Shipsy’s geofencing layer draws a polygon around every consignee address, every pickup location, every hub and cross-dock, and every known fraud-prone zone — and validates every operational event (attempt, delivery, pickup, scan) against those polygons in real time. When the polygon says no, the system says no: the event doesn’t close, and the anomaly is surfaced for review. Enterprises deploying the full geofence layer typically see fraud flag rates drop 60%+ while false positives stay below 2%.
Why we built this
GPS alone is not enough. A driver who marks a shipment “delivered” from 400 meters away either hasn’t actually delivered, has parked inconveniently, or is committing fraud. Without a geofence, you cannot tell the difference. Without a mechanism to act on the difference, your “delivered” rate is meaningless.
Enterprises were either ignoring the problem (and eating the chargebacks) or enforcing it with blunt threshold rules that broke in the real world (high-rise lobbies, gated communities, atoll island deliveries). We built the geofence layer as a tuned, per-address, self-improving mechanism — strict where it can be, flexible where it must be.
How it works
The geofence layer has four pieces:
Piece 1 — Polygon library. Every address, hub, pickup, and sensitive zone in the network has a stored polygon. Polygon shapes are learned: the initial polygon is a radius around the geocoded point, but it tightens and reshapes automatically as successful-delivery GPS captures accumulate. After 5–10 successful deliveries to an address, the polygon typically tightens from a default 150m circle to a 50m custom polygon around the true building footprint.
Piece 2 — Real-time event validation. Every operational event in the driver app (arrive, attempt, deliver, scan, pickup) is validated against the relevant polygon in under 100ms. In-polygon events proceed; out-of-polygon events fail the validation and require a reason code. Reason codes are typed (building access denied, customer meeting at street, consignee requested contactless at distance, security gate), each with a policy for whether the delivery can still close.
Piece 3 — Fraud-pattern layer. On top of the polygon check, a pattern detector watches for fraud signals: a driver whose out-of-polygon rate suddenly spikes, a driver who clusters many “delivered” events at one location (pattern consistent with mass-fake deliveries), repeated delivery events against addresses known to be vacant, GPS signals from devices with spoofing indicators, or photos whose EXIF location contradicts the geofence. These signals score drivers, routes, and shipments — high-score events are flagged even if they individually passed the polygon check.
Piece 4 — Sensitive-zone enforcement. Hubs, cross-docks, pickup points, and known fraud-prone zones have their own policy polygons. A pickup scan must occur inside the pickup polygon. A hub-arrival event must register within the hub polygon. This is how Shipsy enforces that operational events happened where they were supposed to happen — critical for contractor networks and cross-dock operations.
All geofence events and outcomes are retained for audit. Each ePOD artifact carries its geofence decision inline.
Here’s the state flow at a glance:
Early results
Enterprises deploying geofence validation typically report, within 60 days:
- 60%+ drop in fraud flags detected (driver fake-deliveries, contractor short-cycling, wrong-location pickups).
- COD dispute volume down 30–50% because “delivered from inside the geofence, with OTP and photo” is hard to contest.
- Sub-100ms validation latency per event, with no driver-visible lag.
- False-positive rate under 2% thanks to learned polygons and typed reason codes.
A global alco-bev leader operating across 70+ countries uses geofence enforcement on hub arrivals and primary distribution events to validate settlement-critical execution evidence.
What’s next
Three upgrades: NFC tag validation to supplement GPS at the door (NFC tap confirms physical presence), indoor micro-polygons for high-rises and malls where GPS alone is unreliable, and satellite-image verification for high-value deliveries, pulling recent imagery to confirm the building exists and is reachable.