Promises That Deliver
Source: https://www.tue.nl/en/news-and-events/news-overview/13-03-2026-promises-that-deliver Parent: https://www.tue.nl/en/research
Share
How smarter time windows and routing can make next day and same day delivery work for customers and cities
Promises That Deliver
March 13, 2026
TU Eindhoven researcher Şifanur Çelik introduces models that make time window promises fair, reliable, and efficient in last mile logistics using optimization and deep learning.
Şifanur Çelik, photo by Vincent van den Hoogen
Şifanur Çelik of the Operations, Planning, Accounting and Control group defended her PhD on March 12, 2026. In this work, Çelik examined how delivery services should decide which time windows to promise and when to communicate them.
The research, carried out in the OPAC group at the Department of Industrial Engineering and Innovation Sciences, shows that time window decisions shape both customer experience and operational reality.
Why it matters\ Cities are grappling with congestion, emissions, and curb space pressure as e-commerce grows. Consumers expect narrow and reliable delivery windows, yet those promises ripple through fleets, depots, and neighborhoods. Companies must keep service convenient without pushing up cost, traffic, or driver stress. Çelik’s thesis examines how commitments can be set to support both reliability for households and workable routes for carriers.
Real pressures\ Retailers face fluctuating demand, uncertain travel times, and a constant stream of new orders while customers book. Offering a tight window too early can later limit routing choices. Waiting too long frustrates customers. The study looks at three common settings that firms must balance every day. These include next day planning with considering disruptions in the route, next day booking with later confirmation, and same day delivery where customers choose a slot in real time.
Earlier promises\ When time windows must be promised before travel time uncertainty is known, Çelik analyses the Time Window Assignment Vehicle Routing Problem with Stochastic Travel Times. The work introduces a two-step Benders decomposition with scenario clustering that can solve benchmark instances to optimality. Results indicate that carefully chosen flexible windows and route structure can absorb uncertainty without eroding service.
Smarter timing\ The thesis then studies a setting where providers delay communicating the exact time window until additional information is available. By modelling the decision as a semi Markov process, Çelik designs policies that anticipate downstream routing cost. The findings show that a considered delay can improve efficiency while treating customers fairly, helping operators avoid over promising during peak periods.
Customer choice\ For same day systems, slot design and routing are intertwined. Çelik proposes a transformer based model that learns which slot menu to offer while building feasible routes in real time. By capturing spatial and temporal patterns with attention network, the approach serves more customers with the same fleet size and improves service level compared with common heuristics, even when demand shifts across neighborhoods or time of day.
Societal value\ Better time window policies can reduce extra miles, cut idle curb time, and lessen failed deliveries, making streets calmer and cleaner. For entrepreneurs, the methods help grow delivery without runaway cost or driver churn. For municipalities, they support policy goals on congestion and emissions. For households, they raise the chance that promised visits arrive when expected.
What it shows\ Across three settings, the thesis presents a unified view of time window assignment as sequential decision making. It integrates stochastic optimization, anticipatory control, and machine learning to guide when and how commitments should be made. For the OPAC group and the department, the work illustrates how analytical tools can align fairness to customers with feasible daily operations.
Şifanur Çelik defended her thesis on March 12, 2026
- [### PhD candidate
Şifanur Çelik of the Department of Industrial Engineering and Innovation Sciences
Read more](https://www.linkedin.com/in/sifanur-celik/) - [### Dissertation title
From Stochastic Programming to Deep Learning: Advancing Time Window Assignment and Vehicle Routing in Last Mile Logistics
Read more](https://research.tue.nl/en/publications/from-stochastic-programming-to-deep-learning-advancing-time-windo/) - [### Supervisors
Tom van Woensel, Albert Schrotenboer, Layla Martin
Read more](https://www.tue.nl/en/research/researchers/tom-van-woensel/)
Marc Rosmalen