Although diabetic polyneuropathy (DPN) has a very high prevalence among people with diabetes, gait analysis using cyclograms is very limited, and cyclogram research, in general, is limited to standard measures available in software packages. In this study, cyclograms (movements of the centre of pressure, COP, on and between the plantar surfaces) of diabetics and healthy individuals recorded with a smart insole were compared in terms of geometry and balance index, BI. The latter was calculated as the summed product of standard deviations of cyclogram markers, i.e., start/end points, turning points, and intersection points of the COP. The geometry was assessed by the positions of, and distances between, these points, and the distance ratios (14 parameters in total). The BI of healthy and diabetic individuals differed significantly. Of the fifteen parameters (including the BI), three were suitable as classifiers to predict DPN, namely two distances and their ratio, with false negatives ranging from 1.8 to 12.5%, and false positives ranging from 2.9 to 7.1%. The standard metric of the cyclogram provided by the software packages failed as a classifier. While the BI captures both DPN-related balance and other balance disorders, the changing geometry of the cyclogram in diabetics appears to be DPN-specific.