EURO 2025 Leeds
Abstract Submission

2658. Predictive and optimization-based package consolidation for efficient e-commerce fulfillment at Trendyol

Invited abstract in session TB-38: Forecasting, prediction and optimization 2, stream Data Science meets Optimization.

Tuesday, 10:30-12:00
Room: Michael Sadler LG19

Authors (first author is the speaker)

1. Mutlu Soruklu
Data Science, Trendyol
2. Betül Ahat
Data Science, Trendyol
3. MERT PARÇAOĞLU
4. Fikret Batu Sagay
Data Science - Core Ops, Trendyol
5. Ahmet Çınar
Data Science, Trendyol
6. Meltem Sanisoğlu

Abstract

In traditional e-commerce fulfillment, each item is transferred individually without consolidation, leading to high shipping costs and inefficient cargo movement. This study proposes a hybrid model that integrates predictive analytics and optimization to improve package consolidation efficiency while ensuring timely delivery. The objective is to consolidate shipments up to a certain deci and price limit, thereby reducing overall logistics costs, holding costs, and partial shipment rates while improving lead times, particularly for shipments from Turkiye to international destinations.
The model predicts estimated delivery times to consolidation centers and identifies shipments that might exceed delivery thresholds, preventing consolidation delays and customer dissatisfaction. By analyzing multiple factors of individual shipments such as delivery time, deci and price, the model optimally creates consolidated packages while filtering out delayed shipments. This approach enhances fulfillment efficiency, minimizes unnecessary cargo movement, and optimizes customer experience by reducing partial shipments and improving international lead times. Empirical results demonstrate that prediction-driven package consolidation significantly reduces shipment costs by more than 50% when high customs fees are considered, decreases customer lead times of affected consolidated packages by approximately 25% by filtering out delayed shipments, and ensures a more reliable and scalable fulfillment

Keywords

Status: accepted

Back to the list of papers