The e-commerce industry is constantly flourishing and has a strong momentum of its own. However, behind the prosperity, many e-commerce platforms and merchants are facing challenges brought by the e-commerce business model. As part of the e-commerce sales process, determining the fulfillment time of orders is a headache for many retailers. If delivery is too slow, it often leads to consumer dissatisfaction and complaints, which can affect the store’s overall positive feedback and repurchase rate. If delivery is too fast, this can lead to a decrease in the retailer’s profit margin as some customers place multiple orders in a short time. If the retailer can consolidate these orders, it can effectively reduce logistics costs.
Recently, CHEN Shouchang, a researcher at the Department of Data Science and Management Engineering, School of Management, Zhejiang University, together with Professor YAN Zhenzhen from Nanyang Technological University and Professor LIN Yunfeng from Singapore Management University, developed a new order fulfillment model to help sellers determine the optimal delivery time and reduce costs.
Click here to access the research paper
Their research achievement "Managing The Personalized Order Holding Problem in Online Retailing" was successfully published in the top international journal Manufacturing & Service Operations Management and was awarded "The Finalist of the 2021 MSOM Data Driven Challenge".
|
CHEN Shouchang | 陈寿长 School of Management, Zhejiang University |
||
|
|
||
|
|
|
Academic Background: Dr. Chen is a ZJU100 Young Professor (Assistant Professor) at the School of Management, Zhejiang University. He obtained his Ph.D. from Zhejiang University (2013.9-2019.7). Before joining Zhejiang University, he worked at Nanyang Technological University as a Research Fellow (2019.11-2021.10). His research interests include Data-Driven Decision-Making, especially for online retailing platforms and Production and inventory Management. You can learn more about Assist. Prof. CHEN Shouchang’s academic background here |
|
Does all the money earned go to the logistics company? – Researchers found that "order fulfillment" has a big impact on profit margins |
In online shopping, the entire process of "order fulfillment" includes not only the order by the consumer, but also the allocation of the order warehouse, the sorting of the order, the packaging and finally the delivery to the consumer.
In 2018, Amazon spent around 30 billion dollars on transportation costs, but only generated net sales of around 10 billion dollars. Between 2017 and 2019, JD’s order fulfillment costs accounted for about 7% of JD’s net sales.
Allocating stock orders, sorting and packing orders, consolidating or splitting goods These challenges in e-commerce have attracted a lot of attention from both the industry and academia. CHEN Shouchang and his team have also been working on this topic. After observing a large number of cases, CHEN Shouchang’s team found that the efficiency of the "order fulfillment" process has a great impact on the profit margin of merchants and that reducing the cost of order fulfillment plays a crucial role in e-commerce platforms. They proposed an idea: "Can we determine an optimal order fulfillment strategy to help e-commerce merchants choose the most appropriate time for order fulfillment?"
Since May 2020, after three years of intensive research, CHEN Shouchang’s team has successfully developed a new order fulfillment model and constructed an optimization algorithm to help online merchants achieve a balance between "cost reduction" and "delivery responsiveness" by using this model.
|
Predicting Customers’ Reordering Behavior, Analyzing Different User Profiles and Exploring Four Models and Algorithms |
CHEN Shouchang’s team explained:
|
" |
Our goal is to develop a prediction tool that is both accurate and easy to explain. By using an intelligent order fulfillment system, online retailers can meet the needs of their customers while reducing their fulfillment costs." |
What optimization programs are included in this new order fulfillment model? How do they help companies make the right order fulfillment decisions?
|
01 |
Sequential decision model CHEN Shouchang’s team proposed a "sequential decision model" in their research to predict the subsequent choice behavior of customers belonging to different user profiles when placing an order - i.e. reorder, continue searching and abandon. The sequential decision model helps quantify the trade-off problem of delayed orders, and online retail platforms can use this model to develop personalized order delay strategies for customers. |
|
02 |
A high-dimensional Markov decision model Based on the premise of quantifying and balancing the delayed order problem using the "sequential decision model", the research team modeled the delayed order problem as a high-dimensional Markov decision model and solved the "curse of dimensionality" problem of this high-dimensional Markov decision model using its special structure. |
|
03 |
Speculations on the structure of the optimal strategy to delay ordering Furthermore, this study characterizes the structure of the optimal order delay strategy, namely the personalized threshold strategy. In e-commerce orders, customers belonging to different user profiles also have different personalized delay time thresholds. This threshold can be used to estimate the customer’s delayed order time. This means that if the customer does not place any new orders within this threshold time, there will be no further order delays. The figure below shows that users belonging to different portraits also have significant differences in their personalized thresholds. The personalized threshold for most users is less than 10 minutes, which means that many customers can control the process of order delay within 10 minutes after placing an order. |
Personalized threshold distribution
|
04 |
An analytical expression for personalized thresholds and its visual representation In this study, the team proposed an analytical expression for personalized thresholds in terms of user profiles, which can calculate personalized thresholds more efficiently. In addition, the research team visualized the personalized thresholds using a segmented linear approximation method and summarized the management’s findings. The research showed that the personalization threshold is higher for consumers who are more likely to place orders continuously within a short period of time. For example, merchants should set longer order delay times for business users, members, female users and users from first-tier cities. |
|
An Innovative Research Model Based on the Motto "Small But Powerful" Effectively Saves Costs for E-commerce Platforms and Merchants |
|
■ |
Academic contributions |
Unlike previous research, CHEN Shouchang’s team uses different models and reserves the possibility for practitioners and researchers in other fields to choose different theoretical algorithms.
First, in order to improve the interpretability of the model and the optimization algorithm, this study did not use state-of-the-art machine learning prediction algorithms such as deep learning methods, but innovatively used the "small and beautiful" "sequential choice behavior model" from the field of sales management.
Second, the data-driven deferred ordering model also provides an interface for other prediction methods, laying the foundation for multidimensional research usability. The purpose of this interface is to allow engineers to continue to freely choose the prediction methods they trust or are accustomed to in practice.
|
■ |
Practical value |
1) Has reference value for most e-commerce merchants
The study used big data from JD.com to comprehensively test and evaluate personalized order delay strategies with thresholds. The data includes about 30000 stock receipts and issues (SKUs), 450000 customers, 480000 orders and 20 million click records. Among these, the extensive click data plays a crucial role in analyzing and estimating the length of time customers spend on the platform.
The study fully demonstrates the support of big data in model optimization, and the use of mass data means that this study is applicable and effective for most merchants to predict order inventory.
Data
2) Help e-commerce platforms reduce order fulfillment costs
According to calculations, compared with other strategies, the personalized order delay strategy proposed by the research team can help e-commerce platforms save about 1.2% of order fulfillment costs when orders are delayed by 12 minutes (based on JD’s annual order volume in 2016, the personalized threshold strategy can save 20 million in order fulfillment costs). According to data from users who frequently place orders, personalized delayed order strategies can bring significant cost savings to merchants and platforms.
Performance of the personalized delayed order strategy
E-commerce, as an important industry in the Internet era, is changing the ecosystem in business and society at an unprecedented pace and scale.
The research of CHEN Shouchang’s team closely links the current development status of e-commerce enterprises with the actual problems of the industry, and provides merchants with new solutions for order decision-making through scientific research. The contribution of management know-how to the challenges faced by e-commerce platforms is of great academic and practical value.
- We are pleased that our scholars have taken action to enable innovation and breakthroughs in industry through digital intelligence and management to promote economic and social development.
- You can read the original article in Chinese here
