Method and system for operating brand consulting and advertisement generation ai platform based on customized marketing funnel for smes

Abstract

According to one embodiment, the system collects industry information and business data of small and medium-sized enterprises or small business owners, generates a marketing funnel structure based on the collected industry information and business data, performs AI reverse analysis to analyze the funnel in reverse from a consumer perspective regarding business ideas or marketing strategies entered by a user to derive purchase journeys and churn factors, automatically generates marketing content based on the derived analysis results, automates subscription-based settlement processing, collects and analyzes performance data resulting from the execution of marketing content to calculate industry-specific marketing performance indicators, and optimizes the marketing funnel structure and content based on the calculated performance indicators.

Description

Technology Field

[0001] The present invention relates to the fields of artificial intelligence (AI), digital marketing, and subscription economy platform technology.

[0002] More specifically, this relates to a method and system for operating a brand consulting and advertising generation AI platform based on a customized marketing funnel for small and medium-sized enterprises (SMEs), which designs a customized marketing funnel based on the business data of SMEs and micro-enterprises, pre-validates strategies through AI reverse discussion, and provides automated content creation and subscription settlement processing in a one-stop manner. Background Technology

[0004] With the recent rapid growth of the digital marketing market and the advancement of generative AI technology, there is a growing trend of attempts to reduce the time and cost of content creation. However, 75.7% of domestic small and medium-sized enterprises (SMEs) have monthly sales of less than 1 million won, leaving them unable to hire dedicated marketing personnel. Furthermore, outsourcing to large agencies incurs high costs of at least 5 million won, leaving them in a marketing blind spot.

[0005] Existing solutions, such as freelance marketplaces or DIY marketing tools (e.g., Canva), had limitations in that it was difficult to maintain brand differentiation and consistency due to significant variations in the quality of deliverables or a lack of professionalism.

[0006] Furthermore, existing AI marketing tools are limited to simple image or text generation and lack features such as funnel design leading to actual contract conversions or sales, as well as optimization based on performance analysis data, which prevents them from leading to substantial sales growth for small and medium-sized enterprises. The problem to be solved

[0008] The present invention aims to collect industry information and business data of small and medium-sized enterprises or small business owners, and to automatically generate and optimize a customized marketing funnel structure that reflects the practical experience of marketing experts.

[0009] In addition, the present invention is designed to allow AI to analyze and discuss the funnel in reverse from a consumer perspective regarding business ideas or marketing strategies input by a user, with the purpose of checking and supplementing dropout factors in the purchase journey in advance before implementation.

[0010] The present invention aims to alleviate the cost burden on small business owners and increase the efficiency of platform operations by automatically generating marketing content that aligns with brand consistency based on verified marketing strategies and automating subscription-based settlement processing. means of solving the problem

[0012] According to one embodiment, the system collects industry information and business data of small and medium-sized enterprises or small business owners, generates a marketing funnel structure based on the collected industry information and business data, performs AI reverse analysis to analyze the funnel in reverse from a consumer perspective regarding business ideas or marketing strategies entered by a user to derive purchase journeys and churn factors, automatically generates marketing content based on the derived analysis results, automates subscription-based settlement processing, collects and analyzes performance data resulting from the execution of marketing content to calculate industry-specific marketing performance indicators, and optimizes the marketing funnel structure and content based on the calculated performance indicators.

[0013] The system receives a business profile from small and medium-sized enterprises or micro-enterprises that includes industry category, monthly sales volume, major sales channels, and the age group and interests of target customers; searches for funnel templates corresponding to the industry category and sales volume range of the received business profile from a funnel template database classified and accumulated by industry based on the marketing execution history of existing subscribers; generates a customized marketing funnel structure including awareness, interest, conversion, and repurchase stages by reflecting the target customer characteristics of the business profile in the searched funnel templates; for each stage of the generated marketing funnel structure, selects content types appropriate for that stage based on the industry category and the age group of target customers, matching social media video content for the awareness stage, detailed landing pages for the interest stage, discount or promotional ads for the conversion stage, and retargeting ads for the repurchase stage; calculates a total advertising budget proportional to the monthly sales volume for each matched content type; and refers to industry-specific funnel stage contribution data, which calculates the average ratio of each funnel stage's contribution to final conversion from the marketing execution results of existing subscribers within the same industry category accumulated in the funnel template database, to determine the total advertising budget for each An execution strategy can be established to distribute based on the contribution ratio of each funnel stage.

[0014] In performing AI reverse discussion, the system receives marketing strategy information in text form from a user, including the characteristics of the product or service to be marketed, the expected target customer base, major advertising channels, and key appeal messages; based on the received marketing strategy information, an AI model generates a virtual consumer persona that reflects the demographic characteristics and consumption propensity of the expected target customer base; as the generated virtual consumer persona navigates each stage of the marketing funnel in reverse order from the final conversion stage to the awareness stage, it evaluates whether the information or motivation necessary to maintain purchase intent is satisfied at each stage; during the evaluation process, stages where the purchase intent of the virtual consumer persona is judged to decrease are identified as churn risk zones; cases where the correlation between the key appeal message and the content provided at that stage is below a preset threshold or where differentiated value compared to competing products is not specified are classified as churn factors; for each classified churn factor, improvement measures including content reinforcement, change of appeal message, or switching of advertising channels are generated; and a strategy review report is generated that maps the churn risk zones and improvement measures to the funnel stages and can be provided to the user.

[0015] In automatically generating marketing content, the system receives and stores brand guideline information including the user's brand logo, representative color, font, and tone and manner, and provides industry information, the corresponding stage of the marketing funnel, and brand guideline information as input conditions to an image generation model to generate an advertising image that conforms to the brand's representative color and tone and manner, and provides consumer behavior goals and key appeal messages for the corresponding stage of the marketing funnel as input conditions to a text generation model to generate advertising copywriting in stages, including short hooking phrases to enhance brand awareness in the awareness stage and phrases emphasizing benefits to induce purchasing behavior in the conversion stage, and combines the generated advertising image and advertising copywriting according to the specifications of the advertising channel, and produces channel-specific content according to the resolution, text placement ratio, and file format requirements of each of the social media feed ads, story ads, banner ads, and landing page content, and verifies the color match, font usage, and tone and manner consistency with the stored brand guideline information for each produced channel-specific content, confirms content with a pre-set brand consistency score or higher as distributable, and performs regeneration for content with a score lower than that.

[0016] In automating subscription-based settlement processing, the system sets and stores billing information including monthly subscription fees and payment cycles according to service grades for each subscriber, links with the API of a PG payment system to request automatic billing of subscription fees using a registered payment method for subscribers whose stored payment cycle has arrived, receives a payment approval or failure response from the PG payment system, and for subscription fees for which payment approval has been received, calculates respective settlement amounts according to the pre-set revenue sharing ratio between the marketing service provider and the platform operator who have entered the platform, while calculating subscription service revenue, single project revenue, and platform usage fees separately by item, aggregates the settlement amounts calculated by item according to the pre-set settlement cycle, and generates and provides a settlement report for customers including payment details and service usage status to subscribers, and a settlement report for providers including revenue sharing details and expected deposit amounts to service providers.

[0017] The system performs anonymization processing to remove individual corporate identification information, including company name, representative information, and business registration number, from marketing performance data collected from multiple subscribers belonging to the same industry category, and constructs an industry-specific benchmark dataset by classifying the anonymized performance data by industry category and revenue size range. The industry-specific benchmark dataset aggregates performance data from multiple anonymized companies within the same industry category and revenue size range, functioning as reference data representing the average marketing performance level of the corresponding industry and size range. It matches the industry category and revenue size range extracted from the business profiles of new subscribers with a benchmark group, which refers to a group of companies belonging to the same industry category and revenue size range within the industry-specific benchmark dataset. It then calculates a marketing maturity score by comparing the average ad conversion rate, average customer acquisition cost, and average funnel stage churn rate of the matched benchmark group with the new subscribers' current marketing metrics by item. Based on the calculated marketing maturity score, it identifies the funnel stage that recorded the lowest score relative to the benchmark group's average as a priority target for improvement. For the identified stage, it examines the content types commonly applied by companies that achieved performance corresponding to a pre-set top ratio within the benchmark group, and By analyzing combinations of advertising channels, you can automatically generate and recommend customized action plans to apply to new subscribers.

[0018] After the execution of marketing content is completed, the system calculates performance metrics including click-through conversion rates by advertising channel, cost of acquisition per customer, and return on investment relative to advertising costs based on data collected from advertising platforms regarding impressions, clicks, conversions, and advertising expenditure. It derives the actual churn rate for each funnel stage from the calculated performance metrics and evaluates whether there is a match between the predicted churn zone and the actual churn zone at each funnel stage by comparing the stage identified as a churn risk zone by the virtual consumer persona in the AI ​​reverse discussion with the stage with a high actual churn rate. If there is a stage where the predicted churn zone and the actual churn zone do not match in the evaluation results, it collects consumer behavior data at that stage as additional training data to update the consumption propensity parameters and churn judgment thresholds of the virtual consumer persona. It can then repeatedly execute a feedback loop using an AI reverse discussion model with the updated consumption propensity parameters and churn judgment thresholds applied, so that the accuracy of churn prediction improves compared to before the update when performing reverse discussions on subsequent marketing strategies. Effects of the invention

[0020] The method and system for operating a brand consulting and advertising generation AI platform based on a marketing funnel tailored for small and medium-sized enterprises (SMEs) according to the present invention can achieve an annual cost reduction of 82.5% compared to hiring a regular marketing team, while systematizing a funnel structure that reflects the practical experience of experts, thereby providing high-quality one-stop integrated services to SMEs in marketing blind spots.

[0021] In addition, the present invention enables the pre-examination of the purchase journey and churn factors from a consumer perspective before strategy execution through an AI reverse discussion function, thereby establishing a data-driven decision-making system and structurally improving the waste of unnecessary advertising costs.

[0022] The present invention can resolve the risk of quality instability associated with outsourcing and help small and medium-sized enterprises accumulate long-term brand assets and grow by providing design, marketing, performance analysis, etc., within a single ecosystem based on a subscription model. Brief explanation of the drawing

[0024] FIG. 1 is a diagram illustrating a system using a neural network model according to one embodiment. FIG. 2 is a diagram illustrating the learning of a neural network according to one embodiment. FIG. 3 is a diagram showing a neural network structure according to one embodiment. Figure 4 is a block diagram showing the configuration of a customized brand consulting and advertising AI platform system for small and medium-sized enterprises according to one embodiment. FIG. 5 illustrates the overall operation flow of a customized brand consulting and advertising AI platform for small and medium-sized enterprises according to one embodiment. FIG. 6 illustrates an AI reverse discussion execution operation according to one embodiment. FIG. 7 illustrates a subscription-based settlement automation operation according to one embodiment. FIG. 8 is a flowchart illustrating the detailed execution process of an AI reverse discussion according to one embodiment. Specific details for implementing the invention

[0025] Various embodiments are described in detail below with reference to the attached drawings. However, these embodiments may be modified in various forms, and therefore the scope of the patent application is not limited or restricted by these embodiments. All modifications and equivalent substitutions should be interpreted as being included within the scope of the rights. Specific structural or functional descriptions for each embodiment are for illustrative purposes only and may be modified in various forms. Accordingly, the embodiments are not limited to a specific form, and the scope of this specification includes variations of the technical concept, equivalents, or substitutions.

[0026] The terms "first" or "second" may be used to describe various components, but are intended merely to distinguish each component. For example, the first component may be named the second component, and likewise, the second component may be named the first component. When a component is referred to as being "connected" to another component, this means that it is directly connected or that another component may exist in between. The terms used are for descriptive purposes only and are not intended to be limiting. Singular expressions include the plural unless otherwise interpreted in the context. In this specification, expressions such as "comprising" or "having" indicate the presence of the features or elements described in the specification and should be understood as not excluding the possibility of the existence of other features or elements.

[0027] Unless otherwise specifically defined, all terms used herein have the meaning generally understood by those skilled in the art. Terms defined in commonly used dictionaries should be interpreted in a meaning consistent with the context of the relevant technology and should not be interpreted in an overly formal sense unless explicitly defined in this application. Additionally, when describing the drawings, the same reference numerals are used for identical components, and redundant descriptions are omitted. When describing embodiments, if a detailed description of the relevant technology is likely to unnecessarily obscure the essence of the embodiment, such detailed description may be omitted.

[0028] The embodiment can be implemented in various forms, such as personal computers, laptop computers, tablets, smartphones, televisions, smart home appliances, intelligent vehicles, kiosks, and wearable devices. An artificial intelligence (AI) system is a computer system that implements human-level intelligence; unlike existing rule-based smart systems, it is a system in which the machine learns and makes decisions on its own. As the use of AI systems increases, recognition rates improve and user preferences are identified more accurately, so existing systems are gradually being replaced by deep learning-based AI.

[0029] Artificial intelligence technology consists of machine learning and various component technologies that utilize it. Machine learning refers to algorithms that autonomously classify and learn the characteristics of input data, while component technologies are techniques that mimic human cognitive and judgment functions through machine learning algorithms, such as deep learning. These technologies include linguistic understanding, visual understanding, reasoning and prediction, knowledge representation, and motion control.

[0030] The various fields where artificial intelligence technology is applied are as follows. Linguistic understanding refers to the technology of recognizing and processing human language, including natural language processing, machine translation, dialogue systems, question answering, and speech recognition and synthesis. Visual understanding refers to the technology of recognizing and processing objects, including object recognition, tracking, image search, human recognition, and scene understanding. Inference and prediction refer to the technology of judging and predicting information, including knowledge-based inference, optimization prediction, and recommendation systems. Knowledge representation refers to the technology of automatically processing information from human experience, including knowledge construction and management. Motion control refers to the technology of controlling the movement of autonomous vehicles or robots, including navigation, collision avoidance, and driving control. Generally, machine learning algorithms are trained using a trial-and-error method to apply them to real-life situations. Deep learning, in particular, requires hundreds of thousands of iterations. When it is difficult to execute this in a real-world environment, training is conducted through simulations by implementing a virtual environment on a computer.

[0031] In this invention, Artificial Intelligence (AI) refers to a technology that implements human learning, reasoning, and perceptual abilities on a computer, and includes concepts such as machine learning and symbolic logic. Machine learning (ML) is an algorithmic technology that learns the characteristics of input data on its own. Through such machine learning algorithms, AI technology can analyze input data, learn the results, and perform judgments or predictions based on them. Furthermore, technologies that mimic human cognitive and judgment functions using machine learning algorithms can also be included in the category of AI. For example, fields such as linguistic understanding, visual understanding, reasoning and prediction, knowledge representation, and motion control fall into this category. Machine learning refers to the process of training a neural network model through experience in processing data. Through this, computer software can improve its data processing capabilities on its own. A neural network model models the correlations between data and can be expressed by various parameters. The core of machine learning is to optimize the model's parameters by extracting and analyzing features from given data to derive relationships between the data and repeating this process. For example, a neural network model can learn the relationship between input and output, or learn relationships by deriving regularities solely from the input data.

[0032] Artificial intelligence learning models or neural network models are designed to implement the structure of the human brain on a computer and include multiple network nodes that mimic neurons. These nodes are interconnected to exchange signals and process data through layers of varying depths within the AI ​​learning model. Such models may include artificial neural networks and convolutional neural networks (CNNs). For example, AI learning models can be machine learned through methods such as supervised learning, unsupervised learning, and reinforcement learning. Machine learning algorithms that may be used include decision trees, Bayesian networks, support vector machines, artificial neural networks, AdaBoost, perceptrons, genetic programming, and clustering.

[0033] CNNs are a type of multi-layer perceptron designed with minimal preprocessing, consisting of one or more convolutional layers and general artificial neural network layers. Thanks to this structure, CNNs can effectively utilize 2D input data and demonstrate excellent performance in image and speech fields. CNNs are trained via standard backpropagation and have the advantage of being easier to train and using fewer parameters than other feedforward artificial neural network techniques. Convolutional networks are neural networks containing a set of nodes with bound parameters, and many computer vision tasks have been improved as the amount of training data increases and computational power improves. In today's large-scale datasets, overfitting is not critical, and increasing the network size improves test accuracy. Since the optimal use of computing resources becomes a limiting factor, distributed and scalable implementations of deep neural networks become necessary.

[0035] FIG. 1 is a diagram illustrating a system using a neural network model according to one embodiment.

[0036] As illustrated in FIG. 1, a system (100) using a neural network model may include a plurality of user terminals (110-1 to 110-n), a network (N), a server (120), and a database (130).

[0037] A plurality of user terminals (110-1 to 110-n) may be computing devices used by small and medium-sized enterprises or small business owners to input business profile and marketing strategy information, and to receive automatically generated marketing content, AI reverse discussion-based strategy review reports, and settlement reports from the server (120).

[0038] The network (N) refers to a communication network for transmitting and receiving data between a plurality of user terminals (110-1 to 110-n), a server (120), and a database (130), and may include various wired and wireless communication networks such as the Internet, mobile communication networks, and local area networks.

[0039] The server (120) may be a device that generates a customized marketing funnel structure based on collected business data, performs AI reverse discussion through a virtual consumer persona using a neural network model, and controls the overall automatic generation of marketing content and subscription-based settlement processing.

[0040] The database (130) can store industry-specific funnel templates, user business profiles, brand guidelines, and accumulated marketing performance data required for the operation of the server (120). The database (130) may be configured as a physical device separate from the server (120) as shown in FIG. 1 and connected via a network (N), or it may be implemented by being integrated inside the server (120).

[0042] FIG. 2 is a diagram illustrating the learning of a neural network according to one embodiment.

[0043] Referring to FIG. 2, the server (120) can train and operate an artificial neural network (210) to perform virtual consumer persona generation, AI reverse discussion-based funnel exit risk prediction, and automatic generation of marketing content.

[0044] The server (120) can input input data (220) for learning into an artificial neural network (210) and perform a feedforward operation to derive a prediction result (230). In an embodiment of the present invention, the input data (220) may be a business profile of a small and medium-sized enterprise, marketing strategy information, target customer characteristics, or marketing execution history data of existing subscribers. Additionally, the prediction result (230) may correspond to the probability of a virtual consumer churning at a funnel stage, whether the expected purchase intention is maintained, or the configuration of optimized marketing content.

[0045] The derived prediction result (230) is compared with the truth data, which is the actual marketing execution result (e.g., actual ad conversion rate, churn rate, etc.), and an error value is calculated through a loss function. The calculated error value undergoes a backpropagation process that propagates in the reverse direction of the artificial neural network (210), through which the gradients of each node are calculated and the weights inside the artificial neural network (210) are updated through an optimization algorithm.

[0046] Through this process, the server (120) can continuously improve the prediction accuracy of the consumption propensity parameter of the virtual consumer persona and the reverse discussion model by repeatedly executing a feedback loop that reduces the error between the predicted deviation interval and the actual deviation interval.

[0048] FIG. 3 is a diagram showing a neural network structure according to one embodiment.

[0049] According to one embodiment, a neural network may be composed of an input layer, a hidden layer, and an output layer. The input layer is the layer into which data is input to the neural network. In the hidden layer, a feature map may be generated by performing a multiply-accumulate operation and an activation operation on the input data. The multiply-accumulate operation is a process of multiplying the input data by a corresponding weight and then accumulating the results of these multiplications. The activation operation is a process of generating a final output by passing the result of the multiply-accumulate operation through an activation function, and various types of activation functions may be applied. For example, activation functions may include, but are not limited to, the sigmoid function, hyperbolic tangent function, ReLU function, Leaky ReLU function, Max Out function, and ELU function.

[0050] A hidden layer may consist of one or more layers. For example, if it is divided into a first hidden layer and a second hidden layer, the first hidden layer generates a feature map by performing multiplication-accumulation and activation operations based on data received from the input layer, and this feature map can be used as input data for the second hidden layer. The second hidden layer can receive the feature map generated by the first hidden layer and perform multiplication-accumulation and activation operations. The output layer may be a layer related to the result of the operation performed in the hidden layer.

[0051] In one embodiment, a learning model learns syllable (character) patterns that are frequently used together in a given corpus to automatically identify the boundaries between compound words and named entities, and integrates object information from a first UI source with object information rendered in a browser to generate a learning object information file. Using this learning object information file, data for training a deep learning network is generated, and based on data received from various domains, the data is standardized into an integrated format according to one or more standardization methods suitable for each domain. Subsequently, data of a specific domain is learned and inferred, information necessary for standardization in that domain is determined, and post-processing can be performed on the data received from various domains.

[0052] The first UI source includes an XML file, and the training object information file includes an input JSON file for feature learning and an output JSON file used as label data during the training process. This output JSON file is a file containing HTML DOM Tree information implemented in compliance with web standards. Various domains include at least one of a RAN (radio access network), a transport, or a core, and the post-processing process may include a correlation function.

[0054] Figure 4 is a block diagram showing the configuration of a customized brand consulting and advertising AI platform system for small and medium-sized enterprises according to one embodiment.

[0055] Referring to FIG. 4, the system (400) may include a communication interface (410), a database (412), an input / output interface (414), a processor (420), and a memory (430).

[0056] The communication interface (410) may refer to a means of communication for the system (400) to transmit and receive data with an external device or an external server. The communication interface (410) may be connected to a user terminal (e.g., PC, smartphone, tablet, etc.) via a network to receive the user's business profile, marketing strategy information, and brand guideline information. The communication interface (410) may be linked with an external payment server (e.g., Toss Payments PG server) to transmit and receive requests for automatic payment of subscription fees and payment approval responses. The communication interface (410) may be linked with an external advertising platform (e.g., Meta (Instagram / Facebook) advertising platform, Google Ads, etc.) to transmit requests for the execution of marketing content and receive performance data such as impressions, clicks, conversions, and advertising expenditure amounts that occur after the advertising execution. The communication interface (410) may use a wired communication method (e.g., Ethernet) or a wireless communication method (e.g., Wi-Fi, LTE, 5G, etc.), and the types of communication methods are merely examples and are not limited to this and may vary depending on the network environment.

[0057] The database (412) may refer to a storage means that permanently stores and manages various data required for the operation of the system (400). The database (412) may perform a role distinct from memory (430). While memory (430) performs the role of storing temporary data for instructions and computation processes for the operation of the processor (420), the database (412) may perform the role of permanently storing core data assets that the system (400) accumulates and utilizes over the long term. Data stored in the database (412) may include industry-specific funnel template data, subscriber business profile data, advertising performance indicator data by marketing campaign, anonymized industry-specific benchmark datasets, subscriber billing information and settlement history data, brand guideline information, and analysis history data of AI reverse discussions. The database (412) may be implemented as a relational database (e.g., PostgreSQL) or a non-relational database (e.g., MongoDB), and may be built using a cloud-based database service (e.g., AWS RDS, AWS S3). The types and implementation methods of databases are merely examples and are not limited to this; they may vary depending on the system design.

[0058] The input / output interface (414) may refer to an interface for a user to input data into the system (400) or to visually provide results generated by the system (400) to the user. The input / output interface (414) may be implemented in the form of a web-based dashboard or a mobile application. In terms of input, the input / output interface (414) may provide an input form for the user to input business profiles such as industry category, monthly sales volume, and target customer information, an input window for inputting marketing strategy information as text, and a file upload function for uploading brand logos or image files. In terms of output, the input / output interface (414) may visually display to the user a customized marketing funnel structure generated by the system (400), a strategy review report of AI reverse discussion, a preview of automatically generated marketing content, a subscription settlement report, a marketing performance dashboard, and risk alert notifications. The input / output interface (414) may be implemented as a responsive web in the form of a React.js-based SPA (Single Page Application) and may include a customer dashboard and an administrator panel. The implementation technology of the interface is merely an example and is not limited to this; it may vary depending on the development environment.

[0059] The processor (420) may refer to a computing device that controls the overall operation of the system (400) and performs various operations and data processing by executing instructions stored in memory (430). The processor (420) may include one or more CPUs (Central Processing Units), GPUs (Graphics Processing Units), or NPUs (Neural Processing Units). By executing instructions stored in memory (430), the processor (420) can perform core functions of the system (400), including funnel and strategy analysis, automatic content generation, subscription and settlement control, and performance data optimization.

[0060] The processor (420) can functionally execute multiple software modules. The software modules executed by the processor (420) may include a funnel and strategy analysis module, a content automatic generation module, a subscription and settlement control module, and a performance data optimization module.

[0061] The funnel and strategy analysis module can be responsible for generating a customized marketing funnel structure based on business profile and marketing strategy information received from the user, and for analyzing churn factors by performing AI reverse analysis. The funnel and strategy analysis module can generate a funnel structure by searching for industry-specific funnel templates stored in the database (412), and can perform reverse funnel analysis based on virtual consumer personas using a large-scale language model. The funnel and strategy analysis module can generate the analysis results in the form of a strategy review report and provide it to the user through the input / output interface (414).

[0062] The content automatic generation module can be responsible for generating advertising images and advertising copy suitable for each stage of the marketing funnel by running an image generation model (e.g., Stable Diffusion) and a text generation model (e.g., GPT-4). The content automatic generation module can generate content that maintains brand consistency by referring to brand guideline information stored in the database (412), and can produce channel-specific content in accordance with the specifications of each advertising channel. The content automatic generation module can perform brand consistency verification on the generated content and confirm the content that passes the verification as being ready for distribution.

[0063] The subscription and settlement control module manages billing information of subscription customers, processes automatic payment and collection of subscription fees by calling the API of the PG payment system through the communication interface (410), and can be responsible for calculating the settlement amount based on the revenue distribution between the service provider and the platform operator. The subscription and settlement control module can generate settlement results by separating them into a settlement report for customers and a settlement report for providers, and provide them to each recipient through the input / output interface (414).

[0064] The performance data optimization module can perform the function of analyzing performance data received from an external advertising platform through a communication interface (410), calculating industry-specific marketing performance indicators, and optimizing the marketing funnel structure and content based on the calculated performance indicators. The performance data optimization module can execute a feedback loop that updates the parameters of the AI ​​model by comparing the prediction results of the AI ​​reverse discussion with actual performance data. The performance data optimization module can also perform the function of detecting marketing risks early and generating preemptive response strategies by analyzing the time-series change patterns of the performance data. The performance data optimization module can also perform the function of building and updating industry-specific benchmark datasets and supporting the sharing of high-performance strategies among subscribers through cross-learning.

[0065] The configuration and names of the software modules executed by the processor (420) are merely examples and are not limited thereto, and the modules may be integrated or subdivided depending on the design of the system.

[0066] Memory (430) may refer to a storage means for storing instructions executed by the processor (420) and temporary data generated during the computation process. Memory (430) may include volatile memory (e.g., RAM) and non-volatile memory (e.g., ROM, flash memory). When the instructions stored in memory (430) are executed by the processor (420), they can control the system (400) to perform a series of operations including the collection of business data, the creation of a marketing funnel structure, the execution of AI reverse discussion, the automatic generation of marketing content, the automation of subscription-based settlement processing, and performance data-based optimization. The specific details of the instructions stored in memory (430) will be described in detail through FIGS. 5, 6, 7, and 8.

[0068] FIG. 5 illustrates the overall operation flow of a customized brand consulting and advertising AI platform for small and medium-sized enterprises according to one embodiment.

[0069] The operations described through FIG. 5 can be implemented based on instructions that can be stored in memory (e.g., memory (420) of FIG. 4). The order of each operation of FIG. 5 may be changed, some operations may be omitted, and some operations may be performed simultaneously.

[0070] In operation 510, the system (400) can collect industry information and business data and create a marketing funnel structure.

[0071] Specifically, the system (400) can receive a business profile from a small or medium-sized business or a small business owner that includes an industry category, monthly sales volume, main sales channels, and the age group and interests of target customers. For example, if a business owner in the beauty industry with a monthly sales volume of 30 million won targets women in their 20s and 30s and uses Instagram as a main sales channel, the system (400) can organize this information into a business profile.

[0072] The system (400) may possess a funnel template database classified and accumulated by industry from the marketing execution history of existing subscribers. The funnel template database may store the funnel structure, channel configuration, and conversion performance data of advertising campaigns actually executed across multiple industries during the operation of a subscription marketing service, classified by industry category and sales volume range. The system (400) may search the database for a funnel template corresponding to the industry category and sales volume range of the received business profile.

[0073] The system (400) can generate a customized marketing funnel structure including an awareness stage, an interest stage, a conversion stage, and a repurchase stage by reflecting the target customer characteristics of the business profile in the searched funnel template. Through this, the system (400) can support small and medium-sized enterprises or small business owners without marketing expertise to establish a systematic marketing strategy that reflects the practical experience of experts.

[0074] In operation 520, the system (400) can perform AI reverse discussion to derive churn factors and automatically generate marketing content.

[0075] Specifically, the system (400) can perform an AI reverse discussion in which an AI model analyzes each stage of the marketing funnel in reverse from a consumer's perspective regarding the marketing strategy entered by the user. In the AI ​​reverse discussion, the system (400) can identify churn factors that may lower the consumer's purchase intent at each stage of the funnel and generate improvement measures for the identified churn factors. The AI ​​reverse discussion function may be designed not only to generate content but also to check for vulnerabilities in the strategy in advance at the stage prior to execution of the marketing strategy.

[0076] The system (400) can automatically generate marketing content using an image generation model and a text generation model based on a marketing funnel structure that reflects the results of the analysis of the above-mentioned dropout factors. The image generation model may be a generative AI model such as Stable Diffusion, and the system (400) can generate an advertising image by providing the user's industry information and brand identity as input conditions. The text generation model may be a large-scale language model of the GPT family, and the system (400) can generate advertising copywriting corresponding to consumer behavior goals at each stage of the marketing funnel. The system (400) can produce channel-specific marketing content, including social media advertisements, banner advertisements, and landing page content, by combining the generated advertising images and advertising copywriting.

[0077] In operation 530, the system (400) can collect and analyze performance data to optimize the funnel structure and content.

[0078] Specifically, the system (400) can collect performance data including impressions, clicks, conversions, and advertising expenditure from an advertising platform after the marketing content has been executed. The system (400) can calculate industry-specific marketing performance indicators based on the collected performance data. The industry-specific marketing performance indicators may include click conversion rates by advertising channel, acquisition costs per customer, and a sales return rate relative to advertising costs.

[0079] The system (400) can classify and accumulate the calculated performance indicators in the funnel template database according to industry category and sales volume ranges. The system (400) can analyze the accumulated performance indicators to identify ranges with low conversion rates at each funnel stage and derive optimization measures for the identified ranges, including changing content types, reallocating advertising channels, or modifying appeal messages. By applying the derived optimization measures to the marketing funnel structure and content, the system (400) can implement a data-driven sales cycle structure in which the accuracy of the marketing strategy continuously improves as data is accumulated during the operation of the subscription service.

[0080] The system (400) collects industry information and business data of small and medium-sized enterprises or small business owners, creates a marketing funnel structure based on the collected industry information and business data, performs AI reverse analysis to analyze the funnel in reverse from a consumer perspective regarding business ideas or marketing strategies entered by the user to derive purchase journey and churn factors, automatically generates marketing content based on the derived analysis results, automates subscription-based settlement processing, collects and analyzes performance data resulting from the execution of marketing content to calculate industry-specific marketing performance indicators, and optimizes the marketing funnel structure and content based on the calculated performance indicators. Industry information may refer to information classifying the field of business conducted by the business owner. For example, industry categories such as beauty / cosmetics, F&B (food and beverage), consumer goods, distribution, and education may apply. The above business data encompasses quantitative and qualitative information related to the business operations of the business owner and may include monthly sales volume, major sales channels, target customer age groups and interests, history of past advertising execution, and information on major competitor products. The system (400) can receive the above-mentioned industry information and business data through an input form at the time the user joins the platform, or automatically collect them by linking with an external database based on the business registration number.

[0081] A marketing funnel structure refers to a structure that models the consumer purchase journey step by step, from the point where potential customers first become aware of a brand until the final purchase. Generally, a marketing funnel takes the form of a funnel where the number of potential customers decreases as one moves from higher stages to lower stages, and may consist of an awareness stage, an interest stage, a conversion stage, and a retention stage. The system (400) can analyze the collected industry information and business data to automatically generate a funnel structure suitable for the corresponding industry and sales volume. The generation of the funnel structure may be performed based on funnel templates accumulated by industry from the marketing execution history of existing subscribers. The funnel template may be data that standardizes and stores the stage configuration of a marketing funnel that was actually executed in a specific industry and sales volume range and achieved valid results, the content type for each stage, and the combination of advertising channels.

[0082] Here, AI reverse discussion refers to an analysis method in which an AI model reviews a marketing strategy established by a user from the consumer's perspective to identify weaknesses and areas for improvement in the strategy. While existing marketing analysis methods are limited to designing a funnel in a forward direction from the business owner's perspective, the AI ​​reverse discussion can analyze factors that may cause a consumer to drop out at each stage by tracing back each stage from the point where the consumer makes a final purchase decision to the point of initial awareness. The system (400) can implement the AI ​​reverse discussion by utilizing a Large Language Model (LLM). The system (400) can set a virtual consumer persona in the Large Language Model that reflects the demographic characteristics and consumption tendencies of the target customer group, and configure prompts to simulate scenarios in which the virtual consumer persona maintains its purchase intent or drops out at each funnel stage. As a result of the simulation, the dropout probability, types of dropout factors, and improvement measures for each funnel stage can be derived.

[0083] The purchase journey may refer to a series of touchpoints and decision-making processes experienced by a consumer from the moment they become aware of a specific product or service until they complete the purchase and repurchase. The aforementioned dropout factors may refer to the causes where a consumer's intention to purchase decreases or is discontinued at a specific stage of the purchase journey. These dropout factors may include, but are not limited to, a discrepancy between content and the appeal message, the absence of differentiated value compared to competing products, unclear pricing information, or the complexity of the purchasing process, and may vary depending on the industry and the characteristics of the target customer.

[0084] The system (400) can automatically generate marketing content based on the results of the analysis of churn factors derived above. The automatic generation of the marketing content can be performed by combining an image generation model and a text generation model. The image generation model may be a diffusion-based generative AI model such as Stable Diffusion, and can generate advertising images that match the brand by receiving industry information and brand identity as conditional inputs. The text generation model may be a large-scale language model such as GPT-4, and can generate advertising copy suitable for each stage by receiving consumer behavior goals and key appeal messages at each funnel stage as inputs. For funnel stages identified as vulnerabilities in the analysis of churn factors, the system (400) can generate content in a way that compensates for the corresponding churn factor. For example, if it is analyzed that the correlation between the appeal message and the content is low at a specific stage, the system (400) can reduce the churn rate at that stage by generating copywriting and images that directly reflect the appeal message.

[0085] The system (400) can automatically charge and collect monthly subscription fees from subscribers according to a preset payment cycle by linking with the API of a Payment Gateway (PG) payment system. The system (400) can automatically settle the collected subscription fees according to a preset profit sharing ratio between a marketing service provider that has entered the platform and a platform operator. Through the above-mentioned settlement automation, the management personnel and time required for the operation of the subscription service can be reduced, and settlement errors can be prevented.

[0086] The system (400) can collect and analyze performance data generated after the marketing content is executed through an advertising platform. The performance data refers to quantitative measurements generated as a result of advertising execution and may include impressions, clicks, conversions, advertising expenditure amounts, etc. The system (400) can calculate industry-specific marketing performance indicators from the collected performance data. The industry-specific marketing performance indicator may be an indicator representing the average performance level calculated by aggregating performance data accumulated from multiple subscribers belonging to the same industry on an industry-by-industry basis. The performance indicator may include, but is not limited to, click-through rate (CTR), cost-acquisition cost (CAC), and return on advertising spend (ROAS) per advertising channel, and additional indicators may be included depending on the purpose of analysis.

[0087] The system (400) can optimize the marketing funnel structure and content based on the performance indicators calculated above. Optimization may refer to a process of analyzing performance indicators at each funnel stage to identify stages with low conversion rates or high dropout rates, and adjusting the content type, advertising channels, or budget allocation for those stages. As the performance data is continuously accumulated during the operation of the subscription service, the system (400) can update the funnel template database, and the updated funnel template may be reflected in the creation of the funnel structure for new subscribers thereafter. Through this process, the system (400) can implement a data-based revenue cycle structure in which the accuracy of the marketing strategy improves as data is accumulated. The data-based revenue cycle structure may refer to a structure in which marketing efficiency is gradually improved as the execution of marketing content, collection of performance data, optimization of funnels and content, and re-execution of optimized content are repeatedly cycled.

[0088] According to one embodiment, the system (400) receives a business profile from a small and medium-sized enterprise or a small business owner that includes an industry category, monthly sales volume, major sales channels, and the age group and interests of target customers; searches for a funnel template corresponding to the industry category and sales volume range of the received business profile from a funnel template database classified and accumulated by industry from the marketing execution history of existing subscribers; generates a customized marketing funnel structure including an awareness stage, an interest stage, a conversion stage, and a repurchase stage by reflecting the target customer characteristics of the business profile in the searched funnel template; for each stage of the generated marketing funnel structure, selects a content type suitable for that stage based on the industry category and the age group of target customers, matching social media video content to the awareness stage, a detailed information landing page to the interest stage, a discount or promotion advertisement to the conversion stage, and a retargeting advertisement to the repurchase stage, respectively; calculates a total advertising budget proportional to the monthly sales volume for each matched content type; and calculates an industry-specific funnel stage that calculates the average value of the ratio in which each funnel stage contributed to the final conversion from the marketing execution results of existing subscribers within the same industry category accumulated in the funnel template database. By referring to attribution data, you can establish an execution strategy to allocate the total advertising budget according to the attribution ratio of each funnel stage.

[0089] The system (400) may receive a business profile from a small or medium-sized enterprise or a small business owner that includes an industry category, monthly sales volume, major sales channels, and the age group and interests of target customers. A business profile may refer to data that organizes the business characteristics of the business owner in a structured form. The industry category may be classified into beauty / cosmetics, F&B, consumer goods, distribution, education, etc., but is not limited thereto, and the industry classification system may be expanded or changed depending on the settings of the platform. The monthly sales volume may refer to the average monthly sales volume of the business owner over a recent period, and the system (400) may classify and process the monthly sales volume according to pre-set intervals. Major sales channels refer to online or offline channels through which the business owner sells products or services, and may include the company's own shopping mall, open market, social media, or offline store. The system (400) may have the user directly input the business profile through an input form when signing up for the platform, or may automatically configure it from existing business information.

[0090] The system (400) can search for funnel templates corresponding to the industry category and sales volume range of the received business profile from a funnel template database classified and accumulated by industry from the marketing execution history of existing subscribers. The funnel template database can be constructed from data accumulated by the system (400) in the process of operating a subscription-based marketing service. The system (400) can classify the funnel structure of marketing campaigns actually executed for subscribers, the content type for each stage, the combination of advertising channels, and the conversion performance of the corresponding campaign by industry category and sales volume range, and store them in the database. The search can be performed by querying funnel templates with identical or similar conditions from the database using the new user's industry category and sales volume range as a search key. If multiple funnel templates exist as search results, the system (400) can prioritize selecting the template with the highest conversion performance or present multiple templates to the user for selection.

[0091] The system (400) can generate a customized marketing funnel structure including an awareness stage, an interest stage, a conversion stage, and a repurchase stage by reflecting the target customer characteristics of the business profile in the searched funnel template. The generation of the customized marketing funnel structure can be performed by maintaining the basic structure of the searched funnel template while adjusting the detailed configuration of each stage according to the age group and interests of the target customers. For example, if the target customers are in their 20s and frequently use social media, the system (400) can adjust the funnel structure to increase the proportion of social media-based short-form video content in the awareness stage. On the other hand, if the target customers are in their 40s or older and prefer search-based information exploration, the system (400) can adjust the structure to increase the proportion of blog or review content in the interest stage. The configuration of each stage can be varied in many ways depending on the characteristics of the target customers and may not be limited to the examples.

[0092] The system (400) can establish an execution strategy to select content types for each stage of the generated marketing funnel structure and allocate advertising budgets. The selection of content types can be performed by matching the most suitable content form to each funnel stage based on industry categories and the age group of target customers. The system (400) can match social media video content with high brand exposure effects to the awareness stage, landing pages delivering detailed information about products or services to the interest stage, discount or promotion advertisements that induce purchasing behavior to the conversion stage, and retargeting advertisements targeting existing customers to the repurchase stage. The matching of content types and funnel stages is merely an example and may vary depending on the industry and the characteristics of target customers.

[0093] The allocation of the advertising budget can be performed in two stages. First, the system (400) can calculate the total advertising budget in proportion to the monthly sales volume included in the business profile. The calculation method may involve applying a pre-set ratio of advertising costs to sales for each industry. Next, the system (400) can calculate the average value of the ratio in which each funnel stage contributed to the final conversion by referring to the marketing execution results of existing subscribers within the same industry category accumulated in the funnel template database. The calculated average value may correspond to the industry-specific funnel stage contribution data. By allocating the total advertising budget according to the contribution ratio of each funnel stage, the system (400) can establish an execution strategy so that a relatively large budget is allocated to stages with high conversion contribution. Through the establishment of an execution strategy, small and medium-sized enterprises or small business owners without marketing expertise can perform reasonable budget allocation based on data.

[0095] FIG. 6 illustrates an AI reverse discussion execution operation according to one embodiment.

[0096] The operations described through FIG. 6 can be implemented based on instructions that can be stored in memory (e.g., memory (420) of FIG. 4). The order of each operation of FIG. 6 may be changed, some operations may be omitted, and some operations may be performed simultaneously.

[0097] In operation 610, the system (400) can receive marketing strategy information from the user.

[0098] Specifically, the system (400) can receive marketing strategy information from a user in the form of text, including the characteristics of the product or service to be marketed, the expected target customer base, major advertising channels, and key appeal messages. For example, if a small business owner in the beauty industry wants to run an advertisement on an Instagram channel targeting women in their 20s with natural ingredients as the key appeal point for the launch of a new skincare product, the system (400) can receive the above content as marketing strategy information. The system (400) can parse the received marketing strategy information into a structured data format and use it as input data for a subsequent AI reverse discussion.

[0099] In operation 620, the system (400) can generate a virtual consumer persona that reflects the characteristics of the target customer base.

[0100] Specifically, the system (400) can generate a virtual consumer persona based on the age group, gender, interests, and consumption pattern information of the expected target customer group included in the received marketing strategy information. The virtual consumer persona may be a virtual consumer personality simulated by a large-scale language model that reflects the demographic characteristics and consumption tendencies of the target customer group. The system (400) can set the purchase decision pattern and churn tendency parameters of the virtual consumer persona by utilizing consumer response data of existing customers in the same industry accumulated in the funnel template database as training data. The system (400) can generate multiple consumer type personas, such as price-sensitive, quality-oriented, or trend-following types, depending on the characteristics of the target customer group, so that multifaceted analysis can be performed.

[0101] In operation 630, the system (400) can identify the risk of churn by having a virtual consumer persona navigate the funnel in reverse order.

[0102] Specifically, the system (400) may enable the generated virtual consumer persona to navigate each stage in reverse order from the final conversion stage of the marketing funnel toward the awareness stage. During the reverse navigation process, the system (400) may evaluate whether the information or motivation necessary to maintain the consumer's purchase intent is satisfied at each funnel stage. During the evaluation process, the system (400) may identify a stage where the virtual consumer persona's purchase intent is judged to decrease as a dropout risk zone.

[0103] In identifying the above-mentioned churn risk section, the system (400) may calculate the correlation between the core appeal message and the content provided at the corresponding stage, and classify the case where the calculated correlation is less than a preset threshold as a first churn factor. Additionally, the system (400) may classify the case where the differentiated value compared to a competing product is not specified at the corresponding stage as a second churn factor. The calculation of the correlation may be performed by calculating the cosine similarity between the semantic vector of the core appeal message and the semantic vector of the content at the corresponding stage using a text embedding model.

[0104] In operation 640, the system (400) can generate improvement plans for each deviation factor and provide a strategy review report.

[0105] Specifically, the system (400) can generate improvement plans corresponding to each of the classified churn factors. For the first churn factor, it can generate a plan to reinforce the content of the corresponding stage so that its relevance to the core appeal message is improved. For the second churn factor, it can generate a plan that includes changing the appeal message to emphasize differentiation points from competing products or adding comparison content. Additionally, the system (400) can generate a plan to switch the advertising channel itself when multiple churn factors overlap at a specific funnel stage.

[0106] The system (400) can generate a strategy review report that maps the above-mentioned dropout risk zones and dropout factors and improvement measures corresponding to each zone to each funnel stage. The strategy review report can be configured to quantify and visually display the dropout risk of each funnel stage, and to organize the types of dropout factors and improvement measures by stage. By providing the generated strategy review report to the user, the system (400) can support the user in identifying and supplementing weaknesses in the strategy in advance before executing the marketing strategy.

[0107] According to one embodiment, the system (400) performs an AI reverse discussion by receiving marketing strategy information in text form from a user, including the characteristics of a product or service to be marketed, the expected target customer base, major advertising channels, and a key appeal message; based on the received marketing strategy information, an AI model creates a virtual consumer persona that reflects the demographic characteristics and consumption propensity of the expected target customer base; the created virtual consumer persona explores each stage in reverse order from the final conversion stage to the awareness stage of the marketing funnel, evaluates whether the information or motivation necessary to maintain the purchase intention is satisfied at each stage, identifies a stage where the purchase intention of the virtual consumer persona is judged to decrease during the evaluation process as a churn risk section, classifies cases where the correlation between the key appeal message and the content provided at that stage is below a preset threshold or where a differentiated value compared to a competing product is not specified as a churn factor, generates improvement measures including content reinforcement, change of appeal message, or switching of advertising channels for each classified churn factor, and generates a strategy check report that maps the churn risk section and improvement measures to the funnel stage and provides it to the user.

[0108] In performing an AI reverse discussion, the system (400) may receive marketing strategy information in text form from a user, including features of the product or service to be marketed, expected target customer base, major advertising channels, and key appeal messages. Marketing strategy information may refer to data describing the key elements of a marketing plan established by the user in text form. Features of the product or service to be marketed may include the product's main functions, ingredients, price range, or differentiations from competing products. Key appeal messages refer to the core marketing message intended to be delivered to the target customer and may correspond to the product's value proposition. The system (400) may convert the received marketing strategy information into structured data by parsing the product features, target customer base, advertising channels, and appeal message items, respectively, through natural language processing. The structured data may be used as input for a subsequent AI reverse discussion.

[0109] The system (400) can generate a virtual consumer persona based on the received marketing strategy information, in which an AI model reflects the demographic characteristics and consumption tendencies of the expected target customer group.

[0110] A virtual consumer persona may refer to a virtual consumer profile that models the typical consumption behavior patterns, purchase motives, and information search habits of a specific target customer group. The system (400) can generate a virtual consumer persona by providing the target customer group's age group, gender, interests, income level, and major consumption channel information as prompts to a large-scale language model. The system (400) can set the purchase decision patterns and churn propensity parameters of the virtual consumer persona by utilizing the ad click patterns, purchase conversion history, and churn point data of existing customers within the same industry accumulated in the funnel template database as training data. A churn propensity parameter may refer to a numerical value representing the weight of each factor that influences the probability of a consumer churning at each funnel stage. The system (400) can generate multiple consumer type personas, such as price-sensitive, quality-oriented, or trend-following types, depending on the characteristics of the target customer group. The classification of consumer types is an example and may vary depending on the characteristics of the industry and product.

[0111] The system (400) can evaluate whether the information or motivations necessary to maintain purchase intent are satisfied at each stage as the generated virtual consumer persona navigates each stage in reverse order from the final conversion stage of the marketing funnel toward the awareness stage. Here, reverse exploration refers to an analysis method that traces backward from the point where the consumer completed the purchase to determine what information should have been provided and what motivations should have been satisfied at each stage of the funnel leading up to that purchase. Unlike forward funnel design, the reverse exploration method can have the effect of verifying whether the logical consistency of the strategy and the links between each stage are sufficient by deriving the essential conditions for the final conversion to succeed in reverse for each stage. The system (400) can assign the role of the virtual consumer persona to a large-scale language model and configure prompts to infer the type and level of information that the persona needs to maintain purchase intent at each stage in a question-and-answer format. The system (400) can determine whether the information or motivations required at each stage are sufficiently provided by the current marketing strategy, and if they are not satisfied, identify that stage as a dropout risk zone.

[0112] In identifying the dropout risk zone, the system (400) may classify cases where the correlation between the core appeal message and the content provided at that stage is below a preset threshold as a dropout factor. Correlation may refer to a numerical value of the semantic similarity between two texts. The system (400) may convert the core appeal message into a semantic vector using a text embedding model, convert the content provided at that funnel stage into a semantic vector as well, and then calculate the correlation by calculating the cosine similarity between the two vectors. If the cosine similarity is below a preset threshold, the system (400) may determine that consumer interest may decrease because the appeal message is not effectively reflected in the content at that stage. Additionally, the system (400) may classify cases where the differentiated value compared to competing products is not specified at that stage as a separate dropout factor. Whether the differentiated value compared to competing products is specified can be determined by comparing the product features included in the marketing strategy information with the existence of differentiating elements mentioned in the content at that stage.

[0113] The system (400) can generate improvement plans for each classified churn factor, including content reinforcement, changes to appeal messages, or switching advertising channels. The generation of improvement plans may be performed in different ways depending on the type of churn factor. If the churn factor is a low correlation between the appeal message and the content, the system (400) can generate a plan to add key keywords of the appeal message to the content of that stage or to restructure the appeal message itself to align with the consumer behavior goals of that stage. If the churn factor is the absence of differentiated value compared to competing products, the system (400) can generate a plan to add product comparison content or to change the appeal message to one that emphasizes differentiating points. If multiple churn factors overlap at a specific funnel stage, the system (400) can generate a plan to switch the advertising channel currently in use at that stage to a channel with a higher contact frequency with the target customer.

[0114] The system (400) can generate and provide a strategy review report to the user that maps the churn risk zones and improvement measures to each funnel stage. A strategy review report may refer to a report that organizes the analysis results of AI reverse discussion in a form that the user can understand. The strategy review report may be structured to display the churn risk level for each funnel stage as a numerical value or grade, and organize the types of churn factors identified at each stage and the corresponding improvement measures step by step. The system (400) can visually provide the strategy review report to the user through the platform's dashboard. By providing the strategy review report, the user can identify and supplement weaknesses in the strategy in advance before actually executing the marketing strategy, and can achieve the effect of conducting an expert-level strategy review without the need for marketing professionals.

[0115] According to one embodiment, the system (400) automatically generates marketing content by receiving and storing brand guideline information including a user's brand logo, representative color, font, and tone and manner, providing industry information, the corresponding stage of the marketing funnel, and brand guideline information as input conditions to an image generation model to generate an advertising image that conforms to the brand's representative color and tone and manner, providing consumer behavior goals and key appeal messages for the corresponding stage of the marketing funnel as input conditions to a text generation model to generate advertising copywriting in stages, including short hooking phrases to enhance brand awareness in the awareness stage and phrases emphasizing benefits to induce purchasing behavior in the conversion stage, combining the generated advertising image and advertising copywriting according to the specifications of the advertising channel, and producing channel-specific content according to the resolution, text placement ratio, and file format requirements of each of the social media feed ads, story ads, banner ads, and landing page content, and verifying the color match, font usage, and tone and manner consistency with the stored brand guideline information for each produced channel-specific content, and confirming the content that is equal to or greater than a preset brand consistency score as a distributable state. Regeneration can be performed for content that is less than [amount].

[0116] The system (400) can receive and store brand guideline information, including the user's brand logo, representative color, font, and tone and manner, when automatically generating marketing content. Here, brand guideline information may refer to a set of standards that define the visual identity and communication method of a specific brand. The representative color may include color codes for the primary color and secondary color that represent the brand. Tone and manner defines the atmosphere and tone that the brand intends to convey to consumers, and may be set, for example, as a friendly and casual style, a luxurious and restrained style, or a professional and trustworthy style. The system (400) may configure brand guideline information by allowing the user to directly input brand guideline information into the platform, or by automatically extracting the representative color and font from existing brand assets (logo image, website, etc.) when the user uploads them.

[0117] The system (400) can generate advertising images that match the brand's representative color and tone and manner by providing industry information, the corresponding stage of the marketing funnel, and brand guideline information as input conditions to an image generation model. The image generation model may include a diffusion-based generative AI model such as Stable Diffusion. When configuring prompts for the image generation model, the system (400) may include visual elements corresponding to the industry category, an image atmosphere suitable for the funnel stage, and the representative color and tone and manner of the brand guidelines as conditioning inputs. For example, the prompts can be configured to generate product images with a bright and radiant atmosphere at the awareness stage of the beauty industry, and images with promotion information visually emphasized at the conversion stage. The system (400) can set color conditions so that the brand's representative color is reflected as the main tone of the image when generating images, and can automatically composite the brand logo onto the generated image at a pre-set location.

[0118] The system (400) can generate advertising copy for each stage of the marketing funnel by providing the consumer behavior goal and key appeal message for the corresponding stage of the marketing funnel as input conditions to the text generation model. The text generation model may be a large-scale language model such as GPT-4. The consumer behavior goal may refer to the behavior intended to be induced in the consumer at each stage of the funnel. The consumer behavior goal at the awareness stage may be to recognize the existence of the brand and generate interest, and the consumer behavior goal at the conversion stage may be to induce actual purchasing behavior. The system (400) can generate advertising copy by providing the consumer behavior goal, key appeal message, and tone and manner of the brand guidelines for the corresponding stage of the funnel as prompts to the text generation model. The system (400) can generate short hooking phrases to attract the consumer's attention at the awareness stage, informative phrases conveying the product's key features at the interest stage, phrases emphasizing benefits that induce purchasing behavior such as limited offers or price discounts at the conversion stage, and thank-you messages for existing customers or phrases encouraging repurchase at the repurchase stage. The types of copywriting for each stage are examples and may vary depending on the industry and characteristics of the target audience.

[0119] The system (400) can produce channel-specific marketing content by combining generated ad images and ad copywriting according to the specifications of the ad channel. Ad channel-specific specifications may refer to technical requirements such as the resolution, aspect ratio, text placement ratio, and file format of the content required by each ad platform. For example, social media feed ads may require images with a square (1:1) or vertical (4:5) ratio, story ads may require full-screen images with a 9:16 vertical ratio, and banner ads may require images with a horizontal (16:9 or 728X90, etc.) ratio. The system (400) stores the specification information of each ad channel in a database, resizes the generated ad images to match the resolution and aspect ratio of the corresponding channel, and combines the ad copywriting with the images within the text placement ratio allowed by each channel. The system (400) can produce multiple channel-specific contents in batches, including social media feed ads, story ads, banner ads, and landing page content. The types and specification requirements for advertising channels are examples and may be added or changed depending on platform settings and changes in advertising platform policies.

[0120] The system (400) can determine the distributable content by verifying the consistency of each produced channel-specific content with the stored brand guideline information. Brand consistency verification may refer to a process of automatically checking whether automatically generated content conforms to the user's brand guidelines. The system (400) can extract the main color distribution of the generated content and calculate the degree of color match with the representative color of the brand guidelines. The calculation of the degree of color match can be performed by calculating the color difference between the main color histogram of the generated image and the brand representative color using a standard color difference formula such as CIE Delta E. The system (400) can also check whether a font specified in the brand guidelines is applied to the generated copywriting and whether the tone and manner matches a pre-set style. The verification of the consistency of the tone and manner can be performed by analyzing the tone of the generated copywriting using a text classification model and determining whether it matches the tone and manner category defined in the brand guidelines. The system (400) calculates a brand consistency score by combining color matching, font usage, and tone and manner consistency, and can determine that content with a score above a preset threshold is in a distributable state. For content with a brand consistency score below the threshold, the system (400) can re-perform image generation or copywriting generation until content that complies with brand guidelines is generated.

[0122] FIG. 7 illustrates a subscription-based settlement automation operation according to one embodiment.

[0123] The operations described through FIG. 7 can be implemented based on instructions that can be stored in memory (e.g., memory (420) of FIG. 4). The order of each operation of FIG. 7 may be changed, some operations may be omitted, and some operations may be performed simultaneously.

[0124] In operation 710, the system (400) can set monthly billing information according to the service grade for each subscriber.

[0125] Specifically, the system (400) can set and store billing information, including a monthly subscription fee and a payment cycle, based on the service tier selected by the subscriber upon signing up. The service tier may be divided into multiple tiers depending on the scope of the marketing services provided, for example, a basic tier focused on design, a middle tier including design and marketing, and a premium tier including design, marketing, performance advertising operations, and data analysis. The system (400) can map and store the monthly subscription fee corresponding to each tier to the subscriber's account information. When setting the billing information, the system (400) can store the payment method information registered by the subscriber together and utilize it for subsequent automatic payment processing.

[0126] In operation 720, the system (400) can process automatic billing and collection of subscription fees by linking with a PG payment system.

[0127] Specifically, the system (400) may be linked with the API of a Payment Gateway (PG) payment system. The PG payment system may be an external payment service provider such as Toss Payments. The system (400) may identify a subscriber whose payment cycle set in the stored billing information has arrived, and may send an automatic payment request for the monthly subscription fee of the corresponding tier to the identified subscriber through the API of the PG payment system using a registered payment method. The system (400) may receive a payment approval or payment failure response from the PG payment system. If a payment failure response is received, the system (400) may request payment again according to a preset retry cycle or send a notification to the subscriber instructing them to change their payment method.

[0128] In operation 730, the system (400) can calculate a settlement amount based on the revenue sharing ratio between the service provider and the platform operator.

[0129] Specifically, the system (400) can calculate the respective settlement amounts by applying a pre-set revenue sharing ratio between a marketing service provider and a platform operator to the subscription fee for which payment approval has been received. The marketing service provider may be a professional or agency that provides marketing, design, or advertising operation services to subscribers through the platform. The system (400) may set the revenue sharing ratio individually for each service provider, and different sharing ratios may be applied depending on the service grade and the type of service provided. The system (400) may record the settlement amount for the service provider's share and the settlement amount for the platform operator's share, respectively, calculated according to the revenue sharing ratio.

[0130] In operation 740, the system (400) can aggregate subscription revenue, project revenue, and platform fees by item.

[0131] Specifically, the system (400) can aggregate the calculated settlement amount by classifying it according to sales type. The system (400) can classify revenue from monthly subscription services as subscription revenue, revenue from single-projects performed separately from subscription services as project revenue, and commission revenue from platform operation as platform commission. The system (400) can aggregate the amounts by classified sales type according to a pre-set settlement cycle. The settlement cycle may be monthly, weekly, or a custom cycle set by the user. During the aggregation process, the system (400) can subdivide and track revenue by subscriber, service provider, and service grade, and can store the subdivided revenue data in a settlement database.

[0132] In operation 750, the system (400) can generate and provide a settlement report for the customer and a settlement report for the provider, respectively.

[0133] Specifically, the system (400) can generate settlement reports of different forms depending on the type of recipient based on the aggregated settlement data. For a subscription customer, the system (400) can generate a customer settlement report that includes the subscription customer's payment history, service items used, service grade, and payment amount. The customer settlement report can be configured so that the subscription customer can check their payment history and service usage status through the platform's dashboard.

[0134] The system (400) can generate a settlement report for a service provider that includes the amount of revenue distributed to the service provider, details of revenue distribution, the scheduled deposit date, and the scheduled deposit amount. The settlement report for the provider can be configured so that the service provider can check their settlement status in real time and identify the scheduled deposit amount in advance. The system (400) can provide the generated settlement report for customers and the settlement report for providers to their respective recipients via a platform dashboard, email, or notification message.

[0135] In automating subscription-based settlement processing, the system (400) can set and store billing information including a monthly subscription fee and a payment cycle according to the service grade for each subscriber. A service grade may refer to a classification of subscription products categorized according to the scope and level of marketing services provided by the platform. Service grades may consist of a basic grade focused on design content creation, a middle grade including design and marketing strategy formulation, and a premium grade including design, marketing, performance advertising operation, and data analysis. The types of grades and the scope of services included in each grade are examples and may change according to the platform's operational policy. The monthly subscription fee may refer to a pre-set monthly usage fee corresponding to each service grade. The payment cycle refers to the interval at which the subscription fee is automatically charged; while the default value may be monthly, it may also be set to quarterly or annually depending on the user's choice. The system (400) can generate billing information and store it in a database at the time when a subscriber selects a service grade and registers a payment method. Billing information may include subscriber identification information, selected service tier, monthly subscription fee, billing cycle, registered payment method information, and the first billing date.

[0136] The system (400) can process the automatic billing and collection of subscription fees for subscribers whose pre-set payment cycle has arrived, using a registered payment method by linking with the API of a PG payment system. A PG (Payment Gateway) payment system may refer to an external payment service provider that mediates online payments. The system (400) can be linked with a recurring payment API provided by a PG payment system such as Toss Payments. The system (400) can identify subscribers whose payment cycle has arrived by periodically querying billing information stored in a database. The system (400) can call the API of the PG payment system for the identified subscribers to request payment approval for the monthly subscription fee of the corresponding tier. When a payment approval response is received from the PG payment system, the system (400) can record the payment completion details in the database and send a payment completion notification to the subscribers. When a payment failure response is received from the PG payment system, the system (400) may request payment again according to the preset number of retries and retry interval. If payment is not completed even after exceeding the number of retries, the system (400) may send a notification to the subscriber via email or text message instructing them to change their payment method.

[0137] The system (400) can calculate the respective settlement amounts for subscription fees for which payment approval has been received, according to a pre-set revenue sharing ratio between a marketing service provider and a platform operator who have entered the platform. A marketing service provider may refer to a professional or agency that directly performs marketing, design, or advertising operation services for subscribers through the platform. A revenue sharing ratio may refer to the ratio of the amount distributed to the service provider and the platform operator, respectively, from the subscription fee revenue. The system (400) can individually set and store revenue sharing ratios for each service provider, and different distribution ratios may be applied depending on the service grade or the type of service provided. For subscription fees for which payment has been completed, the system (400) can identify the service provider that provides services to the subscriber and calculate the settlement amount for the service provider and the settlement amount for the platform operator, respectively, according to the revenue sharing ratio applied to the service provider.

[0138] The system (400) can aggregate the calculated settlement amount by classifying it into subscription revenue, project revenue, and platform fees. Subscription revenue may refer to regular revenue generated from the use of a monthly subscription service. Project revenue may refer to irregular revenue generated from projects commissioned individually, such as brand design, video production, or special campaigns, separate from the subscription service. Platform fees may refer to the amount received by a platform operator as compensation for service brokerage. The system (400) can automatically classify for each payment transaction whether the revenue corresponds to subscription revenue, project revenue, or platform fees. The system (400) can aggregate the amount for each classified revenue type according to a pre-set settlement cycle. The settlement cycle may be monthly by default, but can be changed to a weekly cycle or a custom cycle set by the user. During the aggregation process, the system (400) can subdivide and track revenue by subscriber, service provider, and service grade, and can store the subdivided revenue data in a settlement database.

[0139] The system (400) can generate and provide settlement reports of different forms depending on the type of recipient based on aggregated settlement data. A settlement report for customers is a settlement report provided to a subscriber, which may include the subscriber's payment history, service items used, service level, and payment amount. Subscribers can check their payment history and service usage status through the settlement report for customers. A settlement report for providers is a settlement report provided to a marketing service provider, which may include the amount of revenue distributed to the service provider, details of revenue distribution, scheduled deposit date, and scheduled deposit amount. Service providers can identify their revenue status and deposit schedule in advance through the settlement report for providers. The system (400) can provide the settlement report for customers and the settlement report for providers to their respective recipients via a platform dashboard, email, or notification message. The channels for providing the settlement report are examples and may be added or changed depending on the platform settings. Through the automation of settlement processing, management personnel and time required for the operation of subscription services can be reduced, settlement errors caused by manual work can be prevented, and settlement transparency between service providers and platform operators can be ensured.

[0140] According to one embodiment, the system (400) performs anonymization processing to remove identification information of individual companies, including company name, representative information, and business registration number, from marketing performance data collected from multiple subscribers belonging to the same industry category, and classifies the anonymized performance data by industry category and sales size range to construct an industry-specific benchmark dataset, wherein the industry-specific benchmark dataset aggregates performance data of multiple anonymized companies corresponding to the same industry category and sales size range and functions as reference data representing the average marketing performance level of the industry and size range, and matches the industry category and sales size range extracted from the business profile of new subscribers with a benchmark group, which refers to a group of companies belonging to the same industry category and sales size range in the industry-specific benchmark dataset, calculates a marketing maturity score by comparing the average ad conversion rate, average customer acquisition cost, and average funnel stage churn rate of the matched benchmark group with the new subscribers' current marketing indicators by item, identifies the funnel stage that recorded the lowest score compared to the average value of the benchmark group in the calculated marketing maturity score as a priority target for improvement, and for the identified stage, the performance corresponding to the upper pre-set ratio within the benchmark group By analyzing the combinations of content types and advertising channels commonly applied by successful companies, customized action plans can be automatically generated and recommended for new subscribers.

[0141] The system (400) can perform anonymization processing to remove identification information of individual companies from marketing performance data collected from multiple subscribers belonging to the same industry category, and classify the anonymized performance data by industry category and sales volume range to construct an industry-specific benchmark dataset. Anonymization processing may refer to a process of removing or de-identifying information that can identify a specific company, such as company name, representative information, and business registration number, included in the performance data.

[0142] The system (400) protects individual companies' sales information from being exposed externally through anonymization processing, while also enabling the use of the data for statistical analysis at the industry level. Anonymization methods may include deleting identification information fields, irreversibly transforming data using a hash function, or applying k-anonymity techniques, but are not limited to these, and an appropriate method may be selected according to the data protection policy. An industry-specific benchmark dataset refers to aggregating marketing performance data of multiple anonymized companies belonging to the same industry category and the same sales volume range, and may represent reference data indicating the average level of marketing performance in the corresponding industry and volume range.

[0143] The system (400) can construct a benchmark dataset by first classifying anonymized performance data by industry category and secondarily classifying it by sales volume range within each industry category. The sales volume ranges can be set as ranges such as monthly sales of less than 10 million won, 10 million won or more to less than 50 million won, and 50 million won or more to less than 100 million won, but the range and number of ranges may be changed according to the platform's operational policy. The benchmark dataset may include indicators such as the average ad conversion rate, average customer acquisition cost, average funnel stage dropout rate, and average sales return relative to advertising costs of companies belonging to the group. The system (400) can periodically update the benchmark dataset as the number of subscribed customers increases and performance data accumulates, and the statistical reliability of the benchmark may improve as the amount of accumulated data increases.

[0144] The system (400) can match the industry category and sales volume range extracted from the business profile of a new subscriber with a corresponding benchmark group in the benchmark dataset. A benchmark group may refer to a group of companies belonging to the same industry category and the same sales volume range within the benchmark dataset.

[0145] The system (400) can search for benchmark groups that meet the conditions in the benchmark dataset using the industry category and sales volume range of the new subscriber as search keys. For example, if a new subscriber with monthly sales of 30 million won in the beauty industry joins, the system (400) can match a benchmark group that is in the beauty industry and has monthly sales of 10 million won or more and less than 50 million won. If there is no group in the benchmark dataset that exactly meets the conditions of the new subscriber, the system (400) can match a group with the same industry category and the closest sales volume range as a substitute group, or utilize a group of a similar industry category.

[0146] The system (400) can calculate a marketing maturity score by comparing the average ad conversion rate, average customer acquisition cost, and average funnel stage dropout rate of the matched benchmark group with the current marketing metrics of new subscribers item by item. The marketing maturity score may refer to a metric that quantifies the level at which the current marketing capabilities of new subscribers are positioned compared to the average of companies of the same industry and size.

[0147] The system (400) can calculate the difference between the indicator value of new subscribers and the average value of the benchmark group for each comparison item, and normalize the calculated difference to convert it into an item-specific score. The system (400) can calculate an overall marketing maturity score by weighting the item-specific scores. The weight of each item can be set differently depending on the industry category; for example, in an industry where the conversion rate directly affects sales, the weight of the advertising conversion rate item may be set high. The marketing maturity score can be calculated as a value between 0 and 100 points, but the range of the score and the calculation method may vary depending on the platform settings.

[0148] The system (400) can identify the funnel stage that recorded the lowest score relative to the benchmark group average in the calculated marketing maturity score as a priority improvement target. Identifying the priority improvement target may mean the process of finding the funnel stage where the marketing performance of new subscribers is most significantly lacking compared to the industry average. The system (400) can select the stage with the largest deviation relative to the benchmark group average as a priority improvement target by comparing the item-specific scores in each of the awareness stage, interest stage, conversion stage, and repurchase stage. If similar levels of deviation appear in multiple funnel stages, the system (400) can assign priority to improve sequentially starting from the upper stage of the funnel (awareness stage). This is because if the upper stage of the funnel is not improved, the number of potential customers flowing into the lower stage itself may be insufficient, thereby limiting the improvement effect of the lower stage.

[0149] The system (400) can automatically generate and recommend a customized action plan by analyzing the combination of content types and advertising channels commonly applied by companies that have achieved performance corresponding to a preset top ratio within a benchmark group for the identified priority improvement target stage. The preset top ratio may refer to criteria for selecting a certain percentage of companies with excellent performance indicators within a benchmark group. For example, if the preset top ratio is 20%, the system (400) can select companies within the benchmark group that have a conversion rate within the top 20% of the corresponding funnel stage as high-performance companies.

[0150] The system (400) can analyze patterns of content types, advertising channels, and budget allocation ratios commonly used by high-performing companies in the corresponding funnel stage. Based on the analyzed patterns, the system (400) can generate a customized action plan that includes changes in content types, addition or conversion of advertising channels, and adjustment of budget allocation that new subscribers can apply in the priority improvement stage. A customized action plan may refer to a list of specific marketing improvement actions that a customer needs to execute, based on the results of a comparative analysis of the new subscriber's current marketing situation and best practices from a benchmark group. The system (400) can provide the generated action plan to new subscribers through a platform dashboard and can track whether the action plan is executed and changes in performance after execution to reflect in the creation of subsequent action plans.

[0152] According to one embodiment, after the execution of marketing content is completed, the system (400) calculates performance indicators including the click-through rate per advertising channel, the cost of acquisition per customer, and the sales return relative to advertising costs based on data collected from an advertising platform regarding impressions, clicks, conversions, and advertising expenditure amounts, and calculates the actual churn rate at each funnel stage from the calculated performance indicators, and evaluates whether there is a match between the predicted churn range and the actual churn range at each funnel stage by comparing the stage identified as the churn risk range by the virtual consumer persona in the AI ​​reverse discussion with the stage with a high actual churn rate, and if there is a stage where the predicted churn range and the actual churn range do not match in the evaluation results, collects consumer behavior data at that stage as additional training data to update the consumption propensity parameter and churn judgment threshold of the virtual consumer persona, and uses an AI reverse discussion model to which the updated consumption propensity parameter and churn judgment threshold are applied to repeatedly execute a feedback loop so that the accuracy of the churn prediction is improved compared to before the update when performing a reverse discussion on a subsequent marketing strategy.

[0153] The system (400) can calculate performance metrics based on data collected from an advertising platform regarding impressions, clicks, conversions, and advertising expenditure after the execution of marketing content is completed. Performance metrics may refer to figures that quantitatively measure the performance of marketing content execution. The system (400) can calculate performance metrics including the click-through rate (CTR) per advertising channel, cost of acquisition per customer (CAC), and return on advertising expenditure (ROAS). The click-through rate refers to the ratio of clicks to ad impressions and indicates how effectively the advertising content has attracted consumer attention. The cost of acquisition per customer refers to the value obtained by dividing the total advertising expenditure by the number of actual converted (purchasing) customers and indicates the cost incurred to acquire one new customer. The return on advertising expenditure refers to the ratio of sales generated by the advertisement to the amount of advertising expenditure and indicates the profitability of the advertising investment. The system (400) can calculate performance metrics by subdividing them by advertising channel, funnel stage, and content type.

[0154] The system (400) can derive the actual churn rate for each funnel stage from the calculated performance indicators and compare it with the stage identified as a churn risk zone by the virtual consumer persona in the AI ​​reverse discussion. The actual churn rate for each funnel stage may refer to a value calculated from actual performance data regarding the ratio of consumers who churn without proceeding to the next stage at each funnel stage.

[0155] The system (400) can calculate the dropout rate for each stage by comparing the number of inflows at each funnel stage with the number of moves to the next stage. For example, if 1,000 people were exposed to an advertisement at the awareness stage and 100 of them moved to the interest stage (such as visiting a landing page), the dropout rate for the awareness stage can be calculated as 90%. The system (400) can compare the calculated actual dropout rate for each funnel stage with the stages identified as dropout risk zones by virtual consumer personas during the AI ​​reverse discussion process. Comparison may refer to the process of verifying whether the dropout risk stage predicted by the AI ​​reverse discussion matches the stage where the actual dropout rate is high.

[0156] The system (400) can evaluate whether there is a match between the predicted churn interval and the actual churn interval at each funnel stage. The evaluation of the match can be performed by comparing whether the AI ​​reverse discussion predicted the corresponding stage as a churn risk interval for each funnel stage and whether the actual churn rate exceeds a preset churn rate threshold. A stage where both the prediction and the actual are classified as churn risk can be classified as True Positive, a stage where the prediction was churn risk but the actual churn rate is low can be classified as False Positive, and a stage where the prediction was safe but the actual churn rate is high can be classified as False Negative. The system (400) can evaluate the prediction accuracy of the AI ​​reverse discussion model by aggregating the distribution of True Positives, False Positives, and False Negatives for all funnel stages.

[0157] If there is a stage in the evaluation results where the predicted dropout interval and the actual dropout interval do not match, the system (400) can collect consumer behavior data at that stage as additional training data to update the consumption propensity parameters and dropout judgment thresholds of the virtual consumer persona. Consumer behavior data refers to a record of behaviors shown by the consumer during the actual advertising execution process and may include whether the ad was clicked, time spent on the landing page, whether an item was added to the cart, whether payment was completed, etc.

[0158] The system (400) can collect consumer behavior data from funnel stages where mispredictions and nonpredictions occurred as additional training data. The consumption propensity parameter may refer to the weights for each factor applied when determining whether a virtual consumer persona will drop out at each funnel stage. The dropout judgment threshold may refer to a standard value for determining whether a virtual consumer persona drops out at a specific funnel stage. The system (400) can analyze the collected consumer behavior data to identify the cause of the discrepancy between the actual consumer's dropout pattern and the virtual consumer persona's dropout judgment. Based on the identified cause, the system (400) can adjust the weights for each factor of the consumption propensity parameter or adjust the dropout judgment threshold upward or downward. For example, if it is found that the actual consumer is more influenced by brand trust than by price information at the stage where a misprediction occurred, the system (400) can update the consumption propensity parameter by lowering the price sensitivity weight and increasing the brand trust weight of the persona.

[0159] The system (400) can repeatedly execute a feedback loop using an AI reverse discussion model to which updated consumption propensity parameters and churn judgment thresholds are applied, so that the accuracy of churn prediction is improved compared to before the update when performing reverse discussions on subsequent marketing strategies. The feedback loop may refer to a structure in which the execution of marketing content, collection of performance data, comparison of the prediction results of the AI ​​reverse discussion model with the actual results, updating of model parameters, and subsequent prediction by the updated model are repeated cyclically.

[0160] Since new performance data is added as training data whenever the feedback loop is executed repeatedly, the AI ​​reverse discussion model can progressively reflect the actual consumer behavior patterns of the industry and target customer base more accurately. The execution cycle of the feedback loop can be set to a marketing campaign unit, a monthly unit, or a point in time when a certain amount of performance data is accumulated, and may vary depending on the platform's operational policy. Through the feedback loop, the system (400) can not stop at static analysis based merely on past data, but continuously improve the AI ​​reverse discussion model based on actual market reactions, thereby progressively improving the accuracy of pre-checking marketing strategies.

[0162] FIG. 8 is a flowchart illustrating the detailed execution process of an AI reverse discussion according to one embodiment.

[0163] The operations described through FIG. 8 can be implemented based on instructions that can be stored in memory (e.g., memory (430) of FIG. 4). The order of each operation of FIG. 8 may be changed, some operations may be omitted, and some operations may be performed simultaneously.

[0164] In operation 810, the execution of AI reverse discussion may be initiated. The system (400) may initiate operation 810 when a user executes the AI ​​reverse discussion function through the platform's input / output interface (414).

[0165] In operation 820, the system (400) can receive marketing strategy information from a user. The marketing strategy information may include features of the product or service to be marketed, an expected target customer base, a key appeal message, and major advertising channels. The system (400) can receive marketing strategy information entered by the user in text form through an input / output interface (414). The system (400) can convert the received marketing strategy information into structured data by parsing the product features, target customer base, appeal message, and advertising channel items, respectively, through natural language processing.

[0166] In operation 830, the system (400) can generate a virtual consumer persona that reflects the characteristics of the target customer. The system (400) can generate a virtual consumer persona based on the age group, gender, interests, and consumption pattern information of the expected target customer group included in the marketing strategy information received in operation 820. The system (400) can generate a virtual consumer persona by providing the demographic characteristics and consumption propensity of the target customer group as prompts to a large-scale language model. The system (400) can set the purchase decision pattern and churn tendency parameters of the virtual consumer persona by utilizing consumer response data of existing customers within the same industry accumulated in the database (412) as training data. The system (400) can generate multiple consumer type personas, such as price-sensitive, quality-oriented, or trend-following types, depending on the characteristics of the target customer group.

[0167] In operation 840, the system (400) may start a reverse search of the funnel from the final conversion stage. The system (400) may set the final conversion stage of the marketing funnel (e.g., the purchase completion stage) as the starting point for the search and initiate a process of searching each stage in reverse order from that stage toward the awareness stage. The reverse search method can have the effect of identifying logical breaks in strategy or information gaps that are difficult to discover in a forward funnel design by deriving in reverse the conditions that must be satisfied at each stage for the final conversion to succeed.

[0168] In operation 850, the system (400) can evaluate the purchase intent at the current funnel stage and check the risk of churn. The system (400) can evaluate whether the information or motivation required for the virtual consumer persona created in operation 830 to maintain the purchase intent at the current funnel stage is satisfied. The system (400) can assign the role of the virtual consumer persona to a large language model and configure prompts to infer in a question-and-answer format whether the content and appeal messages provided at the current funnel stage are sufficient to maintain the purchase intent of the persona.

[0169] In operation 860, the system (400) can determine whether the virtual consumer persona's purchase intent is maintained at the current funnel stage. Based on the evaluation result of operation 850, if the system (400) determines that the virtual consumer persona's purchase intent is maintained, it can proceed to operation 890. If the system (400) determines that the virtual consumer persona's purchase intent is decreasing or dropping out, it can identify the stage as a dropout risk zone and proceed to operation 870.

[0170] In operation 870, the system (400) can analyze churn factors for the current funnel stage identified as a churn risk section. The system (400) can calculate the correlation between the core appeal message and the content provided at the current stage, and classify cases where the calculated correlation is less than a preset threshold as churn factors. The calculation of the correlation can be performed by using a text embedding model to calculate the cosine similarity between the semantic vector of the core appeal message and the semantic vector of the content at that stage. The system (400) may also classify cases where the differentiated value compared to competing products is not specified at the current stage as a separate churn factor. The types of churn factors may include, but are not limited to, a lack of relevance to the core message and a lack of differentiation compared to competing products, and may vary depending on the industry and the characteristics of the target customer.

[0171] In operation 880, the system (400) can generate a strategy improvement plan corresponding to the churn factor analyzed in operation 870. The system (400) can generate different improvement plans depending on the type of churn factor. If the correlation between the appeal message and the content is low, the system (400) can generate a plan to add the core keywords of the appeal message to the content of that stage or to reconstruct the appeal message itself. If there is a lack of differentiated value compared to competing products, the system (400) can generate a plan to add product comparison content or to change the appeal message to one that emphasizes differentiation points. If multiple churn factors overlap, the system (400) can generate a plan to switch the advertising channel itself to a channel with a higher contact frequency with the target customer. The system (400) can save the generated improvement plan by mapping it to the corresponding funnel stage and churn factor. After operation 880 is completed, the system (400) can proceed to operation 890.

[0172] In operation 890, the system (400) can determine whether exploration for all funnel stages has been completed. The system (400) can check whether the processes of operations 850 through 880 have been performed for all funnel stages from the final transition stage to the recognition stage. If exploration for all funnel stages has been completed, the system (400) can proceed to operation 910. If there are funnel stages for which exploration has not been completed, the system (400) can proceed to operation 900.

[0173] In operation 900, the system (400) can move in reverse order from the current funnel stage to the previous funnel stage. The system (400) can move the search position from the funnel stage currently being explored to a funnel stage one step earlier (in the direction of the awareness stage). For example, if currently exploring the conversion stage, it can move to the interest stage, and if exploring the interest stage, it can move to the awareness stage. After operation 900 is completed, the system (400) can return to operation 850 to perform a purchase intent evaluation and churn risk check for the moved funnel stage. The process of operations 850, 860, 870, 880, 890, and 900 can be performed repeatedly until exploration for all funnel stages is completed.

[0174] In operation 910, the system (400) can generate and provide a strategy review report to the user. The system (400) can generate a strategy review report by synthesizing the dropout risk zones, dropout factors, and improvement plans for each funnel stage accumulated during the repeated execution of operations 850 to 900. The strategy review report may be structured to display the dropout risk of each funnel stage as a numerical value or grade, and to organize the types of dropout factors identified at each stage and the corresponding improvement plans step by step. The system (400) may be structured to visually highlight stages with a high dropout risk so that the user can intuitively identify the sections that need to be improved first. The system (400) can visually provide the generated strategy review report to the user through the dashboard of the input / output interface (414). Once the provision of the strategy review report is complete, the execution of the AI ​​reverse discussion may be terminated. Based on the provided strategy review report, the user can identify and supplement weaknesses in the strategy in advance before actually executing the marketing strategy.

[0175] According to one embodiment, the system (400) can perform an early warning function for marketing risks by analyzing the time-series change pattern of performance data occurring during the operation of a subscription customer's marketing funnel, predicting in advance before a decline in sales or customer churn occurs, and automatically generating a preemptive response strategy.

[0176] Specifically, the system (400) can collect performance data, including funnel stage conversion rates, ad channel click conversion rates, customer acquisition costs, and sales return rates relative to advertising costs, for each subscriber on a daily or weekly basis and store it as time-series data. Time-series data may refer to data in which the values ​​of the same indicator are recorded continuously in chronological order.

[0177] The system (400) can calculate the direction and speed of change of each performance indicator by performing a moving average and trend analysis on the stored time series data. A moving average may refer to a statistical method of averaging data values ​​over a certain period to mitigate short-term fluctuations and identify overall trends. The system (400) can simultaneously calculate a short-term moving average (e.g., 7 days) and a long-term moving average (e.g., 30 days) and detect the point at which the short-term moving average crosses the long-term moving average downward as a signal of performance decline. The setting of the moving average period is an example and may vary depending on the industry and the characteristics of the advertising campaign.

[0178] The system (400) can evaluate the severity of a signal when a signal of performance decline is detected. The evaluation of severity can be performed by combining the magnitude of the decline in performance indicators, the duration of the decline, and the degree of deviation from the average value of the industry-specific benchmark dataset. The system (400) can classify the severity into multiple preset grades, for example, into caution, warning, and danger grades. The types of grades and classification criteria may vary depending on the settings of the platform.

[0179] The system (400) can send risk alert notifications to subscribers based on a severity level. The risk alert notification may include the funnel stage where a decline in performance is detected, the trend of changes in metrics at that stage, and the severity level. The system (400) can automatically generate and provide a preemptive response strategy along with the risk alert. The generation of the preemptive response strategy may be performed by analyzing the current content type and ad channel configuration of the funnel stage where a decline in performance is detected, and by comparing it with the combination of content types and channels applied by high-performing companies at the same stage in an industry-specific benchmark dataset to identify elements that require change. Based on the identified change elements, the system (400) can generate a response strategy that includes replacing content types, adding or switching ad channels, readjusting budget allocation, or modifying appeal messages.

[0180] The system (400) can store the history of risk alerts and preemptive response strategies in a database and evaluate the effectiveness of the response strategies by tracking changes in performance after the response strategies have been executed. The system (400) can accumulate patterns of highly effective response strategies as training data and use them to generate more accurate response strategies when similar situations of performance decline occur in the future.

[0181] According to one embodiment, the system (400) can perform a cross-learning function that creates a collective learning effect within an industry by mutually sharing marketing performance data and content operation strategies among multiple subscriber customers belonging to the same industry category, within a range where the identification information of individual companies is not exposed.

[0182] Specifically, the system (400) can identify a subscriber as a best-performing case who has achieved a conversion rate at a specific funnel stage within the same industry category that is higher than a preset ratio compared to the benchmark group average. The system (400) can extract the content type, advertising channel, key keywords of the appeal message, and budget allocation ratio applied to the corresponding funnel stage from the marketing strategy of the identified best-performing case. The system (400) can remove information that can identify a specific company, such as the company name, brand name, and product name, from the extracted strategy elements and preserve only the structural pattern of the strategy in an anonymized form.

[0183] The system (400) can convert the strategy pattern of anonymized best performance cases into a form applicable to other subscriber customers within the same industry category. Conversion may refer to a process of replacing elements that depend on the brand characteristics of individual companies in the strategy pattern of best performance cases with the brand guidelines and business profiles of the subscriber customers, while maintaining industry-common elements. The system (400) can provide the converted strategy pattern as a recommended strategy to subscriber customers who need performance improvement at the corresponding funnel stage.

[0184] The system (400) can cluster subscribers with similar sales volume ranges, similar target customer bases, and similar sales channels within an industry category into learning groups to maximize the effect of cross-learning. Clustering may mean classifying objects with similar characteristics into the same group. The system (400) can convert attributes such as sales volume, target customer age group, and major sales channels included in each subscriber's business profile into feature vectors and form learning groups by applying K-means clustering or hierarchical clustering algorithms. The types of clustering algorithms are examples, and different algorithms may be applied depending on the characteristics of the data.

[0185] The system (400) can track the results of applying the recommended strategy after the strategy pattern of a best-performing case within a learning group has been recommended and executed to other subscribers. The system (400) can analyze the performance changes of subscribers to whom the recommended strategy was applied to calculate the increase or decrease in conversion rate, the change in customer acquisition cost, and the change in revenue profit margin compared to before the application of the strategy. If the results of applying the recommended strategy are positive, the system (400) can increase the recommendation priority of the strategy pattern, and if negative, it can decrease the recommendation priority or exclude it from the recommendation target. As the process of successful strategy patterns spreading within the learning group and ineffective strategy patterns being weeded out is repeated, the system (400) can have the effect of collectively improving the marketing performance of all subscribers within the industry.

[0187] Although the examples above have been illustrated with specific drawings, experts in the field may apply various technical modifications and variations based on them. For instance, appropriate results can be obtained even if the described techniques are performed differently from the presented order, or if the components of the described systems, structures, devices, or circuits are combined in different forms or replaced with other components or equivalents.

Claims

Claim 1 In a customized brand consulting and advertising artificial intelligence (AI) platform system for small and medium-sized enterprises, a memory for storing instructions; The system includes a processor, and the instructions, when executed by the processor, include: receiving a business profile from a small and medium-sized enterprise or micro-enterprise that includes an industry category, monthly sales volume, major sales channels, and the age group and interests of target customers; searching for a funnel template corresponding to the industry category and sales volume range of the received business profile in a funnel template database classified and accumulated by industry from the marketing execution history of existing subscribers; generating a customized marketing funnel structure including an awareness stage, an interest stage, a conversion stage, and a repurchase stage by reflecting the target customer characteristics of the business profile in the searched funnel template; performing an AI reverse discussion that analyzes the funnel in reverse from a consumer perspective regarding a business idea or marketing strategy entered by a user to derive dropout factors in the purchase journey; automatically generating marketing content in a direction that complements the identified dropout factors for the funnel stage where the dropout factors were identified based on the derived dropout factors; collecting and analyzing performance data resulting from the execution of the marketing content to calculate industry-specific marketing performance indicators; and, based on the calculated performance indicators, the marketing funnel Optimize the structure and content, and for each stage of the marketing funnel structure generated above, select a content type suitable for that stage based on the industry category and the age group of the target customers, matching social media video content to the awareness stage, detailed information landing pages to the interest stage, discount or promotion ads to the conversion stage, and retargeting ads to the repurchase stage, respectively; and calculate a total advertising budget proportional to the monthly sales volume for each matched content type,A system that controls the establishment of an execution strategy to allocate the total advertising budget according to the contribution ratio of each funnel stage by referring to industry-specific funnel stage contribution data, which calculates the average value of the ratio of each funnel stage to the final conversion from the results of marketing execution of existing subscribers within the same industry category accumulated in the funnel template database. Claim 2 delete Claim 3 In claim 1, the instructions, when executed by the processor, wherein the system performs the artificial intelligence (AI) reverse discussion, receives marketing strategy information in text form from the user, including features of the product or service to be marketed, expected target customer base, major advertising channels, and key appeal messages; based on the received marketing strategy information, the AI ​​model generates a virtual consumer persona reflecting the demographic characteristics and consumption propensity of the expected target customer base; while the generated virtual consumer persona explores each stage in reverse order from the final conversion stage to the awareness stage of the marketing funnel, it evaluates whether the information or motivation necessary to maintain purchase intent is satisfied at each stage; and during the evaluation process, identifies a stage where the purchase intent of the virtual consumer persona is judged to decrease as a churn risk zone, wherein cases where the correlation between the key appeal message and the content provided at that stage is below a preset threshold or where differentiated value compared to competing products is not specified are classified as churn factors; generate improvement measures including content reinforcement, change of appeal message, or switching of advertising channels for each classified churn factor; and the churn risk zone and A system that controls the generation and provision of a strategy review report to the user, mapping improvement plans to each funnel stage. Claim 4 In claim 1, when the instructions are executed by the processor, the system automatically generates the marketing content by receiving and storing brand guideline information including the user's brand logo, representative color, font, and tone and manner; providing industry information, the corresponding stage of the marketing funnel, and the brand guideline information as input conditions to an image generation model to generate an advertising image that conforms to the representative color and tone and manner of the brand; providing consumer behavior goals and key appeal messages for the corresponding stage of the marketing funnel as input conditions to a text generation model to generate advertising copywriting in stages, including short hooking phrases to enhance brand awareness in the awareness stage and phrases emphasizing benefits to induce purchasing behavior in the conversion stage; combining the generated advertising image and advertising copywriting according to the specifications of the advertising channel, and producing channel-specific content according to the resolution, text placement ratio, and file format requirements of each of the social media feed ads, story ads, banner ads, and landing page content; and for each of the produced channel-specific content, the degree of color matching with the stored brand guideline information, the use of font, and the tone and manner A system that verifies consistency, confirms content with a preset brand consistency score or higher as distributable, and controls the regeneration of content with a score lower than that. Claim 5 delete

Images