Considering customer self-pick-up, home delivery with time window and flexible service demands in the last mile delivery, a location routing problem driven by multiple personalized demands is proposed to improve the service level. To solve the problem effectively, a variable neighborhood search algorithm is designed, which combines an initial solution construction algorithm, a differential neighborhood usage strategy and an adaptive shaking mechanism. Experimental results based on different scales of instances show that the improved variable neighborhood search algorithm has better solving efficiency and robustness. Through the sensitivity analysis of the key parameters, it is found that the proportion of personalized demands and pick-up cost have significant impacts on operating cost. It is of great practical significance to comprehensively consider these factors when designing the last mile delivery system.