We Read 500 Pitch Decks So You Don't Have To NEW YORK - The conventional wisdom about pitch decks is mostly wrong. We proved it with data. Over three months, we trained an ML model on 500 Series A pitch decks submitted to top-tier VC firms between 2023 and 2026. Every deck was parsed across 47 features - slide count, design choices, word frequency, team pedigree, market sizing, revenue claims, competitive positioning - and correlated against actual fundraising outcomes. ==takeaways Market narrative beats product detail: 2.3x more raised on average Previous exit correlates with 1.8x higher raises - FAANG experience shows zero correlation "Risks and mitigations" slide is worth 1.6x more capital raised Decks under 12 slides outperform longer decks by 1.4x Revenue trajectory (direction) beats revenue magnitude (size) as a signal Team slides with photos outperform those without by 1.3x ==/takeaways **Market narrative beats product detail by a