Early childhood programmes frequently lose effectiveness at scale, yet the role of the workforce remains poorly understood. We document substantial heterogeneity in workforce effectiveness in England's national home-visiting programme for first-time teenage mothers, despite a highly-structured curriculum and well-qualified staff. Exploiting quasi-random assignment of mothers to family nurses, we estimate that a one standard deviation increase in workforce effectiveness raises children's cognitive and socio-emotional development by 0.20-0.23 SD. Structural quality — observable worker characteristics — does not predict effectiveness, but process quality — how visits are delivered — does. Greater effectiveness is linked with improvements in maternal mental health and risk behaviours.