Beyond the financial reports, Pagoda (02411.HK): When algorithms start deciding the fate of a piece of fruit

In the fresh produce retail industry, there has long been an unwritten rule: Losses are an unavoidable cost, and experience is the only weapon. Pagoda is rewriting this rule line by line with code—when a fruit-selling company increasingly bases its core decisions on algorithmic models in servers rather than human intuition, the most fundamental rules of the game in this industry are being redefined.

The public perception of Pagoda (02411.HK) $PAGODA GP (02411.HK)$ mostly stays at the green signboards along the street: whether franchisees are making money, whether prices are too high, and whether the so-called 'wave of store closures' is really happening. These topics are important, of course, but they only depict the first half of the company’s story. In fact, Pagoda's digital transformation has quietly gone through four stages: from introducing POS and ERP systems in 2003 to achieve basic management informatization, to establishing Pagoda Tech with hundreds of engineers in 2014, then fully integrating into the WeChat ecosystem in 2019 to build a precision marketing system, and now moving towards the deep waters of full-chain intelligence. From the timing of fertilizer application for farmers at the origin, logistics routes after harvest, store assignments for deliveries, to whether or not to discount unsold inventory at night, every decision point is being recalculated by AI.

Looking solely at the financial data from the past two years, Pagoda has indeed gone through a difficult adjustment period. The company reported total revenue of 8.174 billion yuan in 2025, down 20.4% year-on-year, with a net loss of 317 million yuan attributable to the company's owners, narrowing by 17.8% compared to the 386 million yuan loss in 2024. The total number of stores fell from a peak of 6,093 at the end of 2023 to 4,468, with a net closure of 1,625 stores over two years, equivalent to cutting more than a quarter of its terminal network. However, a closer look at the annual trend reveals a stark contrast between the first and second halves: 741 stores were closed in the first half, resulting in a net loss of 340 million yuan; in the second half, the store network saw a net increase of 82 locations, marking the end of two years of store closures, with a net profit of 23 million yuan in the second half alone, successfully reversing losses at the operating level. Gross margin also rebounded from a low of 5% in the first half to 10% in the second half, showing clear signs of profit recovery. Yu Huiyong defined 2025 as the 'Year of Consolidation and Strengthening,' using this time to complete overall optimization and internal adjustments, which have now been largely completed. The company has set the tone for 2026 to fully open up supply chain capabilities, evolving from a retail channel to an industry platform, deepening B2B sales (which account for about 16% and maintain high growth), accelerating overseas market expansion, and promoting dual-track development of 'channel brands + category brands.'

The fresh produce industry has a well-known 'impossible trinity' - low wastage, high turnover, and good quality, where only two out of the three can be achieved at most. Pagoda's intelligent ordering system is challenging this fate with computational power. Traditional store ordering heavily relies on the manager’s experience, but human judgment struggles to handle hundreds of variables simultaneously: the manager may know it will rain tomorrow but might not remember that in the past three years, sales of imported grapes in this business district have tripled just before Labor Day. The AI system incorporates historical sales data, business district attributes, weather changes, and holiday patterns into its predictive model, providing precise order suggestions down to single-digit numbers. After implementing this system, the ordering process has been reduced from several hours to mere tens of minutes. One franchisee summarized it succinctly: 'Humans look at whether they can sell more today; the system looks at whether they can earn more over seven days on average.' Currently, Pagoda has scaled up systems such as AI + marketing, AI + supply chain, and AI-based smart ordering, covering thousands of stores. The core of its AI strategy is to 'enhance people' rather than 'replace them' - aiding management in making accurate decisions and empowering frontline staff to improve efficiency and reduce costs.

For chain enterprises with thousands of stores, traditional manual inspections are inefficient and prone to high omission rates, with nighttime monitoring being particularly challenging. Pagoda has introduced an AI inspection system driven by large models, covering more than 20 scenarios such as empty shelf detection, overnight product identification, and employee uniform compliance. The accuracy for empty shelves and overnight product recognition reaches 99%. The system connects existing store cameras to an AI analysis box, capturing violations in real-time and issuing automatic alerts, reducing issue resolution time from 'hours' to 'minutes,' significantly cutting labor costs. In the supply chain, AI technology is also applied to fruit quality checks and label recognition - using a combination of cameras and algorithms to achieve intelligent sorting of deformed or spoiled strawberries, saving over 200,000 yuan annually per production line.

Moving upstream in the industrial chain, Pagoda’s investments run deeper. The company has made significant efforts to introduce BLOF ecological harmony farming technology from Japan, a technique that follows natural laws and improves soil conditions from the source, addressing long-standing agricultural issues like compaction and acidification. This technology has been officially recognized by the Ministry of Agriculture and Rural Affairs as one of the key technologies to promote by 2025. In trial demonstrations conducted in areas severely affected by continuous cropping obstacles, such as Beijing and Shanghai, the use of BLOF technology reduced soil salinity by 35%, increased beneficial microorganism populations by 42%, cut chemical pesticide usage by 76%, lowered planting costs by 35%, and improved marketable yield by 22%. By the end of 2024, nearly 30,000 mu of soil had been improved using BLOF technology. In Guangling County, Shanxi Province, Pagoda partnered with the local government to establish the country's first large-scale melon plantation applying BLOF technology, creating a replicable 'Guangling Model,' which aims to extend its success to major arid-region fruit-producing areas like Minqin in Gansu and Zhongwei in Ningxia, building domestic premium brands. For categories with substantial data accumulation, such as citrus and grapes, Pagoda has developed AI recognition tools to provide farmers with comprehensive services ranging from pest and disease identification to efficacy evaluation.

Pagoda has a unique 'four dimensions, one taste, one safety' grading system for fruits - sugar-acidity balance, freshness, crispness, tenderness, aroma, and safety - breaking down the subjective sensation of 'tastiness' into quantifiable indicators. This system represents a core asset refined over more than a decade. Based on these metrics, Pagoda is constructing an interactive, searchable AI knowledge system utilizing advanced model tools to integrate internal text-to-image, text-to-video materials, and knowledge bases for training. Once mature, this system will convert over two decades of accumulated expertise into readily accessible digital assets, aiding functions from employee training to quality control assistance, procurement decision-making, and consumer communication. When a retail enterprise transforms years of industrial data into reusable digital assets and explores their value in empowering the industry, its business logic undergoes a qualitative transformation.

Implementing paid memberships in the fresh produce sector is notoriously difficult, yet Pagoda has delivered impressive results. As of the end of 2023, there were 83.91 million regular members and 1.17 million paid 'Heart Enjoy' members, with a monthly repurchase rate of 76%; by the end of 2025, total membership exceeded 95.3 million, contributing over 70% of consumption. The underlying support for this achievement is a tag system built in collaboration with WeChat, which predicts customer preferences based on purchasing behavior, forming thousands of audience segments for precise targeting. A typical case involves specifically promoting domestic high-end kiwifruit information to members who have purchased Zespri but not MiZong, driving continuous sales breakthroughs. The membership tier design - including Regular, Silver, Gold, and Diamond cards - subtly leverages the psychological 'progress effect,' turning the membership system itself into an engine for repeat purchases.

Efficiency, standards, and customer relationships - Pagoda didn't build these through slogans alone. Intelligent ordering and AI inspections have minimized losses to levels peers find hard to match, representing compound benefits over time; 'four dimensions, one taste, one safety' and its algorithmic models form a barrier created by 26 years of data accumulation, something money can't buy; nearly 100 million members and thousands of audience tags give Pagoda the confidence to operate without relying on public domain traffic - while others are still buying traffic, Pagoda is already managing its own flow.

Yu Huiyong, Chairman of Pagoda, often says: 'Good fruits come from cultivation; without full control of the industry chain, you cannot build a brand.' On the surface, the fruit business is retail, but in essence, it's agriculture. And agricultural problems ultimately require technological solutions. From establishing a tech company in 2014 to embedding AI across the entire chain today, this journey has taken over a decade. It's not a story of investing today and seeing results tomorrow; it's more like digging a moat - laborious while digging, but once completed, it becomes difficult for others to catch up. While the broader fresh produce retail industry is still searching for the next direction, Pagoda has already provided its answer with a team of hundreds of engineers and an operational AI system.

In the world of algorithms, selling fruits is no longer just about depending on the weather or making purchase decisions based on experience.

24K Views