Following recent studies of visual attractiveness in vehicle routing, we investigate the inclusion of shape and compactness penalties in computing solutions to the Vehicle Routing Problem using the Adaptive Large Neighbourhood Search. Visually attractive routes are sought predominantly for two reasons. First, operators are reluctant to implement solutions that exhibit overlapping routes, or unacceptable shape. Second, the visual compactness of routes is indicative of the operational robustness of solutions. We are the first to investigate the concept of bending energy as a solution penalty in this setting. We are also the first to investigate a search that leverages the geographic center of every route encountered during search.