We are a leading data intelligence application software company in China. According to Frost & Sullivan, we are the largest data intelligence application software provider in China in terms of total revenue in 2024. Leveraging our core technologies and industry insights, we offer data intelligence products and solutions, covering marketing and operational intelligence and encompassing online and offline scenarios. We are dedicated to transforming enterprises’ marketing and operational strategy design and decision-making processes leveraging large models, industry-specific knowledge, and multimodal data.
OUR KEY PROPRIETARY TECHNOLOGIES
Our success is built on our innovative and key proprietary technologies, particularly in the fields of data intelligence, enterprise knowledge graph, and data privacy. Our technologies have received widespread recognition. As of June 30, 2025, we had 2,322 patents and 596 patent applications, and have received over 460 domestic and international awards. Specifically, as of June 30, 2025 we owned 1,296 invention patents, encompassing the fields of data intelligence, enterprise knowledge graph and data privacy, among others.
We lead in the use of meta-learning for face recognition in the data intelligence application software industry. Meta-learning refers to a machine learning technology where automatic learning algorithms are applied to metadata, training artificial intelligence models to understand and adapt to new tasks on their own. Introduced in 2019, meta-learning for face recognition reduces the reliance on large datasets, enhancing influencer identification in marketing intelligence and helping businesses to effectively monitor designated store areas at night to ensure security in operational intelligence. Also launched in 2019, we leverage knowledge-graph technologies for sales strategy optimization, which automates the analysis of customer interactions to provide real-time sales insights and helps identify market trends and new business opportunities.
In 2021, we innovatively introduced a meta-learning model capable of developing derivative models to competently conduct various types of tasks. The model creates and trains specialized mini-models to handle different tasks effectively with less computer power than a single large model. Applied in the smart store operating system, the model transforms the inventory management process with less computing power and empowers automated food quality inspection, allowing store employees to focus on other priorities.
We enhanced our knowledge graph modeling in 2022 by the incorporation of “events,” “time,” and “space.” This upgrade enriches the context and relevance of the knowledge graphs and optimizes the storage of time-series data, making them more dynamic and informative. As a result, in marketing intelligence, enterprises can visualize the consumer journey, identifying where consumers interact with a particular brand, and better understand what drives consumer decisions and how these factors evolve over time. By identifying the most effective touchpoints and media channels, enterprises can allocate their marketing budget more efficiently. In operational intelligence, through dynamically linking extensive data sets such as equipment performance data and drawing correlations between factors in the supply chain management such as delivery time, order accuracy, and the quality of goods supplied, the technology enables us to help businesses in proactive maintenance and data-driven decision-making. Specifically, this technology connects real-time information from in-store equipment (such as a restaurant’s oven or a printer) with data from the supply chain (such as shipping delays or order quality). By analyzing these different data streams together, it can spot hidden patterns. For example, it might find that a recurring drop in food quality is preceded by a slight temperature change in a fridge, which was itself caused by a delayed maintenance check. These insights allow businesses to fix problems before they cause breakdowns (e.g. by taking predictive maintenance) and make smarter decisions based on hard evidence, not just guesswork.
In 2023, we further upgraded our knowledge graphs by introducing hypergraph retrieval augmented generation (HRAG), a technique that upgrades the knowledge graph by retrieving and using not just the text format, but also images, speech, and videos without the need to go through a process to convert other data format into text, to efficiently retrieve and connect more diverse data types, offering more precise analyses and richer insights. See “Business—Our core competencies—Key technologies” for more information. For instance, a retail brand can use HRAG to discover that customers in northern regions prefer watching videos with snowy backgrounds during winter, and that violin music in these videos can increase conversion rates by 15%. By analyzing historical sales patterns, weather forecasts, and customer engagement data, the brand can identify periods when customer interest typically peaks during snowy weather. This data-driven approach allows them to launch advertising campaigns at optimal moments, when customers are most likely to respond.
Finally, our application of multimodal large language model (MLLM) beginning in 2023 can deduce the causal relationship between text, images and videos in advertising materials and their impacts, helping enterprises identify the advertising contents that are likely to generate the best marketing performance. For example, the technology can analyze a set of advertisements and determine that advertisements with a red background and concise, conversational text tend to achieve a higher click-through rate than those with a blue background and formal language. It can also identify more complex patterns, such as finding that advertisements featuring pets in videos, when combined with informal text, are significantly more effective in engaging viewers. Our proprietary hypergraph multimodal large language model (HMLLM) launched in 2024 further integrates diverse data types, including EEG and eye movement, supporting enterprises in analyzing more diverse elements in advertising materials, including subjects, emotions, effects, scenes and audiences, to enhance marketing performance, generating marketing content that are predicted to perform well, and even suggesting scripts, visual layouts, and background music for video advertisements.
Our key proprietary technologies are solely developed by us. We do not share any of these technologies with or license any of these technologies to third-parties. For further details on the introduction time, main features, examples of application scenario, and significance for our key proprietary technologies, see “Business—Our Key Proprietary Technologies.”
Multimodal Data and AI in Vertical Scenarios as the Foundation of Data Intelligence
With the iterative progress of big data and artificial intelligence technologies, especially the rapid development of general large models, various industries and enterprises are increasingly focusing on business digitalization and intelligence. The deep integration of data intelligence into business decision-making has become the future trend. According to the Frost & Sullivan Report, China’s data intelligence application software market is expected to have a promising growth prospect, with a projected growth from RMB32.7 billion in 2024 to RMB67.5 billion in 2029, achieving a CAGR of 15.6%.
Currently, general large models still face challenges. One significant challenge is susceptibility to “hallucinations,” where the models generate outputs or information that appear plausible but are factually incorrect or nonsensical. Additionally, these models exhibit decision-making deficiencies in complex scenarios and demonstrate insufficient coordination and controllability in practical applications. Given these limitations, more targeted large models tailored to specific vertical domains have become increasingly essential. The transformation of general large models into those suited for vertical domains hinges on the availability of substantial “high-value” multimodal data with specific industry attributes. With our 19 years of experience in marketing and operational intelligence across multiple industries, we have accumulated a wealth of multimodal data, granting us a unique advantage in developing large models for marketing and operational intelligence applications. Our advanced large models in marketing and operations, deployed across extensive business scenarios, have produced significant amounts of result data. This feedback serves as a valuable resource for finetuning our models. By analyzing result data and client feedback across business scenarios, we continuously optimize our models, which in turn empowers us to generate more accurate and satisfactory results in application scenarios for the clients.
For many years, we have been devoted to enterprise services and the data intelligence application software industry, amassing the industry’s leading multimodal data accumulation, industry insight capabilities, and technical expertise. At the core of our offerings are our multimodal data integration, multimodal data insights, and data-driven, AI-based decision- making capabilities. Leveraging these industry-leading data, insights, and technologies, we provide clients with advanced marketing intelligence and operational intelligence application software. These software products integrate and connect the complex online marketing and offline operational data of enterprises, building a comprehensive data network platform for enterprise operations. This platform transforms marketing and operational data into actionable business insights and provides supporting execution tools, enabling marketing and operational businesses to mutually reinforce each other.
Since our establishment in 2006, we have been persistently exploring new data sources and enterprise needs, constantly innovating in data-driven products and services. Our total revenue increased from RMB1,269.3 million in 2022 to RMB1,462.0 million in 2023. Our total revenue declined from RMB1,462.0 million in 2023 to RMB1,381.4 million in 2024, mainly due to a decrease in revenue from our operational intelligence business. We recorded a total revenue of RMB565.1 million for the six months ended June 30, 2024 and RMB643.8 million for the six months ended June 30, 2025. In marketing intelligence, we have extended our AI capabilities across a wider range of functions—from planning and strategy generation, to content production and execution. By incorporating AI agents into our integrated services, we have attracted new clients, leading to increased revenue in the six months ended June 30, 2025. In operational intelligence, we have driven sales growth in the six months ended June 30, 2025 through enhanced product standardization, expanded AI capabilities, broader scenario coverage, precise customer need fulfillment, and diversified sales channels.
In 2023, we adopted a more standardized product-focused strategy within the operational intelligence domain, exercising greater caution in signing customized service contracts while actively enhancing the development and sales of our standardized products. Customized services involve creating solutions tailored to the unique requirements of individual clients, which can be resource-intensive and less scalable. By contrast, standardized products entail pre-developed services that can be widely adopted by multiple clients with minimal customization, which are typically more cost-efficient and scalable. For example, clients can purchase different modules of the standardized products to address their unique pain points, such as ad monitoring and marketing content generation in the context of marketing intelligence and IT operations management and inventory management in the context of operational intelligence. This strategic shift, being implemented more systematically in 2024, led to an increase in revenue from standardized products, which partially offset the decline in revenue from customized services. Consequently, the revenue structure within our operational intelligence business showed a more balanced composition in 2024 despite a decline in absolute value. In 2025, our enhanced product capabilities and AI innovation had attracted more customers and driven revenue growth. Moving forward, we believe our product strategy will yield further visible results and support sustainable long-term expansion.
For 2022, 2023 and 2024 and the six months ended June 30, 2024 and 2025, our gross profit margins were 53.2%, 50.1%, 51.6%, 50.6% and 55.9%, respectively. Our R&D expenses for 2022, 2023 and 2024 and the six months ended June 30, 2024 and 2025, were RMB750.9 million, RMB480.8 million, RMB353.0 million, RMB173.6 million and RMB150.4 million, respectively. We had operating losses of RMB1,008.9 million, RMB210.9 million and RMB132.3 million for the years ended December 31, 2022, 2023, and 2024, respectively. We had operating loss of RMB84.5 million for the six months ended June 30, 2024, as compared to operating income of RMB6.1 million for the six months ended June 30, 2025.
We had a net profit of RMB1,637.6 million, RMB318.4 million, RMB7.9 million in 2022, 2023 and 2024 and a net loss of RMB98.7 million and RMB203.9 million in the six months ended June 30, 2024 and 2025, respectively. Our net profit position was mainly driven by fair value changes of preferred shares, warrants and convertible notes of RMB2,815.4 million, RMB585.5 million, RMB186.0 million in 2022, 2023 and 2024 respectively, which are in connection with our Company’s value.
Source: Mininglamp-W (02718) Prospectus (IPO Date : 2025/10/23) |