NearDeal: Real-Time Geofenced Deals Marketplace for Local Retail
Building NearDeal across four phases: a marketing site and web app with automated business identity verification, a real-time geofencing engine that alerts nearby consumers to live deals within 60 seconds, native iOS/Android apps, and a Claude-powered engine that emails businesses three data-grounded suggestions after every deal expires.
Client
Location-Based Retail Marketplace Startup (NDA, Australia)
Project Value
Confidential, Active Engagement
Rating
The Challenge
The founder wanted consumers to discover time-sensitive local deals the moment they were near a participating business, and wanted business owners to post a deal and see exactly how it performed, without either side needing manual vetting or a data analyst.
The core technical risk was scale: matching a live deal against every nearby consumer in real time, at a range the client had only bracketed as "500 to 5,000 concurrent users," a 10x spread with materially different infrastructure implications at each end.
The Goal
Sequence a provisional Phase 1 launch (a non-real-time matching layer to reach market faster) behind an automated, real-time-capable architecture underneath, so Phase 2 upgrades the matching engine transparently, with existing users and deals migrating with no re-registration, no visible cutover, and no rebuild.
Solution & Implementation
1Analysis
Mapped the full deal lifecycle across three actors (consumer, business, and platform admin) and identified that server-side matching was not optional: mobile OS platforms cap simultaneous background-geofence monitoring per app at a level that does not scale once a consumer follows more than a handful of businesses, which pushed the entire matching problem onto the backend rather than the device.
2Designing Solution
Built the geofencing engine on PostGIS spatial queries (ST_DWithin) with geohash-based spatial partitioning, so a deal broadcast does not trigger a full-table scan against every registered consumer location. New deals push to the matching engine over WebSocket/SSE the instant they are posted, and matches fan out to push notification services (FCM for Android, APNs for iOS) with an offline-delivery fallback queue.
3Customizing Business Logic
Designed multi-branch fan-out: a business with several locations selects which branch (or branches) a given deal applies to, and the geofencing engine evaluates matches from each branch independently while still sending a consumer within range of multiple branches a single, de-duplicated notification.
Automated business-identity verification runs against a national business registry API at sign-up, with a clear rejection message and an admin override path for edge cases, replacing a manual approval queue entirely.
4Scale & Optimize
Closed the loop for business owners with a Claude API integration: within one hour of a deal expiring, the system assembles that deal's funnel metrics against the business's own history and its category benchmark, and emails three suggestions that must reference the business's actual figures, generic advice is rejected and retried once before falling back to a metrics-only email if the API is unavailable, so the business always receives a summary regardless of third-party uptime.
Results & Impact
Notification Latency (Target)
Concurrent Load (Target)
Crash-Free Sessions (Target)
AI Suggestion Turnaround (Target)
Business Onboarding (Target)
Server-side geofencing was an architectural requirement, not a preference, mobile platforms' per-app background-geofence caps make device-side matching non-viable once a consumer follows more than a handful of businesses
Multi-branch fan-out evaluates matches from every selected branch independently but still sends each consumer a single, de-duplicated alert: no duplicate notifications regardless of how many branches are in range
Automated business identity verification against a national business registry replaces a manual approval queue, with an admin override path reserved for genuine edge cases only
Claude API suggestions are output-validated before sending, generic, non-data-grounded advice is rejected and retried once, with a deterministic metrics-only fallback if the API is unavailable
The Phase 1 provisional matching layer is designed to be replaced transparently by the Phase 2 real-time engine, existing users and deals migrate with no re-registration and no visible cutover
4-phase engagement sequencing (web foundation → geofencing engine → native apps → AI insights) lets the client reach market on Phase 1 while the real-time and native layers are still in build
Key Technologies
Project Gallery
Technical Approach
NearDeal's geofencing engine runs server-side PostGIS spatial queries (ST_DWithin) against active consumer locations and live deal radii, combined with geohash-based spatial partitioning so a deal broadcast queries only the geohash cells that could plausibly be in range rather than scanning the full user table.
This was an architectural requirement, not a preference: both major mobile platforms cap simultaneous background-geofence monitoring per app at roughly 20 regions, which makes device-side matching non-viable once a consumer follows more than a handful of businesses.
New deals push into the matching pipeline over WebSocket/SSE the instant they are posted, and matched consumers receive alerts through Firebase Cloud Messaging (Android) and Apple Push Notification service (iOS), with an offline-delivery fallback queue for unreachable devices.
The target designed and load-tested against is sub-60-second delivery from broadcast to device at a confirmed concurrency figure, not a ballpark range, since infrastructure decisions like connection pooling and queue partitioning differ materially across a wide load estimate.
Multi-branch businesses select which location(s) a deal applies to, and the engine evaluates each branch's radius independently while de-duplicating at the notification-assembly stage so a consumer in range of several branches gets exactly one alert. Business onboarding is fully automated against a national business registry API, replacing a manual approval queue with an admin override reserved for edge cases.
The full architecture, including why device-side geofencing doesn't scale and how the push pipeline is built, is covered in our real-time geofencing architecture guide.
Frequently Asked Questions
About Ortem Technologies
Ortem Technologies is a premier custom software, mobile app, and AI development company. We serve enterprise and startup clients across the USA, UK, Australia, Canada, and the Middle East. Our cross-industry expertise spans fintech, healthcare, and logistics, enabling us to deliver scalable, secure, and innovative digital solutions worldwide.
Related Case Studies
Multinational Transportation & Field Services Organization (NDA)
OpsCommand: Enterprise Fleet, Logistics & Intelligent Operations Platform
Confidential US-Based Manufacturing Organization (NDA)
PlantOps: Enterprise Manufacturing Operations & Industrial IoT Platform
Ready to Build Something Like This?
We've built retail tech & hyperlocal marketplace solutions like this for teams across the US, UK, and Middle East. Fixed scope. Fixed price. Senior engineers from day one.
Free 30-minute scoping call: no commitment, no sales pitch.