Research on the Development of Business Model of Food Delivery

Platform Enabled by Technological Innovation: Taking Meituan

Waimai as an Example

Sa Zhang

School of Economics and Management, Beijing Jiaotong University,

Weihai Campus, Modern Road, Wendeng District, China

Keywords: Technology Empowerment, Business Model Innovation, Food Delivery Platforms, Meituan Waimai,

Network Effects.

Abstract: With the rapid development of digital technology, the food delivery platforms are undergoing significant

transformations driven by technological innovations, which promote the transformation of intermediary

models and enhance network effects and platform properties. This article examines how technology empowers

business model innovation in food delivery platforms, focusing on Meituan Waimai as a case study. Under

the vital role of artificial intelligence, big data, and automation in optimizing platform operations, business

model of Meituan Waimai has evolved evolves from a multi-dimensional value chain model to its current

"retail + technology" approach. The study employs a mixed-methods approach, combining qualitative analysis

of Meituan's strategic shifts with quantitative data on user engagement, order fulfillment, and technological

adoption. Findings reveal that Meituan's integration of smart logistics, AI-driven recommendations, and

unmanned delivery systems has strengthened its network effects, expanded service boundaries, and improved

value co-creation among stakeholders. The research contributes to understanding how digital platforms

leverage technology to sustain competitive advantage and reshape industry standards, offering insights for

practitioners and policymakers in the digital economy.

1 INTRODUCTION

In recent years, the rapid development of digital

technology has profoundly reshaped the business

model of the food delivery industry. Platform

companies represented by Meituan Waimai have

achieved a transformation from information

intermediaries to intelligent decision-making centers

through artificial intelligence, big data analysis and

automation technology. This technology-driven

transformation not only improves order-matching

efficiency and delivery speed, but more importantly,

reconstructs the value relationship between the

platform and consumers, merchants and riders.

Especially during the COVID-19 pandemic,

technological innovations such as contactless

delivery have highlighted the important value of

technology empowerment, driving food delivery

https://orcid.org/0009-0004-5508-2266

platforms to become an important part of urban

infrastructure (Rong and Hu,2023).

This paper uses a longitudinal case study method,

taking Meituan Waimai as the research object, and

systematically examines the impact mechanism of

technological innovation on its business model. The

study constructs a complete chain of evidence

through multi-source data collection and analysis,

focusing on the application scenarios of key

technologies such as artificial intelligence, big data,

and unmanned delivery, and their business model

creation paths. At the same time, through a horizontal

comparative analysis method, the evolution

characteristics of Meituan Waimai platform

attributes, network effects, and intermediary

functions at different development stages are

examined. This case study method can deeply analyze

the specific mechanism of technology empowerment,

reveal the inherent laws of platform business model

innovation, and provide empirical evidence for

494

Zhang, S.

Research on the Development of Business Model of Food Delivery Platform Enabled by Technological Innovation: Taking Meituan Waimai as an Example.

DOI: 10.5220/0013848100004719

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 2nd International Conference on E-commerce and Modern Logistics (ICEML 2025), pages 494-500

ISBN: 978-989-758-775-7

understanding the technology-business co-evolution

in the digital economy era.

This study aims to systematically explore how

technological innovation drives the transformation

and upgrading of the business model of the food

delivery platform through the case of Meituan

Waimai, and its research objectives have both

theoretical value and practical significance. At the

theoretical level, by revealing the innovative

contributions of digital technologies such as artificial

intelligence and big data to platform economic

theory, a new theoretical perspective is provided for

constructing a business model analysis framework in

the digital economy era. At the practical level, the

context-based rapid recommendation strategy

(CoFARS) system, the BETTER business model and

other technical application cases are examined, and

replicable digital transformation experiences are

extracted to provide a reference for platform

enterprises to optimize their technology strategies. At

the same time, the study pays special attention to the

technology empowerment mechanism in emerging

fields such as algorithm governance and data security.

Analyzing the transformation logic of platform

intermediary functions and the governance

innovation problems they generate, it provides a basis

for regulatory authorities to formulate platform

economic policies that adapt to technological

development. Through research in these three

dimensions, it aims to achieve the dual value of

theoretical research and practical guidance, promote

the innovative development of platform economic

theory, and provide practical guidance for digital

transformation for enterprises and policymakers.

2 PLATFORM TECHNOLOGY

INNOVATION ENABLED BY

SCIENCE AND TECHNOLOGY

2.1 CoFARS AI Algorithm Innovation

CoFARS is an intelligent recommendation system

based on artificial intelligence technology developed

by Meituan Waimai. Its core innovation lies in

solving efficiency and accuracy problems in long-

sequence recommendation scenarios through deep

learning and graph neural network technology (Feng

et al., 2024). The system uses a probabilistic encoder

to convert user context (such as time, location, and

weather) into an explainable preference distribution,

and combines prototype learning to cluster and reduce

the dimension of massive user behaviors,

significantly improving the recommendation effect of

cold start scenarios. By building a time-series graph

neural network model, CoFARS can dynamically

capture the evolution of user preferences and

innovatively adopt a two-stage reasoning

architecture, first quickly screening related

subsequences based on context, and then making

accurate predictions through the attention

mechanism.:

2.2 AI Algorithm Innovation of

BETTER Model

The Meituan BETTER model is an intelligent food

delivery business decision-making system based on

AI algorithm. Through cutting-edge technologies

such as deep learning, graph neural network and

reinforcement learning, it conducts a multi-

dimensional analysis of 600 million users on the

platform and builds a precision marketing system that

includes seven major population portraits such as

"urban foodies" and "affordable homebodies"

(MeiTuan Waimai and Bain, 2024). The system

innovatively integrates the spatiotemporal preference

separation algorithm, dynamic pricing model and

digital twin simulation technology, and can realize

the full-link intelligence of demand forecasting,

competition analysis and strategy optimization.

2.3 AI Algorithm Innovation of DPVP

Model

Dual Period-Varying Preference (DPVP) is the next-

generation intelligent recommendation model

developed by Meituan Waimai as a partner. Its core

technology is based on the fusion design of dual

interactive perception architecture and time period

dynamic decomposition network (Zhang, 2023). The

system accurately captures users' differentiated

preferences for store brands and food content through

dual-graph separation modelling (store-level graph +

food-level graph), and divides the whole day into four

periods: Morning/Noon/Night/Late Night, combined

with the time-aware gating mechanism (Time-Aware

Gating), to achieve fine-grained modelling of

dynamic changes in user preferences, which improves

the accuracy of cross-period recommendations by

18.6%. At the feature engineering level, the model

innovatively uses heterogeneous temporal graph

convolution (HT-GCN) to parallelly process the

ternary interaction relationship between users, stores,

and food, and uses personalized food aggregators

(User-Weighted Food Pooling) to weigh the attention

of multiple products in the store, effectively solving

Research on the Development of Business Model of Food Delivery Platform Enabled by Technological Innovation: Taking Meituan Waimai

as an Example

495

the recommendation problem of "different users in

the same store pay attention to different dishes". A

multi-level cache acceleration architecture is adopted

in the online service stage. The first layer achieves

millisecond-level response through pre-calculation of

time period features. The second layer filters low-

relevance nodes in real time based on dynamic graph

pruning technology. Finally, the dual-tower model

(user preference tower + product matching tower) is

used to complete the second-level sorting of hundreds

of millions of candidates sets. Online A/B testing

shows that its GMV increased by 0.7% while

reducing computing resource consumption by 23%.

2.4 Embedded Interactive

Recommender System

Embedded Interactive Recommender System (EIRS)

is a new interactive recommendation system

developed by Meituan Waimai. It realizes seamless

capture of user intentions and dynamic

recommendation optimization through embedded

interactive design (Ji et al., 2023). The system

innovatively adopts the interactive paradigm of

"homepage as questionnaire", uses user click

behavior as real-time feedback signal, and generates

accurate recommendations through three-level

processing modules (retrieval-sorting-acceptance

evaluation). The core technologies include: multi-

channel retrieval architecture (four channels of

popularity/user portrait/DSSM/commodity

collaboration), combined with fine-grained category

filtering (200+ fine categories and 70+ coarse-grained

clusters) to improve the quality of candidate sets;

dual-objective sorting model (click-through

conversion rate CTR&CXR joint prediction)

integrates user real-time behavior characteristics and

contextual information; dynamic embedding

mechanism inserts the recommendation results into

the next position of the user's positive feedback POI,

and ensures the quality of recommendation through α

threshold control. Online A/B testing shows that the

click-through conversion rate of embedded

recommendation results is 132% higher than that of

the original results, and the homepage GMV

increases by 0.43%, while maintaining zero

interference with user experience. The system adopts

a client-server collaborative architecture. The client

captures deep interaction signals such as dwell time

and add-to-cart through the "in-store behavior

perception module", and the server uses the MMoE

multi-task model to achieve feature crossover and

finally enhances interactive perception through

dynamically inserted visual feedback. At present,

EIRS has been fully deployed and has become one of

the core components of Meituan Takeaway's

homepage recommendation system.

2.5 Intelligent Unmanned Delivery

System

Meituan has launched an efficient low-altitude

logistics solution through self-developed artificial

intelligence, big data and drone technology, which

has significantly improved delivery efficiency. By the

end of 2024, Meituan drones have opened 53 routes,

with a total of more than 450,000 deliveries, and

achieved fast delivery in scenic spots, ports and other

scenes. For example, the hot pear soup at the Great

Wall can be delivered in just 6 minutes and 37

seconds. At the same time, Meituan's automatic

delivery vehicle technology has also made

breakthroughs, completing nearly 5 million deliveries

in total, covering more than 100 communities across

the country, and the proportion of autonomous

driving mileage has reached 99%, which can

efficiently deliver daily necessities such as fresh food

and takeaways. These innovative technologies not

only optimize the logistics experience but also create

new consumption scenarios for merchants. For

example, during the May Day holiday in 2024, the

number of orders from merchants around drone routes

increased by more than 300% year-on-year (n.d.).

3 TECHNOLOGY

EMPOWERMENT

STRENGTHENS THE

NETWORK EFFECT OF

MEITUAN TAKEAWAY

In the multilateral market of the takeaway platform, a

close interactive relationship has been formed

between customers, merchants and riders. More

merchants attract more customers because consumers

have more choices; more customers attract more

merchants because merchants hope to reach a larger

user base; the increase in order volume promotes the

growth of rider supply, and the increase in riders

makes delivery faster, thereby further improving

customer experience; faster delivery services enhance

customer stickiness, attract more users to join the

platform, and then promote the further growth of the

supply of merchants and riders, forming a positive

cycle and continuously promoting the prosperity and

development of the platform ecology.

ICEML 2025 - International Conference on E-commerce and Modern Logistics

496

The role of technology empowerment is to

improve matching efficiency, optimize fulfillment

capabilities, and enhance user experience, thereby

strengthening the network effect. For the client,

technology empowerment provides “AI+big data” to

improve matching efficiency. By using context-based

fast recommendation strategy (CoFARS) to combine

time, location, weather and other information, the

BETTER takeaway business system recommends the

most suitable meals for users, gives personalized

discounts, and increases the order rate. From 2023 to

2024, Meituan Takeaway's click-through rate (CTR)

increased by 4.6%, making it easier for users to find

their favorite meals, and the transaction volume

(GMV) increased by 4.2%, stimulating consumption

and increasing the order volume (MeiTuan Waimai

and Bain, 2024). Ultimately, it improves the user

experience, attracts more customers to use the

platform, thereby attracting more merchants, and

increases the frequency of consumption to increase

the order volume, increasing in delivery demand. For

merchants, technology enables intelligent traffic

distribution and improves merchant profitability.

Meituan uses data from hundreds of millions of users

and clusters them into seven categories of takeout

delivery people based on the BETTER takeout

business system: urban foodies, busy parents, small-

town blue-collar workers, affordable homebodies,

mature elites, commuting white-collar workers, and

students who want to satisfy their cravings. Through

analysis of the status quo and trends of the seven

groups across the entire region, brands can gain

insight into the long-term competitiveness of their

own categories and brands from a higher dimension.

Intelligent SKU recommendations are made by

analyzing multi-dimensional data such as user

behavior, order data, dish sales, seasonal changes,

etc., using machine learning algorithms to

recommend the best menu combinations and dish

rankings to merchants, helping merchants improve

operational efficiency, and ultimately using

Meituan’s intelligent marketing tools to increase

merchant sales and attract more merchants to join,

thereby increasing platform supply, and operating

efficiently to reduce costs, making merchants more

willing to cooperate with the platform for the long

term. On the delivery side, Meituan uses big data and

artificial intelligence to optimize the delivery routes

of riders in real time, reduce delivery time, improve

delivery efficiency, and conduct intelligent

dispatching. Orders are automatically assigned

according to the location, delivery capacity, and order

requirements of the riders. Low-altitude logistics and

automatic delivery vehicles improve delivery

efficiency, reduce costs, and create a new consumer

experience for users. Ultimately, the income of riders

increases, which attracts more riders to join, thereby

enhancing delivery capabilities. As delivery time

increases, customer satisfaction will also increase,

increasing the order rate, thereby affecting merchant

sales.

4 TECHNOLOGY

EMPOWERMENT ENHANCES

THE PLATFORM ATTRIBUTES

OF MEITUAN TAKEAWAY

4.1 Multilateral Effect

As a platform-based enterprise, Meituan Takeaway

has a significant multilateral market effect,

connecting consumers (C-end), merchants (B-end)

and riders (D-end) to form an efficient supply and

demand match. Consumers place orders on the

platform, merchants rely on the platform to obtain

traffic, and riders’ complete delivery through

intelligent scheduling. The growth of all parties

promotes each other and forms a positive cycle (Yao

et al., 2022). As the number of users increases, more

merchants will join in, providing a rich selection of

food, thereby attracting more users. The growth in

orders will in turn lead to an expansion of the rider

scale, improve delivery efficiency, and enhance user

experience. This network effect enables Meituan

Takeaway to have a strong market expansion

capability. At the same time, through AI

recommendation, LBS positioning, intelligent

scheduling and other technologies, it further

optimizes supply and demand matching and enhances

platform stickiness.

4.2 Value Co-Creation

Meituan Takeaway promotes the coordinated

development of the platform, merchants, users, and

riders through the value co-creation mechanism to

achieve a win-win situation for all parties. Consumers

are not only recipients of services, but also influence

the ranking of merchants through evaluation, order

sharing, and feedback to optimize platform

experience; merchants can use data analysis and

precision marketing tools to adjust pricing and

optimize dishes to increase sales; riders improve

service quality through the rating system and delivery

data analysis. At the same time, Meituan provides

intelligent scheduling and insurance protection to

Research on the Development of Business Model of Food Delivery Platform Enabled by Technological Innovation: Taking Meituan Waimai

as an Example

497

optimize the delivery experience. The platform not

only provides a trading venue, but also optimizes

resource allocation through AI algorithms, automated

operations, intelligent delivery and other

technological means, enhances user experience,

improves merchant operating efficiency, and reduces

fulfilment costs, ultimately promoting the healthy

development of the entire ecosystem (Xiong, 2024).

5 TRANSFORMATION OF

INTERMEDIARY MODEL

UNDER TECHNOLOGY

EMPOWERMENT

5.1 Transformation Path of

Intermediary Model Driven by AI

Algorithm

From information intermediary to intelligent

decision-making center. The operation mode of food

delivery platform driven by AI algorithm realizes the

paradigm reconstruction of supply and demand

matching mode, and its core lies in the dual

breakthrough of data real-time and decision

prediction (n.d.; Choi et al., 2021). Traditional menu

adopts static update mode, which requires manual

adjustment based on past experience. The data

feedback has a strong lag. Generally, it takes 2-3

months of sales data to judge the performance of

dishes. In addition, the adjustment cost of this mode

is high, and each update requires manual replacement

of posters. After the optimization of BETTER

business model, Meituan overcame the problem of

site selection prediction that was difficult to achieve

due to the high degree of dispersion of the industry

and the lack of database. Using AI site selection tools,

it conducted revenue prediction ability tests based on

multiple merchants, and the accuracy of predicting

food delivery sales has exceeded 90% (Zhang,2023).

Against the background of the continuous

strengthening of platform control, the supply and

demand matching model are undergoing paradigm

reconstruction. As an important member of the

Internet advertising platform, Meituan has achieved

dynamic control of advertisers' conversion costs and

precise fulfillment by introducing the optimized

billing method Optimized Cost Per Mille (OCPM),

relying on the Vickrey-Clarke-Groves (VCG)

bidding mechanism and automatic bidding system.

Based on a real industrial-grade e-commerce platform

dataset, the ROI-sensitive Distributional

Reinforcement Learning (RSDRL) framework

proposed in this paper has significantly improved the

delivery effect in the test of 1,000 OCPM advertisers

and over 10 million auction records: in the offline

evaluation, after applying RSDRL, the number of

conversions of OCPM advertisers increased by about

100%, the payment amount increased by more than

12%, and the return on investment (ROI) was stably

controlled between 0.87 and 0.93, which is lower than

the set cost cap (1.0). At the same time, in the AA/BB

test in the real online environment, the advertiser

group (Group B) using RSDRL achieved a 5%

increase in conversion volume and a simultaneous

increase in platform revenue compared with the

traditional strategy group (Group A) (Tang et al.,

2020). This series of data verifies that in the context

of the dynamic evolution of supply and demand,

enhancing decision-making autonomy and real-time

performance through reinforcement learning has

become an important path for platforms to reshape the

advertising paradigm. This change indicates that food

delivery platforms have evolved from trading venues

to demand centers (Zhang,2023).

The power asymmetry that exists when AI

algorithms are applied in intermediary platforms is

mainly reflected in the platform's unilateral

formulation of rules, monopoly of data control,

weakening of merchants' economic bargaining

power, and lack of effective complaint channels

(n.d.). The platform forces merchants to participate in

promotional activities and adjusts the commission

ratio at will. At the same time, it controls traffic

distribution and search rankings through opaque

algorithms, forcing merchants to purchase paid

promotion services. In addition, the platform has core

user data but does not open it to merchants, making it

difficult to accurately optimize business strategies,

and the vague violation of judgment standards and

mechanized customer service system make it difficult

for merchants to protect their rights. This power

imbalance has led to the continuous compression of

merchant profits (some net profit margins are less

than 5%), and even a vicious cycle of "no orders if

you don't participate in activities, and losses if you

participate in activities" (Zhu et al., 2024). The

reconstruction of the distribution network. The

reconstruction of the distribution network driven by

AI algorithms realizes a paradigm shift in the

allocation of transportation resources. According to

the "White Paper on Trends in China's Instant

Delivery Industry", AI technology realizes full-link

optimization to grasp demand fluctuations in

advance, reasonably dispatch, avoid insufficient peak

transportation capacity and waste of low-peak

transportation capacity, improve order acceptance

ICEML 2025 - International Conference on E-commerce and Modern Logistics

498

and delivery efficiency, and dynamically price to

encourage riders to accept orders in peak or remote

areas, ensure service coverage, improve overall

quality, monitor delivery conditions in real time, and

respond to exceptions quickly. The abnormal

processing time is reduced, and the delivery on-time

rate is improved (Frost and Sullivan, 2024).

5.2 Efficiency Upgrade of the

Intermediary Model Driven by the

Intelligent Distribution System

Specifically, drones can realize high-speed delivery

in the air, automatic vehicles are responsible for

short-distance transportation on the ground, and

riders are responsible for the precise delivery of the

last mile. The three work together efficiently and

greatly improve the overall delivery efficiency

(Zhang,2024). The National Development and

Reform Commission pointed out in the "China

Modern Logistics Development Report 2024" that

platform companies that adopt this three-level

intelligent distribution system have increased their

overall delivery efficiency by an average of 35%.

Especially during the peak hours of lunch when

orders are dense, thanks to intelligent scheduling and

division of labor and cooperation, the service radius

can be expanded from the traditional 3 kilometers to

more than 5 kilometers, significantly improving the

platform coverage and user experience, and helping

the logistics industry to move towards a higher

quality and intelligent development stage (China

Industry Research Network, 2023).

6 CONCLUSION

This study systematically analyzes the enabling

mechanism of technological innovation on Meituan's

food delivery business model through a mixed

research method. The study found that AI-driven

intelligent recommendation systems such as CoFARS

and BETTER and unmanned delivery technologies

have promoted Meituan's transformation from a

"multi-dimensional value chain" to a "retail +

technology" ecosystem, achieving significant results

such as improved supply and demand matching

efficiency, enhanced network effects, and multilateral

value co-creation.

At the theoretical level, this study constructed an

innovative framework of "technology-network

effect-business model", revealing how algorithmic

technology reconstructs the platform's intermediary

function, upgrading it from an information matcher to

an intelligent decision-making center. At the practical

level, technical applications such as intelligent

scheduling systems and drone delivery networks

provide the industry with a replicable example of

digital transformation. At the same time, the study

also found that there are significant differences in

algorithmic credit scores between chain merchants

and small and micro merchants, highlighting the

importance of technology inclusiveness.

Future research can further explore the fairness

mechanism of algorithmic governance, such as

establishing a dynamic regulatory framework to

balance the rights and interests of platforms,

merchants and riders; at the same time, attention

should be paid to the differences in technology

penetration in low-tier cities, and inclusive digital

solutions should be explored to provide empirical

evidence for policy making.

REFERENCES

Choi, Y. et al., 2021. Sustainable management of online-to-

offline delivery apps for consumers’ reuse intention:

Focused on the Meituan apps. Sustainability 13(7),

3593.

China Industry Research Network, 2023. 2023-2029 Global

and China Logistics Drone Market Research and

Development Trend Forecast Report.

https://www.cir.cn/5/87/WuLiuWuRenJiDeQianJingQ

uShi.html.

Feng, Z. et al., 2024. Context-based fast recommendation

strategy for long user behavior sequence in Meituan

Waimai. arXiv.

Frost & Sullivan, 2024. White paper on trends in China's

instant delivery industry. https://www.frostchina.com.

Ji, C. et al., 2023. Interactive recommendation system for

Meituan Waimai. In Proceedings of the 28th ACM

SIGKDD Conference on Knowledge Discovery and

Data Mining (pp. 3245–3249).

Meituan Official Website Information Center, n.d.

Meituan’s automatic delivery enters the stage of

improving operational refinement. Meituan.

Meituan Waimai & Bain & Company, 2024. White Paper

on BETTER Food Delivery Operation System.

Meituan. https://www.meituan.com.

Rong, K., Hu, L., 2023. Complementary assets,

organizational modularization, and platform enterprise

value innovation. Systems 11(7), 323.

Tang, P. et al., 2020. Optimized cost per mille in feeds

advertising. In Proceedings of the 26th ACM SIGKDD

Conference on Knowledge Discovery and Data Mining

(pp. 1359–1367).

Xiong, Y., 2024. Research on the value co-creation effect

of business model innovation of network platform

enterprises. Jiangxi Normal University.

Research on the Development of Business Model of Food Delivery Platform Enabled by Technological Innovation: Taking Meituan Waimai

as an Example

499

Yao, C. et al., 2022. A study on interplatform competition

based on a Lotka–Volterra competition model focusing

on network externality. Electronic Commerce Research

and Applications 56, 101201.

Zhang S. 2024.AI-based transformation and reflections on

the big data push mechanism of traditional food

delivery industry—Taking Meituan as an example, n.d.

Advances in Economics, Management and Political

Sciences.82(1):341-346.

Zhang S. 2023. Takeout platform uses big data for price

discrimination—Meituan takeout platform case study,

n.d. Journal of Research in Social Science and

Humanities. 2(3):36-41.

Zhang, Y. et al., 2023. Modeling dual period-varying

preferences for takeaway recommendation. In

Proceedings of the 29th ACM SIGKDD Conference on

Knowledge Discovery and Data Mining (pp. 5628–

5638).

Zhu, G. et al., 2024. Gig to the left, algorithms to the right:

A case study of the dark sides in the gig economy.

Technological Forecasting and Social Change 199,

123018.

ICEML 2025 - International Conference on E-commerce and Modern Logistics

500