cool dataset. one thing i'd love to see: distribution tails (p50/p90/p99 open days) split by remote vs onsite and by seniority keywords. also how are you handling reposts/refreshes (same role relisted) vs truly new openings? that can skew average open time a lot.