Calculation of Transport Cost for Freight Carriers on the Last Mile: Conducting a Case Study in the Municipality of Delft to Validate and Improve Usage of the Last-Mile Scan Calculation Model

Urban freight distribution problems vary in nature and can be characterized as economically, environmentally or socially grounded. Urban freight distribution operates in urban areas were space is in short supply. These residency areas are the homes of many people. Since most vehicles still drive on conventional energy sources, greenhouse gas emission is a big problem in cities. Safety and noise issues regarding old vehicles operating in urban areas also have a negative impact on liveability of cities. The social need to sustain urban freight distribution and to develop and improve the urban freight distribution system in high-density urban areas is large. The concept of sustainable developments can be divided in three principles, which include economic growth, social equity and environmental protection. The context of urban freight distribution is rather complex. This complexity is created by the multi- actor context of urban freight distribution and the fact that stakeholders do not always carry (social) cost and benefits both. Cargo owners and distributors are at the centre of the system and possess a high degree of power to influence the UFD system and therefore cannot be left out of the process. Overall, they have a neutral attitude towards sustaining urban freight distribution (which is the main goal of the problem owner, the municipality of Delft). Their attitude is neutral since they cannot earn money directly in operating in a sustainable manner. This makes it difficult to implement solutions, create sense of urgency to activate involved stakeholders to make a change. ‘Maatwerk Distributie’ (MD) has developed a Last Mile Scan (LMS). This is a calculation model that is developed to calculate and predict cost structures of last mile trips in urban areas. The main purpose of development is to gather information for both the municipality of Delft as well as participants of the LMS, which are freight carriers. The research objective is: to identify the needs of stakeholders involved in the urban freight distribution system of Delft and to conduct a validation study on the LMS calculation model to be able to provide recommendations to the municipality of Delft on how to improve usage and extent lifespan the LMS model in order to sustain UFD in Delft. This thesis seeks to answer the main research question: How can data and information collected by the LMS be of extended usage to the municipality of Delft to serve the purpose of improving urban freight distribution taking into account all stakeholder interests such as economic inefficiencies, (greenhouse) emissions, noise nuisance, air quality, safety and congestion issues? The conceptual framework developed by Sargent describes the process of how to get from one entity to another (R. G. Sargent, 2005). This conceptual framework is an approach that could be applied and help in building valid and credible simulation model. This framework is used as guidance in the validation process of the LMS model. In order to ensure the LMS is validated properly the four phases of the validation process of the framework are conducted. These phases are conceptual model validation, operational validation, computerized model verification and data validity. In addition, a sensitivity analysis is conducted. Sensitivity analysis can support validation of a model. The LMS model uses different variables and constants to estimate and give an indication of cost of a specific shipping. The black box and the CRD are conceptual models developed to conduct conceptual model validation. The relations between variables used by the model are logic. Although only the direction of the relation is evaluated at this stage, the LMS model seems to be fairly reasonable, but simplified representation of the real situation. Total distance of a trip is calculated using an API tool developed by Google. This application ensures a high accuracy of distances since it uses precise coordinates to calculate covered distances. The formula used for calculating distance per stop is plausible because distance per stop can be calculated by multiplying the number of stops times an average distance per stop. Total time is dependent on driving and unloading time. Driving time is dependent on average velocity of a vehicle and total distance covered. The general dimension of velocity is kilometre per hour; velocity = distance / time; time = distance / velocity. This relation is correct. Unloading time is dependent on the amount of small and large packages delivered multiplied by a constant ‘average unloading time’ per package/pallet. Unloading time can be estimated by using a constant average time for unloading small/large packages because every unit takes time to deliver. Total costs are calculated and dependent on total time multiplied by cost per hour and total distance multiplied by cost per kilometre. This is a plausible method. The formulas to calculate distance, time and cost of a last mile trip and relations between variables used to conduct these calculations are plausible. The benchmark methodology and strict rules that Research Company Panteia practise ensures data collection is done properly and is significantly reliable. Input data on assumptions like velocities of vehicles and cost per minute and kilometre seems correct. Data collected throughout field research is less reliable. It is difficult to state that these assumptions correspond with real world values, because background information on scope of the field research is not available and unknown. Constants are still assumptions and not valid as truth-values. Operational validity is impossible due to a lack of real data, which makes it impossible to crosscheck output of the LMS model with real data from the field. Freight carriers themselves do not know the cost of the last mile. This validation step cannot be conducted. In the sensitivity analysis, all output variables are higher in case of scenarios where constants are higher. This means that all hypotheses are correct and that the LMS model gives the expected results in light of sensitivity of changing different assumptions. This argument strengthens validity of the model because, the model generates realistic and logic outputs when varying input variables and constants. Overall, concluded from the validation process the LMS model is a reasonable representation of the problem entity and conceptualization and computerized model verification phases have been conducted properly. Data validity is not valid (yet) and should be improved on certain assumptions and the scope of the field research that supports these assumptions can be more extensive. Field research methods are basic and the actual size of the research is too small to gather a significant and crucial amount of reliable information and data. Operational validity is impossible due to a lack of real data and lack of knowledge of freight carriers on the exact cost of their last mile trip. Therefore, the LMS model cannot be labelled as valid because not all validation phases are successfully completed. The majority of trips, around 75%, cover a smaller distance then 10 kilometres. This means that the most last-mile trips do not cover a large distance. The average time of all trips is 100 minutes. This is relatively long since the average distance covered in 100 minutes is only 8,59 kilometres. The cost of all trips is distributed evenly between 0 and 100 euro. The centre of gravity lies between 10 and 35 euro. The average cost of all trips is 36,56 euro. All scatterplots of output variables cost, time and distance show expected relationships and results. The structure of the LMS model is linear based. All formulas between variables used in the LMS model are linear. Outcomes of graphs of the relationships between distance, time and cost show linear featured as well. These types of relationships are in line with expectations and therefore strengthen validity of the LMS model slightly. Interesting output variables that are not directly calculated by the LMS model yet, but can be created by applying a simple mutation of output variables of the data sample. These mutated variables can be seen as criteria or key performance indicators to see how different freight carriers are performing. KPIs that are explored in this paragraph are cost per unit; cost per minute; cost per kilometre; cost per m3. The trend line of total cost of PostNL is a significant steeper linear relationship as the trend line of total cost LMS. This implies that as shipping load increases (packages + pallets) alternative shipping method PostNL is more expensive in comparison to a shipment conducted by a freight carrier internal. The intersection is interesting and lies at a shipping load of 5 units (packages + pallets). In the majority of cases (data points) a shipping load of 5 or smaller (relatively smaller shipments) are cheaper to be shipped by an alternative shipping method (PostNL). In majority of cases, larger loading ratios (relatively larger shipments) benefit to be executed by a freight carrier internally. Distributors can be classified based on all parameters in the LMS model or parameters not yet present but easily created by a simple mutation. The possibilities are endless since all attributes and KPI’s can be used as inputs and/or evaluation fields both. The right setting for the municipality is probably a combination of economic and environmental parameters. A conclusion drawn from the cluster analysis is that clusters with a large shipping load score well on environmental aspects (defined environmental key performance indicators). In addition, these groups of carriers also score well on economic aspects (defined specific economic key performance indicators) concluded from the cluster analysis on economic aspects. This strengthens the conclusion that smaller shipments perform less good in general (both in economic and environmental terms) and can better use an altern