Signal Source Catalog (Visual)

Definition

The visual signal source catalog is a structured inventory of satellite imagery, street-level photography, and computer vision capabilities that detect physical "revenue signatures" -- observable deterioration patterns on commercial properties (thermal stains, pothole clusters, facade cracks, roof membrane failures, vegetation encroachment) that indicate maintenance needs, asset distress, or operational decline. [src4] By cross-referencing GPS coordinates with county tax assessor APIs and corporate registry data, these visual signals can be resolved to specific property owners and decision-makers, creating sales intelligence that proves physical need rather than inferring intent. [src5]

Key Properties

• Satellite Imagery (Maxar/Planet): Daily refresh cycles at 30-50cm resolution -- sufficient to detect roof deterioration, parking lot damage, and facility changes across entire metro areas [src1] [src2]
• Street-Level Data (Google Maps/Mapillary): Ground-level photography captures facade condition, signage changes, and loading dock activity from public rights-of-way [src3]
• Revenue Signatures: Visual patterns correlating with spending needs -- thermal stains on flat roofs ($15K-150K repair), pothole clusters ($25K-500K repaving), facade crack propagation (structural settlement) [src5]
• Vision-LLM Defect Detection: 2026-era multimodal models classify defects through prompt engineering alone -- no custom training required for initial deployment [src4]
• Owner Resolution Pipeline: GPS coordinates from detections feed into county tax assessor APIs (CoreLogic, ATTOM) and corporate registries to identify responsible decision-makers [src5]
• Temporal Change Detection: Comparing imagery across periods reveals deterioration trajectories -- accelerating degradation indicates deferred maintenance and increasing urgency [src1]

Constraints

• Satellite imagery subscriptions cost $5K-50K/year depending on coverage and resolution [src1]
• Vision-LLM defect detection has 10-20% false positive rates requiring human verification [src5]
• Strictly limited to commercial/industrial property exteriors -- no residential surveillance or interior assessment [src4]
• Weather, season, and image quality create noise -- snow cover, shadows, and construction staging produce false signals [src2]
• Owner resolution requires separate data infrastructure for GPS-to-owner matching [src5]

Framework Selection Decision Tree

START -- User needs visual/physical signal sources for B2B intelligence
├── What type of property defect?
│   ├── Roof deterioration (commercial)
│   │   └── Satellite imagery (Maxar/Planet) + thermal analysis
│   ├── Parking lot / pavement damage
│   │   └── Satellite + street-level combination
│   ├── Facade / structural condition
│   │   └── Street-level imagery primary, satellite supplementary
│   └── General facility health monitoring
│       └── Visual Signal Catalog ← YOU ARE HERE
├── What geographic scale?
│   ├── Single metro (50 sq miles)
│   │   --> Planet Labs daily monitoring is cost-effective
│   ├── Regional (state-level)
│   │   --> Maxar archive + periodic tasking for high-value areas
│   └── National
│       --> Start with Google Maps/Mapillary, add satellite for priority zones
└── Real-time monitoring or point-in-time?
    ├── Real-time --> Planet Labs daily refresh ($10K-50K/year)
    └── Point-in-time --> Maxar archive imagery ($500-5K per assessment)

Application Checklist

Step 1: Define Target Defect Taxonomy

• Inputs needed: Target service vertical, geographic focus, typical contract values
• Output: Visual defect classification schema mapping patterns to service needs and estimated contract values
• Constraint: Start with 2-3 high-confidence defect types -- attempting everything simultaneously overwhelms quality control [src5]

Step 2: Select Imagery Sources by Coverage and Resolution

• Inputs needed: Geographic focus, defect taxonomy, budget constraints
• Output: Imagery source stack combining satellite (area coverage) and street-level (detail resolution)
• Constraint: Minimum 50cm resolution for roof defects; minimum 10cm for facade cracks [src4]

Step 3: Build Vision-LLM Detection Pipeline

• Inputs needed: Imagery access, defect taxonomy, 100+ labeled examples per defect type
• Output: Automated pipeline processing imagery tiles with confidence scores and GPS tags
• Constraint: Human-in-the-loop verification until false positive rate drops below 15% [src5]

Step 4: Implement Owner Resolution and Outreach

• Inputs needed: GPS-tagged detections, county tax assessor API access, corporate registries
• Output: Complete lead records linking defects to identified owners with contact info and repair estimates
• Constraint: Owner resolution accuracy must exceed 85% before automated outreach [src5]

Anti-Patterns

Wrong: Launching automated outreach on raw Vision-LLM detections

Sending unverified "damage detected" notifications creates credibility damage when false positives reach owners who see no problem. [src5]

Correct: Human-verify all detections and include raw source imagery

Every pre-emptive bid package must contain the dated source image, putting final verification on the human rep. [src4]

Wrong: Monitoring residential properties or property interiors

Extending detection beyond commercial exteriors creates privacy violations and community backlash. [src4]

Correct: Focus strictly on commercial zones and publicly visible exteriors

Limit monitoring to commercial/industrial zoning and conditions visible from public vantage points. [src3]

Wrong: Using a single imagery source for all detection types

Satellite alone cannot detect facade defects; street-level alone cannot assess roofs or provide area-wide coverage. [src2]

Correct: Layer multiple imagery sources matched to defect type

Use satellite for overhead defects (roofs, parking lots) and street-level for vertical surfaces (facades, signage). [src1]

Common Misconceptions

Misconception: Visual signal detection requires custom-trained computer vision models.
Reality: 2026-era Vision-LLMs classify most commercial property defects through prompt engineering alone. Custom training improves edge-case accuracy but is not required initially. [src5]

Misconception: Satellite imagery is too expensive for lead generation.
Reality: Planet Labs daily monitoring covers 50 sq miles for $5K-10K/year. At $150+ per verified lead, ROI turns positive with fewer than 50 leads annually. [src1]

Misconception: Visual signals only apply to construction and property maintenance.
Reality: Visual detection applies to insurance underwriting, municipal infrastructure, commercial real estate due diligence, and facility management intelligence. [src4]

Comparison with Similar Concepts

When This Matters

Fetch this when a user asks about using satellite imagery for sales intelligence, detecting physical property damage at scale, building computer vision pipelines for lead generation, or understanding how visual data complements regulatory and behavioral signals.