Powered By Industry Leading Data & Technology

Our Data

In today's rapidly evolving fashion landscape, organisations are increasingly reliant on market intelligence to inform their strategic decisions – but not all fashion data sources are equal. Missing attributes, vague titles, duplicate listings, counterfeits and the lack of a standardised taxonomy all obscure the insights that matter.

Our data undergoes rigorous validation through expert fashion annotators and systematic quality controls to meet our exceptionally high standards.

The same cultural expertise and collaborative annotation methodology that powers our data services for leading AI companies ensures every data point reflects genuine fashion intelligence. Only verified, expert-annotated data enters the TRUSS Data Layer - our single source of fashion market truth.

At our core is our live system, tracking product performance, brand equity flow, and market positioning to offer the most current view of any fashion entity throughout its entire market lifecycle.

Data Coverage

We collect from 22 major resale and commerce platforms and dozens of smaller sites. Our coverage increases monthly as we integrate more sources. Weekly updates capture market movements in real-time, while daily monitoring of select high-velocity platforms ensures nothing is missed.

But collection is only the beginning. Raw scraped data is unusable for serious analysis

The TRUSS Data Processing Engine

Our cloud pipeline includes automated testing, continuous monitoring, and internal dashboards that ensure data flows from source to product without interruption. Each component addresses specific data quality challenges that make raw secondhand data unusable:

Source Accuracy & Collection:

• Weekly scraping across all major platforms, with daily updates for high-velocity sources
• Completion rate monitoring through "resurrection" tracking - measuring how often items reappear after being missed
• Exact sold timestamp extraction where platform APIs allow, with estimated accuracy within days for non-API sources
• Platform-specific delisting logic that distinguishes genuine sales from temporary removals

Listing Deduplication (TLDR™) Our proprietary TRUSS Listing Duplicate Resolver uses multiple signals:

• Metadata correlation: Title, description, price, and timestamp pattern analysis
• Visual recognition: Semantic similarity algorithms identify near-duplicate images even after reformatting, compression, or watermarking
• Global seller database: We track sellers operating across multiple platforms and deduplicate their inventory to avoid double-counting
• Cross-platform validation: The same item listed on multiple sites gets consolidated into single data points

Human-Led Annotation Pipeline:

• Text extraction: >97% accuracy where model information exists in listings
• Visual AI gap-filling: For the ~50% of listings missing critical details (like popular models such as Chanel Boy bags), our computer vision identifies products from images
• Expert validation: Manual annotation and quality assurance for high-value or complex items
• Active learning: Human corrections continuously improve visual recognition accuracy

‍Sold Item Intelligence. Marking items as sold requires different approaches:

• API-connected platforms: Partnerships w/direct confirmation of sold status to the millisecond
• ‍Non-API platforms: Inferred status using delisted time combined with pricing behavior patterns
• ‍Noise filtering:
  • Refreshed listings identified by timestamp reset patterns
  • Re-listing practices detected via repeated metadata from the same seller
  • Account deletions and cleanouts resolved to reduce statistical noise

Counterfeit Mitigation:

• Seller reputation scoring: Platform-specific rating analysis where available
• Price outlier detection: Cross-platform price comparison flags obvious fakes
• OCR & serial number extraction: Automated reading of authentication details when visible
• Anomaly flagging: Pattern recognition identifies suspicious listing behaviors

Quality Assurance & Monitoring

• Automated testing: Continuous validation of data pipeline integrity
• Completion monitoring: Real-time tracking of scrape success rates and data gaps
• Internal dashboards: Live monitoring of data flow health across all sources
• Error handling: Automatic retry logic and failure notifications maintain data consistency

Domain Expertise That Goes Deep

The TRUSS Ontology

Our taxonomy was designed from first principles to map from any data source to yours. Unlike industry taxonomies built for specific use cases, ours covers the full spectrum of applications across fashion and luxury markets.

Component-Level Precision: We extract style codes, serial numbers, material composition, hardware details, sizing measurements, and release years through visual recognition combined with OCR. Nothing is too granular if it drives insight.

Cultural Intelligence: Our annotation teams understand that "Grunge" differs from "Gothic Prep," that vintage Hermès carries different signals than contemporary luxury, and that regional aesthetic preferences shape market behaviour.

Hybrid Expert Annotation "Centaur Approach"

The TRUSS Annotation Ecosystem combines domain expertise with machine learning efficiency. Fashion experts use our full AI stack in what we call a "centuar annotation approach" - delivering accuracy that pure crowdsourcing can't match and speed that manual processes can't achieve.

Why This Matters?

Across TRUSS' suite of products, from application layer dashboards to market intelligence APIs to foundational data annotation, we ensure pinpoint accuracy with insights and data that can be trusted. Every data point has been validated, contextualised, and prepared for decision-making or model training.

Data Accuracy Determines Decision Quality & Model Performance

Markets move fast, and incomplete or inaccurate data leads to costly mistakes. Our multi-layered approach to data cleaning and validation means the insights you build decisions on are trustworthy.

Consistent Ontology Enables Integration

Your existing systems don't need to change. Our taxonomy maps to industry standards while providing the granularity needed for sophisticated analysis.

Scale Without Compromise

Weekly updates across global platforms, automated quality assurance, and expert validation create a data foundation that scales with your needs while maintaining reliability.