|
Abstract:
|
As Amazon continue to grow, it is critical that Last Mile science and technologies plan routes more intelligently. Accurately predicting the planned duration of deliveries is important for maintaining a high level of customer satisfaction as well as protecting the driver experience and productivity. Underestimated delivery time will force drivers to work longer resulting in missing/delayed packages, driver fatigue, safety issues and leading to risks in both driver and customer experience. Conversely, overestimated delivery time means lower number of deliveries per paid hour, negatively impacting the company’s financials. Time Per Stop (TPS) combines service and transit times for the packages delivered at the stop. TPS model forecasts productivity for a driver delivering to a certain area and guides the planner on the amount of work which can be allocated to a driver in the given neighborhood. It is developed to reduce the Plan-vs-Actual (PvA) error and generate more accurate predictions by using estimators which are less volatile. The current pilot results show that TPS improved route estimation accuracy and lowered workload concerns compared to previous approach.
|
