TCS Corpus
Tata Consultancy Services · 10 tools
Everything TCS publishes about itself — every investor filing from FY2012-13, the quarterly Data Sheet, and 11,800 pages of tcs.com — as a typed fact store plus hybrid search. 19,800 facts keyed by metric, period and dimension, each citing the sheet cell or PDF page it was read from; 2,500 dated business events; 229 resolved clients, partners and analyst firms.
Highlights
Numbers are typed, not retrieved
Revenue for one quarter, one segment, one currency is a five-way lookup that retrieval gets wrong. 19,812 facts keyed by metric × period × dimension × basis × measure across 107 reporting periods, FY2004-05 to the current quarter, each citing its source.
The filings, not just the press releases
1,190 filings catalogued from TCS's own IR servlet — statements under both Ind AS and IFRS, INR and USD earnings releases, fact sheets, shareholding patterns and earnings-call transcripts. TCV, AI revenue and reported margins are lifted from the release text, where the Data Sheet does not carry them.
Coverage is measured, not claimed
Every artifact that should exist was written to a ledger from tcs.com's own sitemap and IR catalog before any fetch, so completeness is a number: 98.8% of pages, 91.4% of linked documents, 87.5% of filings — the shortfall being scanned PDFs with no text layer, whose originals are kept and still readable.