Well-designed financial products improve the overall financial health of users. The design of products is particularly important for low-income customers, for whom product design drives behavior. In this paper, we offer insights on low-income customers’ savings behavior and on how they use their savings accounts. More specifically, we focus on detecting and measuring the effects of a set of explanatory variables on transaction amount. To do so, we use quantile regression (QR) and apply it to a novel dataset collected from a financial institution in Nigeria. The data show individual transactions made using the account over time, along with additional socioeconomic information on each customer. Using these data, we specify a model that incorporates customer age, account age, location, transaction type, gender, and seasonality effects, evaluating their correlation with transaction size. With the QR model, we are able to study the effect of the explanatory variables within each quantile of transaction amount instead of just showing trends on average. This is the first study to examine transaction size among low-income customers through a gender lens using QR. All of the variables incorporated in this model have a significant effect on transaction size. However, among all of the explanatory variables, the season in which a customer places a transaction (seasonality effect) has the largest impact on predicting transaction amounts.