The heterogeneous fleet vehicle routing problem with light loads and overtime: Formulation and population variable neighbourhood search with adaptive memory.

•VNS-triggered memory extraction improves method performance up to 5.2%.•Incorporating real life aspects could improve daily total routing cost up to 8%.•Vehicle capacity and working time utilization could be improved by up to 12.5%.•Real life aspects could improve fleet composition at no extra cost.•Interesting managerial insights regarding real life routing trade-offs.