Wikidata Sparql Query Company Data: Dataset Structure and Field Coverage

One honest limit matters before you write a query: Datadory's research pass on this record verified book-oriented properties (ISBN-13 P212, publisher P123, title P1476, publication date P577) against live responses, while noting that property coverage across arbitrary entities is uneven. Treat the graph as a linkage backbone, then validate the specific P-properties you join against a sample of companies.

1. Write the triple patterns. Bind a subject variable to items classified as organizations or businesses, then chain property paths for each attribute you want. Property paths take the form wdt:Pxxx, and the wikibase label service turns resulting URIs into localized display names.

1. Send it. GET with a URL-encoded SPARQL string, or POST the same query with Accept: application/sparql-results+json. Swap .json for .csv or .tsv` when a spreadsheet is the destination.

1. Read the bindings. Each solution maps variables to typed values - entity URIs like string literals such as ISBN-13 978-87-7934-585-0, and xsd:dateTime values like 2011-01-01T00:00:00Z` for date properties.

1. Iterate interactively first. The browser UI at query.wikidata.org bundles a visual Query Builder, an interactive Query Helper for assembling patterns, and table, map, graph, timeline and image-grid result views - cheaper than debugging in code.

Two etiquette rules apply throughout: send a descriptive User-Agent header, and expect each query to run under timeout and memory limits documented in the WDQS user manual rather than on the landing page.

Wikidata models companies through the same statement machinery it uses for books, and that symmetry is the practical point: the properties Datadory's researchers verified live on this record - P123 resolving a publisher to a publishing-house item, P50 linking work to author item, P31 classifying the entity, P291 recording place, P1476 carrying language-tagged text - are exactly the item-to-item edges that let one query roll a product up to its manufacturer and the manufacturer up to its corporate family. A catalog row keyed on an external identifier becomes a graph node once you join it on the identifier property and walk outward from there.

Fill rates are the caveat worth budgeting for. Datadory's verification notes flag uneven property coverage - many real-world entities lack the full statement set - so measure how many of your resolved companies actually carry the attribute you need before building a pipeline on it. Where coverage thins out, the pool's statistical datasets carry the load instead.

The graph tells you who a company is; official statistics tell you how big its channel is. The U.S. Census Bureau Data API - CBP / Economic Census / Wholesale Trade is a quality-score-9 REST/JSON service over 1,798 datasets, commercial delivery terms, whose County Business Patterns vintages 1986-2023 report establishment counts (ESTAB), employment (EMP) and annual payroll (PAYANN) for NAICS 423430 - computer and peripheral equipment merchant wholesalers - at national, state, county, MSA, congressional-district and ZIP-code levels.