data categories
24
unified on H3 hex index
RIL Steering
Prateek Mathur · Alex Thomas · Gaurav Verma · Advait Pandit
RIL data partners
Biswaketan Kundu · Raghuram Velega

What's running. What's being built. What's next.

Internal Track · UCP / JioMart
Live

Opportunity Explorer.

Built in-house on UCP design language. Anchored on JioMart BAU production data aggregated at hex level. Natural-language workspaces score catchments, flag cannibalisation risk, propose intervention zones. End-to-end agentic decisioning — agents, data and reasoning in one runtime.

retailvista-agentic.ucp.fynd.com
Google Track · Joint MVP
Pilot

Agentic Site Intelligence.

Joint build with Google on Vertex AI, BigQuery, Street View and Maps. Cannibalisation, feasibility and site-selection agents validated end-to-end on live Mumbai catchments. Fynd owns Product, requirements, CUJ evaluation and output-quality feedback. Google owns infrastructure + agent orchestration.

Demo readiness QA loop running
JioGIS Track
Building

Data unlock + Foundation model.

Drives unlock of JioGIS layers and Reliance retail datasets. Foundation model development initiated for the spatial reasoning core. Kentrix MMR procurement anchored here as shared enrichment layer that feeds all three tracks once secured.

Modules running today.

Internal Track surfaces deployed on UCP design language. Each module is the Opportunity Explorer flow that takes a catchment from a scored signal to a pipeline decision. Live as of 01 - May - 2026 on real grocery NSO leads, aggregated on JioMart data.

ModuleStatusCapability
Home · Command CenterLiveKPI summary · Opportunity Intelligence (Comp. Gaps · Whitespace · Active Leads · Approved) · Top Opportunities · Leads · Alerts Requiring Attention · Store Rollout Velocity vs target
Explorer · Discovery + AnalysisLiveFilterable scored opportunity list (Site / LI / Whitespace / Competitor sub-scores) · synchronised Google Maps · Quick Summary · Full Analysis with 10 LI dimensions · Discover full-screen map · Pipeline Kanban (New → Shortlisted → Site Visit → Approved)
Workspace · AI Co-PilotLiveMulti-turn chat for feasibility · catchment · drive-time · brand-network analysis · structured AI outputs (demographics · accessibility · competitor landscape · brand assets) · Google Maps + MapLibre engine toggle · POI dataset and clustering layers
Users · Access ManagementLiveUser table · roles · status · modules · last login · invite flow · permission management
Data & Scoring layerLive10 LI dimensions (GDP · Demographics · Accessibility · Property Rates · Wealth · Footfall · Spending · Order Data · Building Density · Competition) · 4 sub-scores into 0–100 composite · Confidence and Priority labels (High / Medium / Low)

What operators see.

Six surfaces from the Internal Track. Each catchment is scored across 10 Location-Intelligence dimensions; sub-scores roll into a 0–100 composite with Confidence and Priority labels.

RetailVista Home · Command Center with Opportunity Intelligence panel and Top Opportunities list
Home · Command Center
Live

68 opportunities · 0 approved · 0 shortlisted · Opportunity Intelligence panel (3 Comp. Gaps · 65 Whitespace · 0 Leads · 0 Approved) · Top Opportunities (Goregaon East 69 · Dark Store Monginis 62 · Sarai Rohilla 57 …) · Leads feed · Alerts Requiring Attention · Store Rollout Velocity vs Q2 FY26 target.

Explorer · scored opportunity list with synchronised Google Maps
Explorer · scored list + map
Live

Filterable opportunity list (Site / LI / Whitespace / Competitor sub-scores) synchronised with Google Maps. Category and vertical filters. Search. Quick Summary popup on map markers. The scoring surface ground teams use to prioritise visits.

Explorer · Full Analysis with 10 LI dimensions and composite score breakdown
Full Analysis · 10 LI dimensions
Live

Goregaon East · Score 69 · Medium Confidence · Site 52 · LI 60 · Whitespace 90 · Competitor Gap 100. Composite 62/100 across GDP · Demographics · Accessibility · Property Rates · Wealth · Footfall · Spending · Order Data · Building Density · Competition. Each dimension labelled with its data source. Competitor presence: Swiggy Instamart 0.3km · Blinkit 0.4km · Zepto 0.7km · JioMart Quick Absent.

Explorer · Pipeline Kanban with New / Shortlisted / Site Visit / Approved / Dismissed columns
Pipeline · Kanban
Live

New (100) → Shortlisted → Site Visit → Approved → Dismissed. Region filter. Initialising progress indicator. Counts move with operator action. The decision system-of-record tying every scored opportunity to a real-world outcome the agents learn from.

Workspace · AI co-pilot with feasibility analysis output for Andheri
Workspace · AI Co-Pilot
Live

Multi-turn chat. Example: "Please perform feasibility analysis for Andheri for a new Reliance Digital store." Output: ranked hex IDs with rwi_mean (relative wealth), Total POIs, epoch_mean, height_mean. Strategic rationale + key metrics + contrast analysis (why other areas were not chosen). Google Maps and MapLibre engine toggle.

Users · access management table
Users · access management
Live

Member table · role · status · modules · last login. Invite flow · permission management · regenerate password for pending invites. Five active members across the Internal Track build team today; designed to scale to RIL business sponsors per format as adoption rolls out.

From raw signals to activated decision.

Sources → Spatial Aggregation → GIS Visualisation → Agentic Orchestration → Activation. Agents read live signals, run guarded playbooks, write outcomes back for continuous learning. A question asked once becomes a callable agent skill across all three tracks.

Layer 01 · Sources
Live

24 categories.

UCP · JioGIS · Retail store master · Customer locations / demographics / behaviour / intent · Government data (Census, NCRB, RBI, RERA, NFHS) · Third-party POI (Kentrix). See §04.

Layer 02 · Aggregation
Live

H3 hex index.

Hex-first aggregation · no PII · events snapped to H3 cells · only aggregated metrics per hex reach models. One spatial truth, many zooms (resolution 7 · 5.16 km² · resolution 9 · 0.10 km²).

Layer 03 · GIS visualisation
Pilot

Maps + heat overlays.

Per-hex score colouring on Mumbai catchments today · Google Maps + MapLibre engine toggle · Street View for last-mile. Pan-India per-hex Attractability heat is in build.

Layer 04 · Agentic orchestration
Mixed

Agent skills.

Six skills mapped — 3 Live (New Store Opening · Catchment Analysis · Dark-store Drive Time) · 1 Building (Customer Sentiment) · 2 Roadmap (Pricing Promotion · Land Parcel · Transport Optimisation). See §06 for per-skill status.

Layer 05 · Activation
Building

Hand-off surfaces.

Outputs to ALP · Granary · JioMart routing · brand-team workflows. New stores · Pricing · SCM Optimisation · Customer Listening as the activation cells from cornerstone deck.

RetailVista architecture · Sources → Ingest / Transform / Process / Serve / Analyze → Use Cases
Architecture diagram

Sources (UCP · JioGIS · Customer Listening · Google Earth · Reliance datasets · Kentrix · Government · Third-party) → Ingest (Geoson / CSV / Kafka / REST APIs) → Transform / Process / Serve (Databricks · BigQuery · Data Security & Governance · Agentic Orchestration: New Store Opening · Customer Onboarding · Pricing Promotion · Land Parcel Availability · Transport Optimization · Catchment Analysis) → Analyze (Conversational · Maps · Control Tower · Simulation · BI) → Use Cases (New Stores · SCM · Customer Listening · Pricing/Promotions · Land Parcels · TAM · Inventory).

Every signal that lands on the hex.

The shared data inventory feeding all three tracks. Reliance proprietary (UCP, store master), Jio (GIS, telco), Government (Census, NCRB, RBI, RERA, NFHS, SECC) and third-party POI (Kentrix). Volatility ranges from real-time to as-is.

#CategorySourceRefreshGranularityVariables
01Land base & public datasetJioGISAs-isLat / Long36 states · 105M buildings · 137M households · 21.7M POIs · 0.6M villages · 25.6K cities · 3.8M km
02Buildings master (Residential / Commercial)JioGISAs-isLat / Long1.3M km Fiber · 0.3M eNodeB OnAir
03Owned Retail facilities (stores · DCs · DSs · WHs)Retail store masterDailyLat / Long40,000+
04Customer locationsUCP · JioReal-timeLat / Long500M+
05Customer demographicsUCP · JioReal-timeLat / Long20+
06Customer digital behaviourUCP · Jio deviceDailyLat / Long10+
07Customer purchase intentUCP transactionalReal-timeLat / Long150+
08Customer interests & propensitiesUCP · Jio · Media inferredDailyLat / Long700+
09Civic InfrastructureGovt (Mission Antodaya · ODP)As-is200m – 1km240+
10Commercial · ServicesPOI (Kentrix)30 daysLat / Long38
11Commercial RetailPOI (Kentrix)30 daysLat / Long110+
12Crime statisticsGovt (NCRB)As-is500m110+
13DemographyGovt (SECC · Census) + GeoIQ180 days200m – 2000m230+
14EnvironmentGovt (Aridity · IMD)As-is200m – 500m2
15FinancePOI (RBI data)30 daysLat / Long23
16GeoIQ IndicesGeoIQ engineered180 days500m16
17GeographicalGeoIQ engineered180 daysRegion4
18HealthcareGovt (NFHS · Census · SECC) + POIAs-is / 30 days200m – 1000m140+
19InfrastructurePOI (Kentrix) · OSM · Public30 days / 1 yrLat / Long · 200m – 1000m80+
20Leisure & HospitalityPOI (Kentrix)30 daysLat / Long24
21MSMEGovt (third-party)3 months500m100+
22Mobility & FootfallThird-party30 daysHex 820
23Real EstateGovt (RERA) · Public listings30 / 90 days500m · Lat / Long3+
24Socio-economicGovt (SECC · Census) + GeoIQ180 days200m – 2000m900+
RetailVista Platform Layers · five-tier diagram
Platform Layers v0.1

Sources Layer (UCP + Customer Listening · JioGIS · External / Public) → Spatial Aggregation Layer (H3) → GIS Visualisation Layer (per-hex score colouring across India) → Agentic Orchestration Layer (Customer Sentiment · Pricing Promotion · New Store Opening) → Activation Layer (New Stores · Pricing · Transport & SCM Optimisation · Customer Listening).

H3 hex visualisation · Mumbai Attractability scoring with Ghatkopar 642 popup
H3 hex visualisation · Mumbai

Per-hex Attractability score (0-100 · Low / Medium / High / Very High) across MMR. Sample popup (Ghatkopar 642): Attractability 53/100 · Jio Penetration 43% · Connectivity 47% · UPI Growth +37% YoY · Competitors 4 · Real Estate ₹13.4K/sqft · Rent ₹201/sqft · Cannibalisation Risk 9% · Population 315K.

Where RetailVista is going.

A pan-India geospatial backbone running across 1000 cities, weighted on Google's agentic stack, built around per-household Digital Twins and the full network reference for every Reliance operation.

01 · Scale
Roadmap

1000 cities.

Pan-India spatial coverage at city level. Today's pilot is anchored on Mumbai across the Internal Track and Google MVP. Scale gates on JioGIS data unlock, Kentrix MMR enrichment, and foundation-model maturation.

02 · Primary surface
Pilot

Google as the pan-India agentic surface.

Engineering and partnership weight shifts to the Google joint track. Cannibalisation, feasibility and site-selection agents already validated end-to-end on Mumbai catchments. The Internal Track continues as the agentic capability benchmark.

03 · Foundation
Building

Full-stack geospatial foundation.

A 12-layer foundation platform spanning sources, aggregation, GIS, agentic orchestration and activation. The current 5-layer architecture is v0.1; the full stack closes the gap between raw signal and agent-ready decisioning across every Reliance vertical.

04 · Operating model
Building

Joint execution · dedicated org.

A dedicated RetailVista organisation operating across Reliance and Fynd. Engineering integration with Google folds into a single execution plan. Decisioning lives where the data lives — agent skills callable across every track.

05 · Customer Digital Twin
Roadmap

Per-customer Digital Twin · CDAP-native.

Each customer rendered as a hex-anchored Digital Twin of consumption — wallet, channel mix, household, fibre, mobile. Drives personalised offers at decision time. Sits on Layer 04 Agentic Orchestration once UCP customer lat/long unlocks at scale.

06 · Network reference
Roadmap

One GIS layer for every Reliance operation.

5G + 4G network coverage, dark stores, RIL Neighbourhood stores and Enterprise Premise Connectivity rendered on a single GIS reference. Delivery, expansion and infra teams route off the same substrate.

07 · Household identification
Roadmap

100M households identified · path to 180M owned.

All-India street-map resolution to identify 100M households from the Broadband and Air Fibre footprint. Sequenced path to the next 75–100M, ending at 150–180M owned-home relationships. Anchors the household Digital Twin and the JioGIS unlock.

Every business question becomes an agent skill.

Use cases mapped to the Activation Layer and Agentic Orchestration cells of the cornerstone architecture. First set is Live (NSO + dark-store drive time + catchment intelligence). Remaining cells are Building or Roadmap as agents inherit the data unlock from the JioGIS track. The platform is a single standard across formats — Grocery and Fresh through Digital, Trends, Strip Mall and Dark Store.

Use caseStatusWhat it does
New Store OpeningLiveScore every catchment · pre-score before ground visit · cannibalisation, demand and feasibility checked before any approval. Internal Track on real grocery NSO leads · Google Track Pilot on Mumbai catchments.
Catchment AnalysisLiveHex-level catchment scoring across formats. Workspace AI co-pilot answers feasibility, catchment, drive-time and brand-network questions in natural language.
Dark-store Drive TimeLiveDrive-time isochrones for q-commerce. First set of use cases per cornerstone executive summary.
Customer Sentiment AnalysisBuildingCustomer Listening data agentically aggregated to hex level. Inputs: UCP · Customer Listening interfaces.
Pricing & PromotionRoadmapHex-level price elasticity and promotion effectiveness. Activation-layer cell from cornerstone deck.
Customer ListeningRoadmapSpatial overlay of voice-of-customer signals. Activation-layer cell.
Transport & SCM OptimisationRoadmapOptimise transport routes against the spatial backbone. Activation-layer cell.
Land Parcel AvailabilityRoadmapLand parcel surfacing for Strip Mall and large-format expansion. Agentic Orchestration cell.
Inventory AvailabilityRoadmapHex-level inventory presence vs demand signal. Activation-layer cell.
Total Addressable Market (TAM)RoadmapPer-format TAM at hex resolution. Activation-layer cell.
Team · see /organisation/