The e-commerce industry needs to pay more attention to the essence of transactions, which is to provide high-quality products and services, protect consumer rights, promote fair competition, and improve transparency. The e-commerce industry should return to the basics of transactions, circulation, and closing deals, rather than relying solely on price competition or over-promising services.

The capabilities of large models, such as data analysis, consumer insights, and supply chain optimization, can help the e-commerce industry refocus on providing high-quality products and services, returning to the essence of the industry.

Author | Dou Dou

Editor | Pi Ye

Produced by | Industrialist

"Although sales of rice dumplings during this year's Dragon Boat Festival were not as good as expected, based on market analysis and consumer behavior trends, we expect positive growth in mooncake sales this year, with an estimated growth rate of 15% to 20%."

Xinmai Food is a Shanghai-based OEM factory that has produced mooncakes for brands such as LV, Loewe, Gucci, and Dior for the past seven to eight years. In recent years, the factory has undergone some changes. Most notably, it has opened up sales channels to end-consumers, been seen by more B-end enterprises, seen its store fans grow continuously, and experienced a surge in holiday orders...

There are many factors behind these changes, but some deeper reasons are gradually emerging in the narrative of Dong Fanming, the second-generation factory owner.

"We have been actively promoting the intelligence of our supply chain, for example, by using AI vision technology to automatically detect product defects and using digital twin technology to simulate production processes, optimize production plans, and reduce waste."

In fact, since the emergence of large model technology, changes have been taking place in all industries, especially in the e-commerce industry. The technological changes in the era of large models seem to differ from the simple upgrading of electronic devices in the digital era. If digitalization empowers supply chains, it is like equipping your supply chain with basic electronic devices, such as computers and network connections, to enable rapid information flow and recording.

Large model technology brings revolutionary changes to supply chain management. It is not just an information processing tool but an intelligent assistant that can learn and predict market trends, enabling us to make more precise decisions in product development, inventory management, and marketing strategies.

This also accelerates the upgrade of the supply chain, especially in today's booming e-commerce industry belt, which is becoming an important driver for reconstructing the e-commerce industry belt and helping the e-commerce industry return to its commercial essence.

I. A factory 'transformed' by AI

The eve of the Mid-Autumn Festival is the busiest time for Xinmai, with the factory working overtime. The tighter the schedule and heavier the workload, the more important quality inspection becomes. In addition, there are even more challenging aspects such as product packaging design, especially for customers with more complex requirements.

Facing these challenges, Xinmai solved the problem of large-scale and rigorous quality inspections during holidays through AI intelligent visual quality inspection. Additionally, AI packaging was introduced for packaging design.

It is worth noting that in the e-commerce industry belt, factory operating modes differ from those of merchants on retail e-commerce platforms, with relatively fixed working hours. For example, the workday typically starts at 7 am and ends at 6 pm, resulting in slower customer service response times. This requires us to improve efficiency through technological innovation, such as using AI customer service robots to provide 24/7 customer service.

In the past, customer service robots could only provide basic replies through automated responses or keyword searches. However, with the support of large model technology, customer service robots can now answer more complex questions, providing a better pre- and post-sales service experience.

It can be observed that Xinmai's supply chain is undergoing unique changes in product design, processing, and sales. These changes differ from those seen in the digital era, as the learning and prediction capabilities of large model technology can help make better decisions.

Xinmai's supply chain is undergoing a revolution, with profound changes occurring at every stage from product design and processing to sales. This is not just a simple upgrade of digitization but a breakthrough in the learning and prediction capabilities of large model technology, helping Xinmai make more precise and efficient decisions.

In today's fiercely competitive e-commerce industry, where prices are constantly under pressure, large models are precisely the solution to breaking the cycle of 'internal competition.'

It is important to understand that there are many links in the journey of a product from raw material procurement to delivery to the consumer, where costs can be saved and profits increased. However, this does not mean blindly reducing product quality or engaging in low-price competition.

Logically speaking, there should be no such chaos in industries like e-commerce. The reason behind this lies in the fact that the upgrading and optimization of these links have not yet reached an ideal state, and there are still many unconnected links. This is a recurring issue - supply chain management.

For example, B-end merchants may have inadequacies in customer service and after-sales support, making it difficult to effectively address consumer concerns and doubts. Merchants may lack accurate positioning of target markets, leading to a mismatch between products and market demands;

The existing e-commerce logistics system, structure, and capabilities are not compatible with rapidly changing logistics demands, leading to contradictions between supply and demand, especially in online retail and local life sectors.

Therefore, the fundamental solution lies in optimizing and upgrading each link to restore the normal business model of the entire industry belt and move away from vicious competition.

II. Starting with mooncake marketing,

Weaving an AI e-commerce network

How do large models help upgrade each link?

In fact, due to years of operation, enterprises often accumulate a large amount of data. Taking mooncake OEM factories as an example, they collect sales data for mooncakes over the past few years, including sales volume, revenue, and timing (such as around the Mid-Autumn Festival); acquire user behavior data from e-commerce platforms regarding browsing, searching, and purchasing mooncakes; collect information on raw material price fluctuations and production costs; and gather user evaluations, ratings, and feedback on mooncakes to understand consumer preferences.

Subsequently, different flavors and packaging types of mooncakes are labeled and classified; user evaluations are emotionally labeled to distinguish between positive and negative feedback; and invalid or incorrect data, such as abnormal transaction records and duplicate evaluations, are removed.

Then, time-related features are extracted, such as the impact of festivals and seasons on mooncake sales, and user personas are constructed, including age, gender, purchasing power, and other characteristics, as well as product characteristics such as mooncake flavors, packaging, and pricing.

By collecting, cleaning, and labeling these data, a time series prediction model can be trained to forecast future mooncake sales. Based on the results of the sales prediction model, production plans can be arranged reasonably to avoid excess or shortages, thereby optimizing production planning.

Models can also be trained using user behavior and feedback data to analyze the preferences of different user groups. E-commerce platforms can use consumer preference models to recommend mooncake flavors or packaging that users may like, achieving personalized recommendations.

Furthermore, models can be trained using image recognition technology to automatically identify defects in mooncake production, deploying quality inspection models on the production line to automatically reject substandard mooncakes, improving product consistency, and controlling mooncake quality.

Internally, based on large model technology, the factory can reduce inventory backlog through sales forecasting, lowering storage costs; use consumer preferences and social media data analysis to develop more precise marketing strategies; and monitor raw material price fluctuations and supply chain risks to adjust procurement strategies in a timely manner, empowering its own business.

From a broader perspective, the factory can share market demand predictions based on large models with upstream and downstream supply chain partners, helping suppliers and distributors prepare in advance and reduce inventory backlogs or stockout risks.

Based on the predictions, OEM factories can more accurately formulate production plans, coordinate procurement timings with raw material suppliers, and optimize the overall production rhythm of the supply chain.

OEM factories can manage inventory more effectively, reducing excesses or shortages. This inventory management strategy can be extended to the entire industry belt, improving overall efficiency.

Using large models to analyze market and supply chain risks and adjust strategies in a timely manner, this risk management capability can provide early warning and response plans for the entire industry belt.

Moreover, by improving product quality through large model technology, OEM factories can also drive the improvement of quality standards across the entire industry belt, enhancing consumer trust.

This is a microcosm of how large model technology, as it continues to proliferate, is assisting in the reconstruction of the e-commerce industry belt.

In the era of large models, every seller and link in the e-commerce industry belt, empowered by these models, will more rapidly 'connect the dots' to form an AI network. While enhancing their competitiveness, they will also drive the optimization and upgrading of the entire e-commerce industry belt and supply chain through technology, data, and experience sharing, achieving common development.

However, in this case, the dilemma of large model implementation becomes increasingly clear. Currently, the mainstream applications or relatively mature implementation scenarios of AI large models in e-commerce are primarily in intelligent product selection, AI shopping guides, intelligent customer service, AI digital humans, and intelligent marketing.

These scenarios have high digital penetration rates and standardized data, making them the first choice for large model technology implementation in the e-commerce industry, with enterprises being relatively proactive.

According to CCTV Market Research data, 36% of advertisers have already started using AIGC technology in their marketing activities, indicating that AI applications in marketing have gained widespread recognition and adoption.

It is worth noting that not all participants in the e-commerce industry belt have the capability to integrate large model technology into their businesses. They require assistance from a helper.

III. Behind the reconstruction:

Where lies the correct path for AI e-commerce?

In the digital era, the returns on digital transformation are often not immediate, and enterprises may need to continuously invest funds to sustain it. This is even more so for large models, which are more costly and technically complex.

Moreover, deploying AI often requires significant capital investment, including software, hardware, system integration, and employee training. Additionally, to maintain competitiveness, enterprises need to continuously engage in technological innovation and product upgrades, which require substantial financial support.

Although deploying AI technology requires significant investments in software, hardware, system integration, and employee training, close collaboration with technology partners can effectively reduce these costs and accelerate technology implementation.

For large model technology to integrate with supply chains and make rapid and accurate decisions, standardized data is required across all links, which involves extensive digital transformation.

It is noteworthy that as large model technology gains popularity, more business models are being implemented. Increasingly, e-commerce platforms and social media platforms are deploying their AI businesses, packaging their capabilities with AI and offering them to merchants in the industry belt.

Alibaba, for example, provides AI platform services for enterprises, such as the recently launched AI employees for B-end merchants on 1688, alleviating customer service pressure for B-end enterprises while helping them build their AI applications, such as intelligent customer service robots and product image recognition. In the overseas market, Alibaba leverages AI technology to address issues such as localized marketing strategies in cross-border trade, supporting enterprises in better entering international markets.

Tencent, leveraging its strengths in social e-commerce, enhances the effectiveness of social media advertising through AI technology and helps merchants establish closer connections with consumers through services such as intelligent customer service and virtual avatars. This simplifies merchants' operational processes and further enhances the value creation capabilities of each link in the supply chain.

JD.com focuses on the supply chain logistics sector, using large model technology to optimize key aspects such as inventory management, demand forecasting, and logistics routing, significantly reducing costs and improving efficiency. Its integrated supply chain solutions promote digital transformation across the entire industry chain.

In summary, platforms such as Tencent, JD.com, and Alibaba have deeply integrated large model technology into their respective business processes, not only enhancing their competitiveness but also providing strong support for the digital transformation of the entire industry chain.

From localized services for cross-border e-commerce to personalized recommendations for social e-commerce to intelligent management of supply chain logistics, with the support of these platforms' technologies, the e-commerce industry is being rapidly reconstructed.

This is also enabling the e-commerce industry to gradually return to its essence.

IV. Large models help the e-commerce industry belt return to its essence

In the e-commerce industry, the controversy surrounding the phenomenon of 'refund only' continues unabated, causing distress to merchants and posing new challenges to the protection of consumer rights. Players in the industry need to fundamentally re-examine and reconstruct the service system of e-commerce.

"I'm not going upstairs" "Wait for my legal advisor" "Can we get back the items that were refunded only?" "Don't talk to the customer's face"...

On a short video platform, a video of a merchant defending their rights has gone viral. In the video, the merchant is accompanied by a legal advisor, videographer, and mediator, demonstrating a strong desire to seek justice amidst the chaos of 'refund only' disputes. Meanwhile, the hesitant and evasive consumer makes the outcome of this 'confrontation' apparent from the start.

The prevalence of the 'refund only' phenomenon not only troubles merchants but also affects the healthy development of the e-commerce industry.

Using big data analysis to identify and prevent malicious refund behavior can maximize the protection of the interests of both merchants and consumers.

Merchants have established mutual aid associations to collectively address the issue of malicious refunds. This model allows merchants from different regions to assist each other. When a merchant encounters a 'refund only' dispute, they can contact another merchant in the consumer's location through the mutual aid association, who will represent the former in offline communications with the consumer to demand the return of goods or payment.

In recent years, with the introduction of the 'refund only' policy, tactics by 'free riders' and 'wool pullers' have emerged in various forms. "Even lipsticks bought years ago can be returned," complained one merchant on a social platform.

However, from the consumer's perspective, there are also differing views. One buyer recalled her first experience with a 'refund only' in the comments section: "I once bought tangerines, but several were rotten, and the others were dry and hard-skinned."

It can be said that under the chaos of 'refund only,' both buyers and consumers have suffered to some extent, with only the malicious refunders and shoddy merchants seemingly benefiting.

Essentially, 'refund only' is a 'competition' to enhance merchant service and platform service capabilities. It was originally intended for perishable goods, which are often expensive to return due to logistics costs and are prone to damage, making them unsellable a second time.

However, this path seems to be veering off course, gradually deviating from the essence of e-commerce - transactions, circulation, and closing deals.

Merchants in the industry belt seem to find it difficult to do normal business and make money, turning it into a luxury.

"Many merchants are selling at a loss on a domestic e-commerce platform" "Many merchants rely on volume sales, earning advertising fees by inserting free game trial promotion cards from game companies in packages" In recent years, merchants have invested most of their revenue in advertising and traffic generation. They compete on exposure, pricing, and then on service when they can't compete on price anymore.

The e-commerce industry needs to pay more attention to the essence of transactions, which is to provide high-quality products and services, protect consumer rights, promote fair competition, and improve transparency. The e-commerce industry should return to the basics of transactions, circulation, and closing deals, rather than relying solely on price competition or over-promising services.

The data analysis, consumer insights, and supply chain optimization capabilities possessed by large models can help the e-commerce industry refocus on providing high-quality products and services, returning to the essence of the e-commerce industry.

An idealized supply chain that can be seen is one in which every link in the supply chain becomes an individual in a social network, optimizing the "communication" and "cooperation" between them through intelligent algorithms to improve the overall supply chain's collaboration efficiency.

Large models will be an essential driving force in this evolution process.