More than 10% of children in the age range 7–11 turn out to be poor text comprehenders: they have difficulties in reasoning about events of a story. TERENCE is the first adaptive learning system with stories and companion smart games, for reasoning about stories, and developed for primary–school poor comprehenders. The development of TERENCE followed both the user–centred and the evidence–based design methodologies, by placing users at the centre of the development process and by choosing a design because of evidence of its effectiveness for the users. This paper presents the TERENCE smart games: it explains how such methodologies were used for iteratively designing and evaluating incrementally improved versions of the TERENCE smart games.