This study aims to investigate and develop the efficiency of a decentralised dynamic system for last-mile parcel distribution. A better understanding of the efficiency, and thus the feasibility of this decentralised logistics method, contributes to a better understanding of the potential of these systems as alternatives to the current centrally organised system. The main research question is: To what extent can the efficiency of a decentralised dynamic parcel distribution method be improved by applying different heuristics and adapting the simulation environment with congestion and a mixed fleet? To answer this question, a review is made of existing literature on last-mile logistics, self-organisation in logistics and auction methods in logistics. The current model is evaluated, benchmarked and validated. Model improvements are defined and applied in an experimental setting using a small dataset, after which the extended model is applied to a real-scale case dataset based on 729 delivery addresses in Delft. The applied modification strategies are random insertion, parcel swapping and k-means clustering. To account for the behaviour of the model under different conditions, the method was tested for a heterogeneous fleet of vans and cargo bikes and different levels of congestion. In general, the heuristics were able to improve the performance of the model, but the improvements in travel distance were minimal. When the combination of the best performing heuristics is applied to the case data, a 4.6% improvement in vehicle travel distance can be achieved compared to the original method, but still with high computation times. Comparison with an external centrally organised OR solver shows that although the decentralised method can be improved, it does not perform well, as the OR solver can find a better solution in a fraction of the time needed for the improved decentralised auction method. This study provides a proof of concept of the method and a more efficient model has been constructed. However, given the limitations of the method, further research should carefully consider whether it makes sense to continue with this particular method.