US federal award intelligence (weekly)

Weekly recipient-resolved intelligence on the 1,000 largest US federal awards by cumulative obligated dollars with at least one transaction in the trailing 12 months. Two keyless sweeps of the official USA Spending search API (contracts A/B/C/D + assistance 02/03/04/05, trailing 365 days, top 500 each by obligated amount descending) are merged and deduped on the generated award id. Recipient names are entity-resolved (normalization + corporate-suffix stripping merges variants like 'LOCKHEED MARTIN CORP' with 'LOCKHEED MARTIN CORPORATION'); each entity carries its total obligated dollars and award count as peer context, as do the awarding agency and the place-of-performance state. A closed ordered-rule taxonomy labels every recipient sector (tribal, university, government, nonprofit, foreign, business — first match wins, business is the default). The documented 0-100 attention_score = 100 * (50% min-max-normalized log10(obligation) + 30% min-max-normalized log10(recipient-entity total) + 20% newcomer bonus), ranked as attention_rank (1 = hottest; ties: obligation desc, award id asc). new_entrant flags awards starting in the trailing 90 days; mega_award_flag marks >= $1B obligations; breakout_flag marks new entrants scoring >= 80. Every row carries the exact USA Spending award detail URL. Columns: ISO week, as-of date, award id, generated award id (primary key), award URL, award class (contract/assistance), award type verbatim, recipient name verbatim, canonical recipient entity key, recipient sector, description, start/end dates, place-of-performance state, awarding agency, obligation in USD, entity award count + entity total, agency award count + agency total, state award count + state total, new-entrant/mega/breakout flags, attention score + rank, country code (USA), row hash. Nullability: award_type, description, start_date, end_date, place_state and awarding_agency are null where the API returned none — kept, never invented. Caveats: obligated_amount_usd is the award's cumulative obligation across all its transactions (not the trailing window's spend); amounts tick as agencies post modifications, so re-fetches can mint new snapshots; the time filter covers awards with at least one transaction in the trailing window; attention_score is within-snapshot relative; entity resolution is name-based and can merge distinct local offices of one recipient. Sample use: order by attention_rank for the week's biggest federal money moves, filter breakout_flag for fresh giants, or group by recipient_sector to see who captures federal dollars.

Les titres et les descriptions proviennent des sources de données, en anglais.