Luxury Watch Market Intelligence Platform with AI-Assisted Research
WeyBee built the first MVP of a luxury-watch data product and continues to develop the platform: data ingestion from 1,600+ sources, search and valuation, AI-assisted research, web and mobile apps, subscriptions and internal operations, for about 199,000 active users a month.
AIThe client set out to become the neutral, data-first reference for anyone buying, selling or collecting luxury watches. Prices in this market are scattered across dealers, marketplaces and auction houses, in different formats and of varying quality, and buyers had no single view of what a watch is really worth.
- Bring listings and auction records from 1,600+ dealer, marketplace and auction sources into one clean, deduplicated catalogue.
- Turn that data into search, valuation and market views that buyers, collectors and professionals can trust.
- Make complex research easy for people who do not know every filter, reference or chart.
- Serve guests, registered users and paying members across web and mobile, with the right access for each tier.
- Keep the platform fast and maintainable as coverage, traffic and the database grew to 859 tables.
WeyBee built the first MVP and has stayed with the product ever since. A dedicated WeyBee team owns the .NET APIs, background jobs and email services, the React and Next.js web apps, the Flutter mobile apps, the Python and LangGraph AI services, and the data pipelines that collect, normalise, enrich and index market data. The same team builds the internal tools the client's staff use to run content, crawling and AI quality review.
Automated crawling and ingestion of listings and auction records from 1,600+ sources, with normalisation, enrichment, cross-source duplicate detection and search indexing in one catalogue.
Faceted search across dealer inventory and auction results by brand, model, reference, price, year, condition and materials, with structured brand, model and reference navigation.
Price history, median-price series, trends, performance over selected periods, value retention and variant comparisons.
A natural-language assistant that interprets watch and market questions, queries platform data and answers with listings, comparisons and performance charts.
Collectors organise the watches they own and follow their valuations over time, with performance work for larger collections.
Saved searches and alerts, with scheduled email delivery and open, bounce and complaint handling.
Guest, registered and paid experiences with authentication, registration prompts, Stripe subscription billing and tier-based access.
Flutter apps for iOS and Android with mobile-specific navigation and analytics.
A separate seller application for watch submissions, valuation requests and consignment intake.
Photo search that finds matching watches when the user does not know the name or reference.
Public brand and model pages, market reports and magazine content that bring in organic visitors.
Internal analysis and bespoke analyst reports for business clients.
Internal tools for content management, crawling operations, enrichment workflows and AI-response review.
Analytics across web, mobile and back end, and onboarding that captures each user's interests.
AI-assisted research for
the watch market
Watch research used to mean many filters and charts. We built a natural-language assistant on LangGraph that understands watch and market questions, queries the platform's own data and answers with listings, comparisons and performance charts, with every answer checked by an automated quality system.
Python services route each question to the right data tools and assemble a structured answer.
The assistant queries listings, auction records and price series, and shows comparisons and performance charts.
An LLM-based evaluation system and an internal audit dashboard let the team inspect and improve responses.
Users photograph a watch to find matches when they do not know the name or reference.
- Back end: .NET and C# for web APIs, background processing, scheduled jobs and email services.
- Web: React and Next.js in JavaScript and TypeScript, plus a separate React and TypeScript seller-advisory app.
- Mobile: Flutter and Dart for iOS and Android.
- AI services: Python with LangGraph LLM orchestration, a TypeScript conversational interface, structured data and tool querying, and automated LLM-based answer evaluation.
- APIs: REST and GraphQL between applications and platform services.
- Data: Microsoft SQL Server with stored procedures, and OpenSearch for indexed search and retrieval.
- Data engineering: automated crawling, scheduled ingestion, normalisation, enrichment, cross-source deduplication, indexing and queue-based synchronisation.
- Cloud and email: AWS, with Amazon SES for delivery and bounce and complaint feedback, and Brevo for contact synchronisation.
- Analytics and BI: PostHog, Mixpanel and Google Analytics across front end, mobile and back end; Tableau and Google Data Studio for reporting.
- Quality: Playwright test automation, AI answer evaluation and an internal assistant audit dashboard.
- About 199,000 active users a month across web and mobile.
- Automated ingestion from 1,600+ dealer, marketplace and auction sources.
- 886,000+ historical auction records from about 500 auction houses available for research and comparison.
- About 220,000 server-side emails sent a month through the platform's alert and notification services.
- Mobile monthly active users grew from 989 in April 2026 to 2,922 in August 2026, about 195% growth.
- A partnership in which WeyBee grew with the product, from first MVP to a platform with 859 database tables.
Fine Art Market Intelligence and Subscription Platform
Planning something similar?
Tell us what you're building and we'll suggest the right team, engagement model and first AI use case.