This study analyzes the effect of information overlap in groups discussing a complex problem on individual post-discussion complex problem solving (CPS). We hypothesize that information distribution among group members has an inverse u-shaped effect on individual post-discussion performance, favoring groups with a medium informational heterogeneity. As CPS is presumably correlated with experience, we also assume that exposure to the problem before the actual task leads to higher performance than less or no exposure. Experimental results support the first hypothesis: A medium overlap of instructional text paragraphs in dyads led to higher performance in a computer-simulated complex problem than complete or no overlap. The second hypothesis is not supported. Limitations of the study and practical implications are discussed.