Many planning tasks involve complex reasoning about time: what must happen in sequence and what may happen in parallel. One hundred ten online participants were provided with a simple planning scenario (to design a calling tree) and asked to manipulate different diagrammatic representations of the problem. More important than the initial representation was the participants’ transformed representations: if time was encoded in the lengths of tree links then inference was more accurate. This finding suggests that diagram transformation may be a useful way to elicit representation strategies, and that such transformations from different starting conditions may be useful as diagnostics and as design aids.