BACKGROUND: BCG has been recommended at birth in countries with a high tuberculosis burden for decades, yet delayed vaccination is widespread. To support a WHO guidance review, we estimated the potential global tuberculosis mortality benefit of administering BCG on time and consequences of later administration. METHODS: We estimated age-specific BCG coverage in 152 high-burden countries using data from large, nationally representative household surveys, to parameterise a static mathematical model, calibrated to global childhood tuberculosis deaths in 2016. 12 hypothetical scenarios explored the effect of BCG delivery at birth, 6 weeks, 6 months, or 9-12 months, on tuberculosis deaths per global birth cohort by age 15 years, including delivery at the time of the first diphtheria-tetanus-pertussis vaccine (DTP1) or the first measles-containing vaccine (MCV1). We assumed constant vaccine efficacy by age, but varied coverage and degree of vaccination delay, including no delay. FINDINGS: In 152 high-burden countries, we estimated that BCG coverage in 2016 was 37% at 1 week of age, 67% at 6 weeks, and 92% at 3 years. Modelled scenarios in which 92% BCG coverage was achieved at birth reduced tuberculosis deaths in the global birth cohort by 5449 (95% uncertainty range 218-15 071) or 2·8% (0·1-7·0) by age 15 years. 100% coverage at birth reduced tuberculosis deaths by 16·5% (0·7-41·9). Later administration increased tuberculosis deaths-eg, BCG vaccination at 6 weeks, the recommended age of DTP1, increased tuberculosis deaths by 0·2% (0-0·4), even if BCG reached DTP1 coverage levels (94% at 3 years). INTERPRETATION: Reducing delays and increasing coverage at birth would substantially reduce global paediatric tuberculosis mortality. Modelled scenarios whereby BCG was administered later in the infant schedule were all estimated to increase tuberculosis deaths, even with increased coverage. The WHO recommendation for BCG at birth should be maintained and emphasised. FUNDING: WHO.