What problem does it solve? Lead databases store company names the way registrars and scrapers found them — with legal suffixes, taglines, dba entities, parenthetical descriptors, and shouting capitals. Pasted into email copy, these strings read like an obvious mail merge and kill reply rates. This playbook turns each raw string into the short human form a person would say out loud, ready to drop into a {{company_clean}} variable. ## Core Features & Use Cases - Tiered cleaning chain: Defaults to Clay's free Normalize Company Name action, with an optional AI toggle (locked gpt-4o-mini prompt) for lists with taglines, dba entities, and second-language duplicates that a normalizer cannot fix. - Deterministic safety guards: A placeholder blocklist catches junk strings like Self-employed and Private Practice, and a substring verifier guarantees no word is ever invented — no second model call needed. - Benchmarked output contract: Returns company_clean, changed, and confidence per row at a measured 98/100 usable rate for roughly $0.15 per 1,000 rows, with abstain (empty string) instead of guesses. - Use Case: A cold-email campaign's first line reads "Noticed {{company_clean}} is hiring." Run the playbook over a 5,000-row lead list so 318, Inc dba Hamiltons Bud and Bloom becomes Hamiltons Bud and Bloom, while junk rows are excluded and routed to review. ## Quick Start Clean the company names in my lead list into short spoken forms and flag any rows that should be excluded from the campaign.
