How the SousLab corpus is sourced, normalized, and audited.

The corpus is sourced from three feeds, internally tagged tier1_a, tier2_a, and tier2_b. Each feed has its own ingestion pipeline, schema normalizer, and freshness audit. The public API exposes the abstract data_tier filter — buyers can filter to first-party (tier1) only, or include aggregator-curated (tier2) for breadth. Underlying feed identities are not surfaced publicly.

• tier1_a: daily refresh, monitored. Failure alerts fire within the same day.
• tier2_a, tier2_b: weekly refresh on a staggered schedule.
• Chain entity layer: revalidated weekly. Ticker mapping reconciliation runs on a quarterly schedule.
• Premium tier (Scale+ and Enterprise): same-day refresh on tier1_a + daily Parquet drops.

Restaurants are normalized to consumer-facing brands (e.g. multiple McDonald's storefronts → one chain entity). Brands are then mapped to their corporate parents and, where applicable, public tickers. The chain coverage page lists the top 25 publicly-traded restaurant tickers we cover at the entity level. Mapping methodology, location-count audit, and ticker selection criteria are detailed in the methodology PDF.

Freshness, completeness, and coverage-by-metro audits run alongside refresh. Numbers worth publicly stating: 449,716 restaurants covered, 44.3M+ menu items indexed, top 25 publicly-traded restaurant tickers mapped at the entity level. Per-source freshness, accuracy spot-check methodology, and known coverage gaps live in the methodology PDF — granularity intentional, sent under a soft NDA on request.

Versioning

This is methodology v1.0. Major revisions trigger a re-issued PDF and notification to all licensed customers; minor revisions are tracked inline in /changelog. If you have an active license or a pending request, you'll always have the methodology version that matches the data version you received.