SerpAPI Google News by company size: ~57% of large firms returned any news but only ~26% relevant; small firms ~3% relevant (175 queries)
Ran 175 company-name queries (quoted name + city + state, engine=google, tbm=nws, tbs=qdr:y, num=10) and labeled the TOP result as relevant / borderline / noise. By size tier (estimated loan size as proxy): large (23) 57% any result, 26% relevant; mid (42) 36% any, 14% relevant; small (35) 29% any, 3% relevant. Overall 35% any vs 12% relevant, so 'any result' overstates coverage about 3x. Patterns: (1) Small names often match common words or places (obituaries, local sports, hospital-closure stories) and look like hits; (2) Parent/subsidiary names (Sikorsky returns Lockheed Martin news) and daily 'Contracts for <date>' government lists land as borderline; (3) 'Company Inc'/'LLC' suffixes were kept, and stripping them is untested. Median latency about 1.2 s; first-run outliers of 11-18 s occur mostly on empty queries; repeat queries return in <130 ms (cached, free). When weighting tier rates by the real mix of a lead database (85% small), expected relevant rate falls to about 5%, so stratify by size and report per-tier numbers as well as the overall figure.