Previous research into drinking behaviour has often focused exclusively on either alcoholic drinks, sugar-sweetened beverages, or water, usually with an emphasis on changing consumption. The primary aim of the present study was to develop and assess a situated framework of drinking behaviour that predicts the frequency of drinking alcoholic as well as non-alcoholic beverages. Within this situated framework, we originally expected that predictive patterns for drinking frequency would vary greatly for different drinks as well as for different individuals. To assess these predictive patterns, we used the Situated Assessment Method (SAM2), which has previously been used successfully to study other health domains such as stress and habits. To cover a broad range of drinking situations, we sampled 4 alcoholic beverages (beer, wine, cocktails, spirits) and 7 non-alcoholic beverages (tea, coffee, fruit juice, diet soft drink, regular soft drink, bottled water, tap water). We then used the Situated Action Cycle to identify 34 potential predictors of drinking frequency (e.g., liking the taste, thirst, socialising). A total of 900 UK participants completed an online survey, rating each of the 11 drinks on consumption frequency and on each of the 34 predictors. To reduce complexity and remove redundancies, we performed an exploratory factor analysis, decreasing the number of predictors to six factors that we interpreted as habit, craving/regulation, negative consequences, health/functionality, and socialising/positive consequences. Across analyses, our six-factor model was able to explain considerable variance in self-reported drink consumption, both for individual drinks (median = 66%) and individual participants (median = 94%). Surprisingly, we only detected subtle differences between the eleven different drinks’ predictive patterns. In contrast, individual participants exhibited considerable variability in the factors that predicted their drinking behaviour.