The space syntax Theory of Natural Movement postulates that everything else being equal, land use selects their location based on the asymmetry of accessibility created by the configuration of the street network. In this article, I test the hypothesis whether configurational (syntactic) properties of an urban street network are relevant for the location of land use. If syntactic features are relevant, then land use types may have a ‘syntactic signature’. To identify this blueprint, I apply machine-learning techniques to datasets of syntactic measures, for ten business types. The results are ten models, each with the syntactic blueprint of a type of business. The models are used to predict the existence, or not, of such businesses in segments of the map of London. The performance of the models varies, with fifty per cent reaching statistical significance, including one with ‘good’ prediction ability. The models for Starbucks coffee shops and solicitor’s offices have the strongest prediction ability. The exploratory exercise demonstrates the potential of the machine-learning method Random Forests, when supervised and individually applied to a business activity, to identify the syntactic signature of business types. Such models can be used during planning and design and on location studies. The results strengthen the candidacy of syntactic measurements to location-decision-making. Moreover, they reinforce the theory of Natural Movement.