Engineering designers seek to explore ‘real’ problems that must be solved across design processes. This exploration might be challenging in complex problem situations. An effective way of encouraging design exploration is conjecture-based problem exploration—informing problem re-interpretation by potential solutions. However, little evidence indicated how this process unfolds, especially in complex problem situations. This study addresses this question by articulating the underlying cognitive mechanism of conjecture-based problem exploration. Situated in a creative design practice that tackles real-world, complex problem situations, we employ grounded theory to conduct qualitative coding of interview transcripts and documents elicited from ten multidisciplinary graduate students. We developed a three-phase process model to explain conjecture-based problem exploration: (1) triggering through analogizing, inspiring, evaluating, and questioning; (2) transitioning to problem space expansion; and (3) resulting in problem focus adjustment incrementally or radically. Our explanation contributes to design theory building and encourages engineering designers to embrace a dynamic view of design problems when addressing complexity.