Household air pollution (HAP) from combustion of biomass fuel, such as wood and animal dung, is among the leading environmental risk factors for preventable disease. Close to half of the world's population relies on biomass cookstoves for their daily cooking needs. Understanding factors that affect HAP can inform measures to maximize the effectiveness of cookstove interventions in a cost-effective manner. However, the impact of kitchen and household characteristics, as well as the presence of secondary stoves, on HAP concentrations is poorly understood in Puno, Peru.To explore how household characteristics explain variability of kitchen area concentrations and personal exposures to CO, PM2.5 and BC from biomass cookstoves among women in rural Peru.Household characteristics (including kitchen materials and layout, wealth, and cooking behaviors) and HAP measurements were collected from 180 households in Puno, Peru, from baseline measurements of a randomized trial. Kitchen area concentrations and personal exposures to carbon monoxide (CO), fine particulate matter (PM2.5) and black carbon (BC) were sampled for 48 h. We implemented simple and multivariable linear regression models to determine the associations between household characteristics and both kitchen area concentration and personal exposure to each pollutant.Mean daily kitchen area concentrations and personal exposures to HAP were, on average, 48 times above World Health Organization indoor guidelines for PM2.5. We found that roof type explained the most variability in HAP and was strongly associated with both kitchen area concentrations and personal exposures for all pollutants after adjusting for other household variables. Personal exposures were 27%-36% lower for PM2.5, CO and BC, in households with corrugated metal roofs, compared to roofs made of natural materials (straw, totora or reed) after adjusting for other factors. Higher kitchen area concentrations were also associated with less wealth, owning more animals, or sampling during the dry season in multivariable models. Having a liquefied petroleum gas (LPG) stove and having a chimney were associated with lower personal exposures, but were not associated with kitchen area concentrations. Personal exposures were lower by 21% for PM2.5 and 28% for CO and BC concentrations among participants who had both LPG and biomass stoves compared to those with only biomass cookstoves adjusting for other household factors.Characterizing HAP within different settings can help identify effective and culturally-relevant solutions to reduce HAP exposures. We found that housing roof type is strongly related to kitchen area concentrations and personal exposures to HAP, perhaps because of greater ventilation in kitchens with metal roofs compared to those with thatch roofs. Although HAP concentrations remained above guidelines for all households, promoting use of metal roof materials and LPG stoves may be actionable interventions that can help reduce exposures to HAP in high-altitude rural Peru and similar settings.