Rural households in Ethiopia mostly rely on unimproved drinking water sources. Yet little is known concerning the preferences of rural residents when it comes to planning responses to unreliable water services. This study aims to determine households’ preferences and willingness to pay to improve drinking water services in Eastern Ethiopia. We collected discrete choice experiment data among 456 households. The study employs qualitative data from focus group discussions and key informant interviews to develop the attributes and their levels. Based on this, four attributes and their levels were selected: regularity of water supply, water treatment, predictability of water availability, and water price per 20 L of jerrycan as a water service fee. A mixed logit model was used to analyze the choice data. Results from the choice experiment show that respondents have a higher utility for the improvement alternatives than the status quo. On average, respondents prefer lower prices, regular water supply, treated water, and predictable water supply, with evidence of heterogeneity in the preferences for water service improvements. We found further that risk-averse respondents tend to prefer the improvement options with lower levels of uncertainty. Moreover, being a housewife, attending higher education, and belonging to low-income households were more likely to choose the improvement options. The marginal willingness to pay results indicate that respondents have a higher value for treated water than the other attributes. The findings provide insights into the preferences of rural households and the determinants of demand-driven water services in developing economies like Ethiopia.