Text classification is an important research problem in many fields. We examine a special case of textual content namely, short text. Examples of short text appear in a number of contexts such as online reviews, chat messages, twitter feeds, etc. In this research, we examine short text for the purpose of classification in internet auctions. The “ask seller a question” forum of a large horizontal intermediary auction platform is used to conduct this research. We describe our approach to classification by examining various solution methods to the problem. The unsupervised K-Medoids clustering algorithm provides useful but limited insights into keywords extraction while the supervised Naïve Bayes algorithm successfully achieves on average, around 65% classification accuracy. We then present a score assigning approach to this issue which outperforms the other two methods. Finally, we discuss how our approach to short text classification can be used to analyse the effectiveness of internet auctions.