We investigated the effect of pandemic-related disruption on the frequency of non-grocery brick-and-mortar shopping. We conducted a quasi-longitudinal survey with structured ques- tions that captured shopping experiences before and during the disruption. We employed machine learning algorithms and statistical tests such as chi-square, random forest model, and permutation test. Based on the permutation test, prior to the disruption, online shop- ping frequency was the sole feature statistically associated with brick-and-mortar shopping frequency. During the disruption, perceived safety of online shopping emerged as the only statistically significant feature. Delivery vehicle-induced traffic issues were not statistically associated with brick-and-mortar shopping frequency. Although crowdsourced deliveries were not significant, they exhibited a proportional relationship with shopping frequency ac- cording to SHAP values. Regular retrieval of orders from parcel lockers did not result in more frequent visits to brick-and-mortar stores. We investigated the effect of several aspects of online shopping on brick-and-mortar shopping frequency, including the frequency of online shopping, frequency of online-order deliveries, attitudes toward online shopping, and perceived issues arising from last-mile delivery logistics. Using a quasi-longitudinal survey and two machine learning-based models, we offer insights into how disruptions alter shopping behaviour and attitudes.