Students Project & Theses

In 2017, some research progress has been made by students from different universities in form of projects or graduation theses which have proven to be of value for the research group. The following table lists these student contributions.

Evaluation of Delivery Cost Approximation Methods for Attended Home Deliveries

U. Wachsmuth, 09.11.17, Freie Universität Berlin

Attended home deliveries are delivery agreements where the customer picks a day and time window from a list. The delivery agreement is often made online, e.g. in the context of online grocery shopping. The company guarantees to make delivery during that time period. This service is offered by more and more businesses because it can constitute a competitive advantage. Customers appreciate that deliveries are not deposited at the neighbor’s or have to be picked up at the post office. However, companies offering attended home deliveries face high costs and logistic challenges because the time windows render delivery routes less efficient. The calculation of the cheapest delivery tour is performed by solving the NP-hard Vehicle Routing Problem with Time Windows (VRPTW). The online nvironment forces the company to provide the customer with instant feedback about the availability and price of each time window. This feedback is provided by customer acceptance mechanisms which perform feasibility and profitability checks. Thereby, the mechanisms determine which time windows would lead to “acceptable“ agreements. Usually, the mechanisms are based on the solution of the VRPTW. Realistic VRPTW are so complex that their precise calculation would take too long for an online setting. Cost approximation methods are a means to provide sufficiently accurate information in less time. We present different approaches and evaluate their advantages and disadvantages. By means of a simulation study, we test a novel approach to identify and reject customers who will cause a disproportionate increase of the average costs. The results suggest that it is possible to identify unprofitable regions which should not be served with time window guarantee. However, there is also evidence that the exclusion of customers can produce an adverse side effect. It limits the savings potential of economies of scale. Therefore, it should be used with caution.

Impact of environmental determinants in predictive modeling on forecasting purchase probabilities

E. Schweihofer, 07.12.17, Freie Universität Berlin

The present master thesis investigates the hypothesis that environmental determinants constitute suitable contextual factors in the field of product-related purchase decision-making processes in order to increase the accuracy of predictive models. The framework for this study is provided by the discussion about the impact of the environmental factors weather and holidays on the ordering behavior of customers of the online supermarket AllyouneedFresh. The thesis applies a design oriented research paradigm with the objective to construct an IT artifact. This methodology is supported by a systematic literature review. The IT artifact consists of a component for providing contextual factors and a data-driven forecasting model which is developed with support of the CRISP-DM process. The evaluation’s outcome on selected records indicates a significant contextual variance related to the type of the applied machine learning algorithm and the target variable’s shape. Furthermore the influence of the product category's context affinity to the forecast quality as well the importance of single contextual variables in the predictive model is demonstrated.

Developing an Adaptive Large Neighborhood Search for Route Planning in Home Delivery Services with Time Windows

J. Haferkamp, 08.12.17, Freie Universität Berlin

The online grocery market is of increasing importance. Within a very complex logistic process, the so called last-mile problem stands for a cost-intensive and critical phase of order fulfillment in online food retailing. Coping with the last-mile problem requires the effective and efficient realization of the e-fulfillment process. In this process which works as the basis of this thesis, the Vehicle Routing Problem with Time Windows (VRPTW) has to be solved in order to plan the delivery and to determine delivery capacities. The present thesis focuses on the development of an Adaptive Large Neighborhood Search (ALNS) in order to solve the VRPTW. Furthermore, procedures for the use of the ALNS in the e-fulfillment process are modelled. The evaluation of the developments is achieved via experimental computer simulations. These experiments also investigate the impact of different conditions on the considered e-fulfillment process steps. The results of the experiments show that the developed ALNS can be used successfully during the e-fulfillment process to minimize vehicles and travel times as well as to maximize delivery capacities. The experiments also demonstrate the importance to control the demand ime window within the e-fulfillment process in order to realize existing delivery capacities.

Online Supermarkets – Comparison of Business Models and Strategies for Attended Home Deliveries

P. Roczniak, 29.07.17, European University Viadrina Frankfurt (Oder)

The e-grocery sector has evolved significantly since its beginning in the 1980s. Many of the first e-grocers were pushed out of business due to the very high operational and investment costs, related to picking and attended home deliveries. Focusing on achieving operational efficiency, they neglected the needs and expectations of the customers. The promise of online channel becoming the main one for grocery shopping has not been fulfilled. However, current market trends suggest that there is a lot of potential that is still to be exploited. Companies nowadays apply different strategies to attract and keep the customers. It is sometimes the nuances that make the market leader. This Thesis analyses the business models employed in three culturally diverse countries – Germany, Poland and China. It compares the solutions concentrating on fulfilling customer expectations and addressing the operational issues. Significant differences between European and Chinese companies are spotted mostly with regard to product offering and infrastructure aspects, that may be caused among others by different e-commerce maturity, local business environment and local customers’ expectations. The e-grocery market is a very dynamic one, where new entrants can still present a threat to established players. The companies to stay competitive have to constantly monitor market needs as well as benchmark their solutions with competitors. It will be an especially interesting sector to follow as digital native generations are entering the adulthood and could cause a isruption with their readiness to buy online and try new innovations. Moreover, new markets are opening up for online grocery, such as India, Latin America or Africa, where local characteristics can introduce highly novel perspective on e-grocery. This Paper focused on three markets, however further research could incorporate more geographical locations to gain a broader understanding of the e-grocery sector around the world.