How modelling supply and demand interactions improve services like Thuisbezorgd.nl
On-demand systems (think of Uber or Thuisbezorgd.nl) provide consumers with real-time access to mobility services, as well as services that transport goods. Thanks to the advancements in information technology, these on-demand services can combine real-time information, trip planning, and booking to offer requested services at competitive prices. Operations research applies mathematical analysis to solve problems and optimise processes within businesses like in this case, on-demand service providers.
Operations research and the theory of discrete choice models
With on-demand services, the benefit from the demand side is that consumers have access to customised options and an efficient and integrated transportation network. Customers are provided with the comfort of personalised and fast services at their associated affordable price offered by these operators. On the supply side, companies can benefit from a system that better uses available resources and is profitable by accommodating a larger number of users and offering personalised options. With operations research and the theory of discrete choice models we can achieve this goal.
There are relevant examples for both on-demand mobility and city logistics in the Netherlands. For instance, Mobility Mixx is a company that offers a range of mobility services, including public transit, rental cars, carpooling, bike sharing, and taxis. This company provides tailored services to travellers and offers payment and trip scheduling options. Also, Thuisbezorgd.nl provides customised food delivery services where the delivery options are tailored according to the requests of customers.
Even though demand and supply closely interact in such service companies, these two research fields have evolved independently, without paying too much attention to the existing interdependencies between the two. Indeed, incorporating the preferences and tastes of customers, which are usually characterised with discrete choice models, allows for a better planning of the systems for the operators.
The design and organisation of such systems are typically addressed with optimisation models. Together with my colleagues at EPFL in Switzerland, we propose a general framework that allows integration of behavioural models with optimisation. To illustrate how the framework can be used, an application on revenue maximisation is considered. The performed experiments show that the resulting formulation is a powerful tool to plan systems based on differences in customer behaviour.
In a separate project with my colleagues at MIT, TU Delft, and EPFL, we have proposed an algorithmic framework to optimise demand mobility services. Each time a new customer arrives in the system, a customised travel service is offered. We show that with the help of assortment optimisation and dynamic pricing, the service company can better use the available vehicles and as such reduce operational cost while complying with the preferences of the customers.
'Forty-five percent of consumers will not choose the service if punctual delivery service is not provided'
I have been working on developing mathematical models and algorithms for user-centric last mile delivery applications like Thuisbezorgd.nl. The operation of on-demand logistics is an expensive task which cannot benefit from the economy of scale. In addition, such systems operate in a highly uncertain environment in which demand must be responded to in real-time. Failing to satisfy this demand leads to profit loss. Forty-five percent of consumers will not choose the service if punctual delivery service is not provided. This is a big challenge for such companies. I have started collaborating with TU Delft, Thuisbezorgd.nl, and PostNL within a Dutch Research Council research fund to introduce new models to tackle this problem using their data.
Thuisbezorgd.nl is an on-demand logistics provider which operates in the food industry. Our developed approach uses real-time data and provides a planning approach for its operation. Our second case focuses on providing home healthcare supply which is provided by healthcare organisations. PostNL recently established a logistic service to provide medical devices from pharmacies and medical centres to patients. Our proposed optimisation algorithms aim at improving the services for both companies, taking behavioural aspects of consumers (in the case of Takeaway.com/Thuisbezorgd.nl) and patients (in the case of PostNL) into account.