Price intelligence SaaS platform for ecommerce: the Revaly case

Revaly is a SaaS platform giving an online store price intelligence on its competition. The product promise is also its technical challenge.

Competing on something other than price requires first knowing the other price, and that means reading thousands of product listings on third party sites, each with its own structure, and keeping that reading alive over time.

• Heterogeneous catalogs: every site describes its products differently, and a fixed rule breaks with the first redesign.
• Tenant isolation: one store's data cannot leak into another, and that has to be guaranteed at the API layer.
• Plan based usage: if the product sells by monitored URL volume, those have to be counted and enforced.

Without solving all three at once, the product works in the demo and not in production.

We built the platform on a three service architecture, with separate databases so one service's schema changes cannot alter another's tables.

Tenant and quota administration lives in its own service, queried over HTTP rather than sharing tables.

Three services, separate responsibilities

• Administration service in Java with Spring Boot: tenants, plans and quota control, with its own database.
• Operational back end in Python with FastAPI: extraction, processing and product logic.
• Front end in React on Vite, with the client control panel.

Multi tenant and plans

• Tenant isolation propagated at the API layer through an identification header.
• Authentication delegated to a managed provider.
• Subscription plans differentiated by monitored URL volume, from a free no card tier to enterprise, with subscription billing integrated.

Reading catalogs you do not own

• Extraction of third party product listings with language model assistance, to interpret heterogeneous structures on a custom scraping layer.
• Connectors with the commerce platforms where the client's own store lives, including publishing price changes back to it.

Architecture and technology

• Spring Boot, FastAPI and React on Vite, with PostgreSQL and separate databases per service.
• Full stack runnable locally in containers, with external service keys deliberately disabled.

Learn more about software development and artificial intelligence.

The product sustains third party catalog reading, tenant isolation and volume based billing, which are the three things that make it sellable.

Starting point

• Third party catalogs with no common structure, unreadable with fixed rules.
• Tenant isolation and plan quotas required from day one.

Strategic impact

• Language model assisted extraction survives a third party site redesign where a fixed rule would break.
• Isolation travels in the API layer, not in the discipline of whoever writes the query.
• Quota control lives in its own service, so the plan is genuinely enforced.
• The local environment runs the full stack without touching real data.