Compared to 50% on 42 correct-tool raw-graph cases. The 30-point gain comes entirely from redesigning the attribution target.