From our most basic consumption to secondary needs, our spending habits reflect our life styles. Yet, in computational social sciences there is an open question about the existence of ubiquitous trends in spending habits by various groups at urban scale. Limited information collected by expenditure surveys have not proven conclusive in this regard. This is because, the frequency of purchases by type is highly uneven and follows a Zipf-like distribution. In this work, we apply text compression techniques to the purchase codes of credit card data to detect the significant sequences of transactions of each user. Five groups of consumers emerge when grouped by their similarity based on these sequences. Remarkably, individuals in each consumer group are also similar in age, total expenditure, gender, and the diversity of their social and mobility networks extracted by their mobile phone records. By properly deconstructing transaction data with Zipf-like distributions, we find that it can give us insights on collective behavior.