The existing managerial problem of strategic and operational decision-making pertaining to the demand of promotional products is particularly difficult with the lack of direct historical sales information. Data from social media platforms are an underexploited resource in understanding the impact of user-generated data on product sales and subsequently, modelling the demand for these products. This paper investigates 1) the role that online user behaviours and preferences on social media have on the demand of promotional products and 2) distinct clusters of variables on modelling the demand data for promotional products, which include a variable selection approach based on machine learning, judgmental selection of variables and grouping variables according to their marketing objectives of Awareness, Engagement, Conversion and Consumer metrics. The analysis of a case study suggests that the metrics which represent the sentiments of users’ comments, yield demand models with the lowest error measures when compared to other metrics. Moreover, it is observed that combining metrics from various categories such as users’ awareness, engagement, and paid advertising yield more accurate demand models. The case study also demonstrated the effectiveness of judgmental selection of social media variables for demand modelling.