Despite the postulated importance of analogising to human cognition, the study of analogical problem solving in the laboratory has found disappointing results. Providing an analogue to a participant prior to asking them to solve a problem gives only a small benefit at best. Recently, studies outside the laboratory have suggested that experts frequently use analogies in realworld situations. It is less clear whether novices can also spontaneously invoke and use analogies to solve realistic problems. In the current investigation, undergraduates were observed solving a large-scale management problem over two weeks. It was found that many analogies were produced (on average 4.6 per one-hour session), and that 77% of these analogies reflected a structural rather than a superficial mapping between a base and a target. It was also determined that 56% of these structural analogies involved non-trivial mappings of higher-order relations. Further, it was found that analogies were drawn to serve two different purposes: problem solving and illustration. In generating illustrative analogies, participants frequently made superficial mappings, but when generating analogies to solve problems, they never made purely superficial mappings.