The paper exploits ordered choice logit models with latent classes to account for unobservable consumer heterogeneity in analyzing the preferences for cashless payments in Japan for purchases of different sizes. Using the data of the Survey of Household Finance (2007–2020), we discover that consumers separate into classes of more and less frequent users of cashless payments for each category of purchases. The probability of belonging to the former class is positively related to the fact of consumer taking measures for the protection of their financial assets. The results reveal a statistically different effect of consumer socio‐demographic characteristics (sex, age, income, employment, education, household size), the binary variable for residence in a large city and Kanto region, and the dummies for 2019 and 2020 on the choice of cashless payments in the two latent classes.