Evaluating Geofence Dwell-Time Accuracy in High-Density Urban Clusters

The Urban Canyon Problem

In high-density commercial corridors—such as Sukhumvit in Bangkok, Chang Khlan in Chiang Mai, or Mittraphap Road in Khon Kaen—satellite GPS signals rarely travel in a straight line from orbital constellations to a mobile handset antenna. Concrete high-rise facades, overhead transit guideways, and metal billboard structures reflect and refract microwave signals, creating the well-known multipath error.

For mobile applications attempting to calculate customer dwell times or trigger proximity check-ins, multipath interference generates artificial coordinate scattering ranging from 25 to over 150 meters.

       [ GNSS Satellite ]
              \
               \
  [ Tower A ]   \   [ Tower B ]
    |   |        \     |   |
    |   |   (Reflected)|   |
    |   |---------\    |   |
    |   |          \   |   |
    |   |          [ Handset ]  <-- Calculated point jumps across street!
===================================

Why Simple Circular Geofences Fail

The standard approach taken by many mobile engineering teams is to place a static circular geofence with a 100-meter radius around the centroid of a retail venue or transit stop. In physical practice, this circular geometry creates three severe operational flaws:

1. Street Overlap (The Drive-By Problem): A 100m circle around a 25-meter store front inevitably overlaps the adjacent 4-lane arterial road. Vehicles traveling at 60 km/h trigger “Store Arrival” events while waiting at traffic lights.
2. Back-Alley Bleed: The circular perimeter encompasses residential lanes, loading docks, and adjacent competitor premises behind the store.
3. Centroid Bias: If a large store has multiple pedestrian entrances on different sides, a single center point fails to capture users entering via north or south wings until they penetrate deep into the building (where satellite signals often drop out entirely).

Field Measurements: Jitter and Velocity Distributions

During a 14-day field assessment conducted by Link Cascade Base across 18 commercial venues, we compared raw circular trigger events against synchronized video ground-truth observations. The results revealed striking discrepancies:

• 38.4% of total circular triggers were generated by motorists or cyclists passing within the outer perimeter without stopping.
• 14.2% of triggers were caused by indoor coordinate drift (handsets jumping across the geofence perimeter due to Wi-Fi triangulation fluctuations).
• Only 47.4% of recorded arrivals represented genuine pedestrian store visits.

The Solution: Polygonal Geofences & Velocity Hysteresis

To restore data integrity, mobile systems must implement a dual-layer spatial filtering protocol:

1. Multi-Vertex Architectural Polygons (GeoJSON)

Replace circular approximations with precise 8-to-14 vertex polygons conforming to physical building walls, pedestrian walkways, and courtyard boundaries.

2. Velocity-Gated Hysteresis Windows

Rather than firing an immediate event upon coordinate intersection, the mobile client requires:

• Velocity Check: Handset speed must remain below 5.0 km/h (confirming pedestrian locomotion).
• Temporal Persistence: Handset coordinates must remain within the polygonal boundary for at least 90 consecutive seconds.
• Exit Dampening: Departure events are only confirmed after 45 seconds of continuous outside readings, preventing false exits during temporary indoor signal loss.

By applying these two algorithmic filters, false-positive trigger rates in our field tests fell from 38.4% to less than 1.6%, restoring trust to marketing automation and operational footfall metrics.