The Chair of Analytics & Optimization conducts research which deals with decision problems arising in the areas of pricing and revenue management as well as last-mile logistics and can be solved using operations research.
Our goal is to build simple, practical, and realistic models that allow getting model input directly from the data. By designing computationally efficient methods that are both easy and scalable, we strive to generate interesting insights and bring significant value to practice.
Regularly, we cooperate with industry partners like, amongst others, GLS, IBM, Sixt, TUI, and ZeppelinRental.
Modeling preferences of market participants to capture the effects of choice behavior in various optimization models constitutes the foundation of all research activities at the chair. The actors mainly considered are customers requesting services, e.g., flights or deliveries, but in some settings also service providers like occasional drivers in a crowdshipping system.
To segment these actors depending on their choice behavior and to specify mathematical models accordingly, different methods of data mining and machine learning are employed (e.g., clustering or preference learning). Depending on the research question, either static, deterministic or dynamic, stochastic models are developed. In the case standard software cannot be applied efficiently, we also design specialized optimization algorithms. On the one hand, this includes state-of-the-art heuristics, e.g., evolutionary algorithms and adaptive large neighborhood search, on the other hand, algorithms based on approximate dynamic programming. The latter enable real-time decision support using artificial intelligence.
Key Fields of Application
Pricing & Revenue Management
Revenue management comprises concepts and methods for the optimal configuration of sales processes of services and is applied by, e.g., airlines, car rental companies or hotels. Generally, the availability of products defined on services offered is controlled over time, so that demand can be adjusted to match the remaining levels of capacity. In contrast, the closely related concept of dynamic pricing aims to steer demand by varying prices over time.
The number of people who take up residence in cities is rising steadily. In connection with the increasing importance of e-commerce, this leads to an exponential growth of delivery quantities and thereby to a larger environmental impact. In this context, the term last-mile logistics describes the design and optimization of transportation activities of private logistics providers in urban areas. Considering the existing infrastructure, environmental aspects like air and noise pollution, but also the changing consumer behavior is an important part of last-mile logistics.
Research in this application area includes the following questions:
- When should an airline offer which tickets at which price? How can suitable concepts be transfered to car rental companies, hotels or digital service providers?
- Should car rental companies and hotels offer upgrades? If so, who should recieve these offers and how does that affect the availability of certain products?
- How can optimal pricing of (season) tickets for sports events and concerts be achieved?
- When and to what extent should prices of seasonal goods be reduced? When should additional discount campaigns be launched? How should these campaigns be designed?
- Is it profitable to offer flexible products, that enable the seller to decide on the exact service provision after the sale has been completed (e.g., the itinerary of a cruise)?
- How can ancillary services, e.g., internet access during flights, increase profits? When should which customers be offered such services?
Research in this application area includes the following questions:
- How can the use of different modes of transportation ranging from vans to cargo bikes or drones contribute to the reduction of damage to the environment?
- Where should so-called micro-depots be located, in which the transshipment of goods can take place? Which routes should different vehicles be assigned on a certain delivery day?
- Which opportunities for the generation of revenues or the reduction of costs arise for parcel service providers by offering different delivery options, e.g., home delivery, in-car delivery or delivery to automated parcel terminals?
- How should time windows chosen by the customer in the field of attended home delivery be priced in order to facilitate the construction of efficient delivery routes?
- Which opportunities exist to integrate crowd-sourced delivery into a transportation system?
Lufthansa Systems AG - Forecast of Passenger Flows
In many markets, the liberalization of air transportation has lead to a situation where competitors with greatly differing business models are facing each other. Therefore the development of a tool for forecasting passenger flows, which explicitly accounts for the higher price sensitivity of business and leisure customers, became necessary. Especially endogeneities within the data set had to be addressed. These originate from the influence of the respective competitor's revenue management on the observed prices leading to a mutual interdependency of price and demand.
Sixt AG - Capacity Control in Car Rental Revenue Management
In the course of this project, revenue management methods accounting for specific requirements of the car rental business were developed. These include, e.g., the use of upgrades and the possibility of capacity adjustments at different points in time during the planning horizon. Additionally a simulation environment was implemented, enabling the uniform evaluation of all methods in the specific planning environment of the partner company.
TUI AG, IBM Deutschland GmbH - List Price Optimization in the Tourism Industry
The travel catalogue is still the most important distribution channel for many travel companies. On behalf of IBM, a customer oriented pricing system was developed for the TUI AG, that incorporates preferences of individual customers as well as the specifics of this distribution channel. To conduct a segmentation of customers based on transactional data, methods of data mining were used. The results of this step served as an input for mathematical optimization of prices. In this way, an optimal pricing system considering various constraints could be generated.
Zeppelin Rental GmbH & Co. KG - Pricing in the Equipment Rental Industry
Dynamic markets like the equipment rental market are highly affected by economic cycles. Therefore besides asset management, dynamic pricing which is adapted to the general market behaviour is key. In the course of this pricing project with Zeppelin Rental, a process was defined which even more places the focus on the customer and aims at finding the best prices for each customer. Customer segments were derived using statistical methods and data mining to be able to analyze their requirements and willingness-to-pay. The results of this study served as a foundation for the enhancement of pricing practices within Zeppelin Rental.
Daimler AG - Modeling of Customer Choice Behaviour
In rapidly changing, highly competitive market environments, the explicit modeling of customer choice behavior to optimize the products and services offered is gaining importance steadily. In this light, different discrete choice models, including logit, (cross-) nested logit, probit and mixed logit, were evaluated. The application in practice, together with the efficient development of suitable model specifications, were at the project's core. Conclusively, possible fields of application within the automotive industry were identified.
GLS Paketdienst GmbH & Co. OHG - Optimal Design of Delivery Options
The amount of packages delivered in Germany is increasing constantly. As a result, home delivery is complemented by additional delivery options, especially in the B2C-sector. These include delivery to parcel shops and automated parcel terminals. Offering these options in the course of a targeted demand management not only leads to a reduction of transportation costs, but also of damage to the environment. The project aims at formulating optimization models for the economical assessment of different delivery options. Additionally, due to the resulting problem complexity, heuristics for their solution are developed.