Sepsis is an uncommon disease, especially in infants ≥35 weeks' gestation when the incidence in the general population is 0.5-1.0/1000 live births. Although the incidence is higher in symptomatic infants and those with risk factors, sepsis evaluations are far more common than infants with proven sepsis. How can practitioners identify the one baby with sepsis from the many infants with risk factors or clinical signs (finding a needle in a haystack)? For many decades, clinicians have relied on laboratory testing to decide which well appearing infants require antibiotics at birth. Unfortunately, screening laboratory tests have a poor predictive accuracy and cannot pinpoint which infants need empiric antibiotic therapy. The alternative is to treat every infant with certain risk factors for sepsis such as chorioamnionitis, even if the infant appears well. That was the approach recommended by the Committee on Fetus and Newborn in 2010, which led to overtreatment of a large population of uninfected babies. Treating every symptomatic infant with antibiotics is also problematic because some infants are symptomatic solely because of the physiological transition they undergo in the immediate postnatal period.