CRM-based customer data analysis system and method

Abstract

The invention relates to the technical field of big data, and discloses a CRM-based customer data analysis system and method, and the system comprises a main data view construction module which is used for carrying out the data processing of multi-source customer data, and generating a customer main data view of a target customer; the customer value grouping module is used for constructing a customer grouping feature set of the target customers according to the customer main data view, and performing clustering analysis on the target customers to obtain a multi-level value customer group; the customer data analysis module is used for constructing a multi-layer customer analysis network of target customers, extracting basic information and enterprise characteristics of the target customers, analyzing purchase cycle fluctuation coefficients and service response preferences of the target customers, and calculating potential demand probabilities and loss risk probabilities of the target customers; the customer portrait construction module is used for generating a multi-dimensional customer portrait of the target customer; and the customer analysis report module is used for generating a customer analysis report of the target customer. According to the invention, the comprehensiveness and accuracy of customer data analysis can be improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of big data, and particularly relates to a customer data analysis system and method based on CRM. BACKGROUND

[0002] Customer data analysis refers to using data processing technology, statistical analysis and machine learning model to transform, mine and interpret the original data collected from customers, aiming to extract valuable insights, thereby deeply understanding customer behavior, preference, demand, value and future trends, and finally providing scientific basis for strategic decision-making, precise marketing, customer relationship optimization and product service innovation of enterprises. Customer data analysis can provide objective and quantitative evidence, so that strategic planning, market investment, product development and other decisions are more accurate and predictable, greatly reducing the risk of decision-making.

[0003] The traditional customer data analysis method is a customer analysis method with descriptive statistics and static reports as the core. By qualitatively analyzing the structured transaction data in the enterprise, fixed weekly reports, monthly reports and the like are made, and rules are defined according to experience to group customers and generate analysis reports. This method has the problems of single data source, shallow analysis, and rigid group division, resulting in incomplete and low-accuracy customer analysis results. SUMMARY

[0004] The present application provides a customer data analysis system and method based on CRM, which aims to improve the comprehensiveness and accuracy of customer data analysis.

[0005] To achieve the above-mentioned purpose, the present application provides a customer data analysis system based on CRM, comprising: A main data view construction module is configured to extract multi-source customer data of a target customer from a CRM system, wherein the multi-source customer data comprises structured transaction data and unstructured interaction data, and to perform data processing on the multi-source customer data to obtain processed data, thereby generating a customer main data view of the target customer. A customer value group module is configured to construct a customer group feature set of the target customer based on the customer main data view, wherein the customer group feature set comprises RFM value indicators and behavior sequence features, and to perform clustering analysis on the target customer based on the customer group feature set, thereby obtaining a multi-level value customer group. a customer data analysis module configured to construct a multi-layer customer analysis network of the target customer according to the multi-level value customer group and the customer master data view, extract basic information and enterprise characteristics of the target customer through a static attribute layer in the multi-layer customer analysis network, analyze a purchase cycle fluctuation coefficient and a service response preference of the target customer through a dynamic behavior layer in the multi-layer customer analysis network, and calculate a potential demand probability and a churn risk probability of the target customer through a tendency analysis layer in the multi-layer customer analysis network; a customer portrait construction module configured to generate a multi-dimensional customer portrait of the target customer according to the basic information, the enterprise characteristics, the purchase cycle fluctuation coefficient, the service response preference, the potential demand probability, and the churn risk probability; a customer analysis report module configured to generate a customer analysis report of the target customer according to the multi-dimensional customer portrait.

[0006] Optionally, the calculating of the potential demand probability and the churn risk probability of the target customer through the tendency analysis layer in the multi-layer customer analysis network comprises: extracting a customer feature vector of the target customer through the tendency analysis layer in the multi-layer customer analysis network; identifying a potential demand type of the target customer according to the service response preference of the target customer; quantifying a potential demand value of the target customer for the potential demand type based on the customer feature vector; calculating the potential demand probability of the target customer according to the potential demand value; extracting an activity feature of the target customer; analyzing a behavior change rate and an emotional tendency factor of the target customer according to the activity feature and a behavior sequence feature corresponding to the target customer; calculating the churn risk probability of the target customer according to the behavior change rate and the emotional tendency factor.

[0007] Optionally, the generating of the multi-dimensional customer portrait of the target customer according to the basic information, the enterprise characteristics, the purchase cycle fluctuation coefficient, the service response preference, the potential demand probability, and the churn risk probability comprises: determining a portrait dimension and a portrait label of the target customer according to the basic information, the enterprise characteristics, the purchase cycle fluctuation coefficient, the service response preference, the potential demand probability, and the churn risk probability; fitting a customer feature radar chart of the target customer according to the portrait dimension; mapping the portrait label to the customer feature radar chart to obtain a customer label radar chart; Based on the customer label radar chart, a customer distribution state, a label proportion, and a customer change state of the target customer are analyzed; According to the customer distribution state, the label proportion, and the customer change state, a customer portrait dashboard of the target customer is generated; Based on the customer label radar chart and the customer portrait dashboard, a multi-dimensional customer portrait of the target customer is constructed Optionally, the data processing on the multi-source customer data to obtain processed data comprises: Identifying duplicate values, error values, and abnormal values of the multi-source customer data, Based on the duplicate values, the error values, and the abnormal values, data cleaning is performed on the multi-source customer data to obtain cleaned multi-source customer data; The cleaned multi-source customer data is standardized to obtain standardized multi-source customer data; Extracting multi-source data features of the standardized multi-source customer data; According to the multi-source data features, data fusion is performed on the standardized multi-source customer data to obtain fused customer data; Identifying sensitive factors of the fused customer data to perform desensitization processing on the fused customer data to obtain processed data.

[0008] Optionally, the generation of the customer master data view of the target customer comprises: Extracting key features of the corresponding processed data of the target customer, wherein the key features include identity features, transaction features, and behavior features; Based on the key features, a similarity score between key fields in the processed data and the target customer is calculated; According to the similarity score, a customer primary key of the processed data is generated; Based on the customer primary key, the processed data is clustered to obtain a customer master data wide table; According to the customer master data wide table, a customer master data view of the target customer is constructed.

[0009] Optionally, the construction of the customer clustering feature set of the target customer according to the customer master data view comprises: Extracting an interest graph set in the customer master data view, wherein the interest graph set includes a transaction table, a behavior log table, a member information table, and an order table; Based on the interest graph set, a last purchase time interval, a consumption frequency, and a consumption amount of the target customer are calculated; According to the last purchase time interval, consumption frequency and consumption amount, the RFM value index of the target customer is integrated; Based on the interest graph set, the behavior event sequence of the target customer is generated; Extract the behavior sequence features of the behavior event sequence; According to the behavior sequence features and the RFM value index, the customer segmentation feature set of the target customer is integrated.

[0010] Optionally, based on the customer segmentation feature set, the target customer is subjected to cluster analysis to obtain a multi-level value customer group, including: According to the RFM value index corresponding to the customer segmentation feature set, the customer value score of the target customer is calculated; Based on the behavior sequence features corresponding to the customer segmentation feature set, the anomaly coefficient of the target customer is calculated; According to the anomaly coefficient, the abnormal behavior state of the target customer is identified; According to the customer value score and the abnormal behavior state, the target customer is subjected to multiple classification to obtain a multi-level value customer group.

[0011] Optionally, according to the multi-level value customer group and the customer master data view, a multi-layer customer analysis network of the target customer is constructed, including: According to the multi-level value customer group and the customer master data view, a multi-layer network architecture of the target customer is constructed, wherein the multi-layer network architecture includes a static attribute layer, a dynamic behavior layer and a tendency analysis layer; The customer relationship network of the target customer is analyzed to construct a network environment of the multi-layer network architecture; The communication protocol and the running logic of the multi-layer network architecture are determined; Based on the communication protocol, the running logic and the network environment, the multi-layer network architecture is integrated to obtain a multi-layer customer analysis network.

[0012] Optionally, the purchase cycle fluctuation coefficient and service response preference of the target customer are analyzed through the dynamic behavior layer of the multi-layer customer analysis network, including: Through the dynamic behavior layer of the multi-layer customer analysis network, a purchase interval sequence of the target customer is constructed; The time decay factor of the purchase interval sequence is analyzed; According to the time decay factor and the purchase interval sequence, the purchase cycle fluctuation coefficient of the target customer is calculated; The customer response mode, response time and response frequency of the target customer are identified; calculate a response rate of the target customer based on the customer response mode, the response time and the response times, to construct a response rate distribution map of the target customer; determine a service response preference of the target customer according to the response rate distribution map.

[0013] A CRM-based customer data analysis method, characterized in that the method comprises: extracting multi-source customer data of a target customer from a CRM system, wherein the multi-source customer data comprises structured transaction data and unstructured interaction data, performing data processing on the multi-source customer data to obtain processed data, and generating a customer master data view of the target customer; constructing a customer clustering feature set of the target customer according to the customer master data view, wherein the customer clustering feature set comprises RFM value indicators and behavior sequence features, performing clustering analysis on the target customer based on the customer clustering feature set, and obtaining a multi-level value customer group; constructing a multi-layer customer analysis network of the target customer according to the multi-level value customer group and the customer master data view, extracting basic information and enterprise features of the target customer through a static attribute layer in the multi-layer customer analysis network, analyzing a purchase cycle fluctuation coefficient and a service response preference of the target customer through a dynamic behavior layer in the multi-layer customer analysis network, and calculating a potential demand probability and a churn risk probability of the target customer through a tendency analysis layer in the multi-layer customer analysis network; generating a multi-dimensional customer portrait of the target customer according to the basic information, the enterprise features, the purchase cycle fluctuation coefficient, the service response preference, the potential demand probability and the churn risk probability; generating a customer analysis report of the target customer according to the multi-dimensional customer portrait.

[0014] This invention extracts structured transaction data and unstructured interaction data from CRM systems, processes the data to generate a customer master data view, and constructs a customer segmentation feature set based on this view. Combined with RFM value indicators and behavioral sequence characteristics, cluster analysis is performed to form multi-level value customer groups. Furthermore, by constructing a multi-layered customer analysis network, basic customer information, enterprise characteristics, purchase cycle fluctuation coefficients, service response preferences, potential demand probabilities, and churn risk probabilities are extracted from the static attribute layer, dynamic behavior layer, and tendency analysis layer, respectively. This ultimately forms a multi-dimensional customer profile and generates a customer analysis report. Its beneficial effects include: significantly improving the comprehensiveness and accuracy of customer insights; achieving accurate prediction of customer behavior, needs, and risks through multi-dimensional data fusion and in-depth analysis; supporting enterprises in formulating personalized marketing strategies and risk intervention measures, effectively reducing customer churn rates and increasing customer value; optimizing resource allocation and improving operational efficiency; and providing a scientific basis for enterprise decision-making through automated report generation, enhancing market competitiveness. Therefore, it improves the comprehensiveness and accuracy of customer data analysis. Attached Figure Description

[0015] Figure 1 A functional module diagram of a CRM-based customer data analysis system provided in an embodiment of the present invention; Figure 2 A flowchart illustrating a customer data analysis method based on CRM, provided as an embodiment of the present invention; The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0016] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0017] Furthermore, the timing of the steps in the following method embodiments is merely an example and not a strict limitation.

[0018] In fact, the server device deployed by the CRM-based customer data analysis system can be composed of one or more devices. The CRM-based customer data analysis system can be implemented as a business instance, a virtual machine, or a hardware device. For example, the CRM-based customer data analysis system can be implemented as a business instance deployed on one or more devices in a cloud node. In short, the CRM-based customer data analysis system can be understood as a software deployed on a cloud node, which provides CRM-based customer data analysis services for each user terminal. Alternatively, the CRM-based customer data analysis system can also be implemented as a virtual machine deployed on one or more devices in a cloud node. The virtual machine has application software for managing each user terminal. Alternatively, the CRM-based customer data analysis system can also be implemented as a server composed of a plurality of same or different types of hardware devices, and one or more hardware devices are provided to provide CRM-based customer data analysis services for each user terminal.

[0019] In terms of implementation, the CRM-based customer data analysis system and the user terminal are mutually adapted. That is, the CRM-based customer data analysis system is installed as an application on a cloud service platform, and the user terminal is a client that establishes a communication connection with the application; or the CRM-based customer data analysis system is implemented as a website, and the user terminal is implemented as a web page; or the CRM-based customer data analysis system is implemented as a cloud service platform, and the user terminal is implemented as an applet in an instant messaging application.

[0020] Referring to Figure 1 FIG. 1 is a functional module diagram of a CRM-based customer data analysis system according to an embodiment of the present application.

[0021] The CRM-based customer data analysis system 100 can be set in a cloud server, and in terms of implementation, it can be one or more service devices, or a batch of applications installed on a cloud (such as a CRM-based customer data analysis server, a server cluster, etc.), or it can also be developed as a website. According to the functions implemented, the CRM-based customer data analysis system 100 includes a main data view construction module 101, a customer value grouping module 102, a customer data analysis module 103, a customer portrait construction module 104, and a customer analysis report module 105.

[0022] In the embodiment of the present application, in the tracking of the CRM-based customer data analysis, each of the above modules can be independently implemented and called by other modules. The calling here can be understood as that a module can connect a plurality of modules of another type and provide corresponding services for the plurality of modules connected thereby. In the CRM-based customer data analysis system provided by the embodiment of the present application, the application range of the CRM-based customer data analysis architecture can be adjusted by increasing modules and directly calling without modifying program codes, cluster horizontal expansion is realized, and the purpose of quickly and flexibly expanding the CRM-based customer data analysis system is achieved. In actual application, the above modules can be arranged in the same device or different devices, or in a virtual device, such as a service instance in a cloud server.

[0023] The following will be described in combination with specific embodiments, respectively for each component of the CRM-based customer data analysis system and the specific work flow.

[0024] The main data view construction module 101 is configured to extract multi-source customer data of a target customer from a CRM system, wherein the multi-source customer data includes structured transaction data and unstructured interaction data, and perform data processing on the multi-source customer data to obtain processed data, so as to generate a customer main data view of the target customer.

[0025] In the embodiment of the present application, the multi-source customer data of the target customer is extracted from the CRM system, so that data of different sources can be converged and associated, data islands are eliminated, a consistent and complete data basis is provided for subsequent analysis, and decision deviation caused by data fragmentation is avoided. The CRM system refers to a software platform for managing interaction between an enterprise and a customer. The multi-source customer data refers to a collection of multiple types of data related to a target customer collected from different systems, channels and interaction modes. The structured transaction data refers to customer transaction related data with a clear structure, a standard format and easy to store in a relational database, such as order, contract, repayment and other transaction information. The unstructured interaction data refers to customer interaction information without a fixed format and difficult to store and process directly in a traditional database table, such as text, voice, image and video.

[0026] In the embodiment of the present application, the multi-source customer data is processed to obtain processed data, so that data accuracy can be improved and the interference of noise on the analysis result can be reduced, thereby ensuring the reliability of subsequent modeling and analysis. The processed data refers to a data set meeting specific quality standards and analysis requirements obtained after a series of data processing procedures of the multi-source customer data.

[0027] As an embodiment of the present application, the processing of the multi-source customer data to obtain processed data includes: identify duplicate values, error values and abnormal values of the multi-source customer data, perform data cleaning on the multi-source customer data based on the duplicate values, the error values and the abnormal values, to obtain cleaned multi-source customer data; standardize the cleaned multi-source customer data to obtain standardized multi-source customer data; extract multi-source data features of the standardized multi-source customer data; perform data fusion on the standardized multi-source customer data according to the multi-source data features to obtain fused customer data; identify sensitive factors of the fused customer data to perform desensitization processing on the fused customer data to obtain processed data.

[0028] The duplicate values refer to multiple data records of the same real customer in multi-source data. The error values refer to values that do not conform to basic data format, type or business rule definition, such as misspelled words, negative age, etc. The abnormal values refer to values that significantly deviate from other observed values in data distribution. The cleaned multi-source customer data refers to a data set obtained after processing duplicate values, error values and abnormal values in original multi-source customer data. The standardized multi-source customer data refers to a data set after unifying the format, unit and expression of data. The multi-source data features refer to key information points extracted from the standardized data, which can describe customer attributes, behaviors and states, such as behavior features, preference features, derived features, etc. The fused customer data refers to information about the same customer from different data sources, which is integrated into a comprehensive customer data set. The sensitive factors refer to key information fields that can directly or indirectly identify the identity of a specific natural person, and involve personal privacy, business secrets, financial security, etc., such as name, ID card, bank card number, etc.

[0029] Optionally, the multi-source data features of the standardized multi-source customer data can be extracted through deep learning techniques, such as autoencoders, graph neural networks, etc.

[0030] Optionally, the fused customer data can be fused through feature dynamic weighting fusion of attention mechanism, such as calculating attention weights of the multi-source data features to perform weighted fusion on the standardized multi-source customer data to obtain fused customer data.

[0031] The embodiment of the present application can effectively solve the problem of data island and customer identity fragmentation by generating the customer master data view of the target customer, eliminate the cost of maintaining data copies by each department, and reduce the communication cost and operation failure caused by data inconsistency. The customer master data view refers to a unified, comprehensive, accurate, consistent and authoritative digital archive created for each independent customer by integrating all customer-related data sources within the enterprise.

[0032] As an embodiment of the present application, the generation of the customer master data view of the target customer comprises: extracting key features of the target customer corresponding to the processed data, wherein the key features include identity features, transaction features and behavior features; calculating the similarity score of the key fields in the processed data and the target customer based on the key features; generating the customer master key of the processed data according to the similarity score; clustering the processed data based on the customer master key to obtain a customer master data wide table; constructing the customer master data view of the target customer according to the customer master data wide table.

[0033] The key features refer to the core data attributes that can most effectively and stably represent the identity, behavior and value of a customer extracted from the processed data. The identity features refer to relatively static and semi-static information used to directly identify the personal identity of the customer, such as name, ID card, etc. The transaction features refer to data reflecting the commercial value exchange between the customer and the enterprise, such as transaction amount, transaction time, etc. The behavior features refer to non-transactional data generated in the process of customer interaction with the enterprise, such as APP login frequency, page browsing path, product clicking / collection / purchasing behavior, etc. The similarity score refers to a measure of how much a plurality of customer records from different data sources belong to the same real customer. The customer master key refers to a globally unique and stable identifier assigned to the identified unique real customer (or customer entity) in the data integration process. The customer master data wide table refers to a two-dimensional table integrating all key features from different data sources and different dimensions as columns with the customer master key as rows.

[0034] Optionally, the similarity score of the key fields in the processed data and the target customer can be calculated by a fuzzy matching algorithm, such as Jaccard similarity, Levenshtein distance, cosine similarity, etc.

[0035] Optionally, the customer master data view can be constructed by a logical view algorithm, such as SQL view, materialized view, etc The customer value cluster module 102 is configured to construct a customer cluster feature set of the target customer according to the customer master data view, wherein the customer cluster feature set comprises an RFM value index and a behavior sequence feature, and the target customer is subjected to cluster analysis based on the customer cluster feature set to obtain a multi-level value customer group.

[0036] According to the embodiment of the present application, the construction of the customer cluster feature set of the target customer according to the customer master data view can ensure that all analyses on customer value within an enterprise are based on the same set of standardized indexes, avoid the result conflicts caused by the use of different calculation standards by different teams, and ensure the consistency of decision-making. The customer cluster feature set refers to a set of structured and numerical customer attributes selected, calculated and constructed for the specific purpose of customer clustering from the customer master data view. The RFM value index refers to a quantitative index for measuring the interaction between the customer and the enterprise through three core dimensions (the time of the last consumption, the consumption frequency and the consumption amount). The behavior sequence feature refers to a series of discrete behavior events generated by the customer in the interaction with the enterprise, which are organized in chronological order to form an ordered sequence.

[0037] As an embodiment of the present application, the construction of the customer cluster feature set of the target customer according to the customer master data view comprises: extracting an interest graph set in the customer master data view, wherein the interest graph set comprises a transaction table, a behavior log table, a member information table and an order table; calculating the time interval of the last purchase, the consumption frequency and the consumption amount of the target customer based on the interest graph set; integrating the RFM value index of the target customer according to the time interval of the last purchase, the consumption frequency and the consumption amount; generating a behavior event sequence of the target customer based on the interest graph set; extracting a behavior sequence feature of the behavior event sequence; integrating the customer cluster feature set of the target customer according to the behavior sequence feature and the RFM value index.

[0038] The interest chart set refers to a set of data tables highly related to customer interest and behavior, which is filtered from the customer master data view according to the needs of customer clustering. The transaction table records the detailed data table of each core value exchange between the customer and the company. The behavior log table refers to a detailed log of the customer's interaction with the company's products and services in a non-transactional scenario. The member information table refers to a table that records the static attribute information of the customer as a member. The order table refers to a summary information table that records a customer's complete shopping request and transaction order. The last purchase time interval refers to the time difference between the analysis day and the customer's last successful purchase time. The consumption frequency refers to the total number of purchases made by the customer within a specified time window (such as the past year, a quarter). The consumption amount refers to the total amount of consumption accumulated by the customer within a specified time window (such as the past year, a quarter). The behavior event sequence refers to an ordered list formed by arranging all discrete behavior events of a customer within a period of time in chronological order.

[0039] Optionally, the RFM value indicator of the target customer can be integrated through a supervised learning model, such as logistic regression, XGBoost, LightGBM, etc.

[0040] Optionally, the customer clustering feature set of the target customer can be integrated through a deep learning feature cross and fusion network, such as constructing a behavior sequence tower for the behavior sequence feature and an RFM tower for the RFM value indicator, and fusing the behavior sequence tower and the RFM tower through a fusion network to obtain the customer clustering feature set.

[0041] The embodiment of the present application can abstract and induce from individual to group by clustering analysis of the target customer based on the customer clustering feature set to obtain a multi-level value customer group, and convert massive individual data into manageable and understandable group concepts. The multi-level value customer group refers to a set of customer groups with different value levels and behavior characteristics obtained through clustering analysis.

[0042] As an embodiment of the present application, the clustering analysis of the target customer based on the customer clustering feature set to obtain a multi-level value customer group includes: According to the RFM value indicator corresponding to the customer clustering feature set, the customer value score of the target customer is calculated; Based on the behavior sequence feature corresponding to the customer clustering feature set, the abnormal coefficient of the target customer is calculated; According to the abnormal coefficient, the abnormal behavior state of the target customer is identified; According to the customer value score and the abnormal behavior state, the target customer is classified multiple times to obtain a multi-level value customer group.

[0043] The customer value score refers to the overall value of a single customer to the enterprise in a specific period. The abnormal coefficient refers to the degree of deviation of the behavior pattern of a single customer from the normal behavior pattern to which it belongs. The abnormal behavior state refers to a qualitative description of the degree of deviation of the current behavior pattern of the target customer from its historical normal behavior pattern.

[0044] Optionally, the customer value score of the target customer can be calculated by weighted linear combination.

[0045] Optionally, the abnormal coefficient of the target customer can be calculated by an anomaly detection algorithm such as IsolationForest, Autoencoder, LOF, etc.

[0046] The customer data analysis module 103 is configured to construct a multi-layer customer analysis network of the target customer according to the multi-level value customer group and the customer master data view, extract basic information and enterprise characteristics of the target customer through a static attribute layer in the multi-layer customer analysis network, analyze the purchase cycle fluctuation coefficient and service response preference of the target customer through a dynamic behavior layer in the multi-layer customer analysis network, and calculate the potential demand probability and churn risk probability of the target customer through a tendency analysis layer in the multi-layer customer analysis network.

[0047] The embodiment of the application can realize stereoscopic integration and visualization of customer information, support dynamic behavior analysis and real-time response, and thus improve the intelligent level of the enterprise customer management system, by constructing a multi-layer customer analysis network of the target customer according to the multi-level value customer group and the customer master data view. The multi-layer customer analysis network refers to a data analysis architecture for deep customer insight and intelligent decision-making based on the customer master data view and the multi-level value customer group.

[0048] As an embodiment of the application, the multi-layer customer analysis network of the target customer is constructed according to the multi-level value customer group and the customer master data view, comprising: According to the multi-level value customer group and the customer master data view, a multi-layer network architecture of the target customer is constructed, wherein the multi-layer network architecture comprises a static attribute layer, a dynamic behavior layer, and a tendency analysis layer. The customer relationship network of the target customer is analyzed to construct a network environment of the multi-layer network architecture. The communication protocol and running logic of the multi-layer network architecture are determined. Based on the communication protocol, the running logic and the network environment, the multi-layer network architecture is integrated to obtain a multi-layer customer analysis network.

[0049] The multi-layer network architecture refers to a customer-centered hierarchical organization, structured storage and analysis of customer data of different dimensions and different granularities. The static attribute layer refers to the basic layer in the multi-layer network architecture, which stores the basic information and inherent characteristics of the customer. The dynamic behavior layer records the behavior data of the customer over time, reflecting the interaction process between the customer and the enterprise. The tendency analysis layer refers to a predictive analysis layer constructed by a machine learning model based on static attributes and dynamic behavior data. The customer relationship network refers to a network graph structure constructed based on the association between customers (such as social relationship, enterprise organizational structure, transaction link, etc.). The network environment refers to the infrastructure and data ecosystem supporting the operation of the multi-layer customer analysis network, such as data storage, computing resources, etc. The communication protocol refers to the standardized interface specification defining the data exchange and calling rules between layers and modules in the multi-layer network architecture. The running logic refers to the process and rules of the collaborative work of the modules in the multi-layer customer analysis network.

[0050] Optionally, the customer relationship network can be analyzed by time series dynamic network modeling, such as time series model (such as LSTM, GRU), dynamic graph neural network (such as DySAT, TGAT), etc.

[0051] Optionally, the communication protocol and running logic of the multi-layer network architecture can be determined by digital twinning technology, such as constructing a virtual environment of the multi-layer network architecture by digital twinning technology to simulate the communication behavior and running behavior of the multi-layer network architecture, and determining the communication protocol and running logic of the multi-layer network architecture based on the communication behavior and the running behavior.

[0052] The embodiment of the application extracts the basic information and enterprise characteristics of the target customer through the static attribute layer in the multi-layer customer analysis network, which can convert the extracted static information into structured labels, laying a foundation for subsequent precision marketing, risk control and personalized services. The basic information refers to the descriptive data directly related to the target customer and relatively stable, such as name, enterprise name, unified social credit code, etc. The enterprise characteristics refer to a multi-dimensional structured and unstructured information set describing the internal attributes, business status, market positioning and potential risks of the enterprise customer, such as enterprise name, legal representative, main business, etc.

[0053] The embodiment of the present application can more accurately reflect the current value and future potential of the target customer by analyzing the purchase cycle fluctuation coefficient and service response preference of the target customer through the dynamic behavior layer of the multi-layer customer analysis network, and support differentiated operation strategy. The purchase cycle fluctuation coefficient is a statistical indicator for measuring the stability of the customer's purchase behavior time interval within a period of time. The service response preference refers to the reaction tendency and selection mode of the customer when facing various services provided by the enterprise.

[0054] As an embodiment of the present application, the purchase cycle fluctuation coefficient and service response preference of the target customer are analyzed through the dynamic behavior layer of the multi-layer customer analysis network, which includes: The purchase interval sequence of the target customer is constructed through the dynamic behavior layer of the multi-layer customer analysis network. The time decay factor of the purchase interval sequence is analyzed. The purchase cycle fluctuation coefficient of the target customer is calculated according to the time decay factor and the purchase interval sequence. The customer response mode, response time and response frequency of the target customer are identified. The response rate of the target customer is calculated based on the customer response mode, response time and response frequency to construct the response rate distribution graph of the target customer. The service response preference of the target customer is determined according to the response rate distribution graph.

[0055] The purchase interval sequence refers to the time interval sequence between the target customer's consecutive two purchase behaviors. The time decay factor is a mathematical expression for measuring the gradual weakening of the influence of customer behavior over time. The customer response mode refers to the type of behavioral feedback of the customer to the service and marketing touch, such as clicking, opening, purchasing, complaining, etc. The response time refers to the time interval from service touch to customer response behavior. The response frequency refers to the total number of times the customer responds to service touch within a certain time window. The response rate refers to the ratio of the number of responses to the total number of touches within a certain time window. The response rate distribution graph refers to a chart that visualizes the response rate as the core indicator according to different dimensions (such as time, channel, response mode), which is used to reveal the distribution rule and preference of customer response behavior.

[0056] Optionally, the time decay factor of the purchase interval sequence can be analyzed by deep learning time series modeling methods such as long short-term memory network, gated recurrent unit, etc.

[0057] Optionally, the response rate distribution map can be constructed through dynamic interactive visualization, such as Plotly, Tableau, etc.

[0058] The potential demand probability and the churn risk probability of the target customer are calculated through the tendency analysis layer in the multi-layer customer analysis network, so that the marketing resource allocation can be optimized, the conversion rate can be improved, the customer churn rate can be reduced, and the customer lifetime value can be improved. The potential demand probability refers to the possibility of the target customer to generate purchase, use and consultation behaviors for specific products, services or functions in the future. The churn risk probability refers to the possibility of the target customer to terminate the use of products, services and no longer transact with the brand in the future.

[0059] As an embodiment of the present application, the potential demand probability and the churn risk probability of the target customer are calculated through the tendency analysis layer in the multi-layer customer analysis network, comprising: The customer feature vector of the target customer is extracted through the tendency analysis layer in the multi-layer customer analysis network; The potential demand type of the target customer is identified according to the service response preference of the target customer; The potential demand value of the target customer for the potential demand type is quantified based on the customer feature vector; The potential demand probability of the target customer is calculated according to the potential demand value by using the following formula: ; Wherein, represents the customer feature vector The potential demand probability of the target customer for the th potential demand type, represents the th potential demand type, represents the exponential function with as the base, represents the customer feature vector, represents the customer feature vector The potential demand value of the target customer for the th potential demand type, represents the behavior weight of the target customer, represents the total number of potential demand types; The activity feature of the target customer is extracted; The behavior change rate and the emotional tendency factor of the target customer are analyzed according to the activity feature and the corresponding behavior sequence feature of the target customer; The churn risk probability of the target customer is calculated by using the following formula according to the behavior change rate and the emotional tendency factor: ; wherein, represents the churn risk probability of the target customer under the customer feature vector , the churn event of the target customer, represents the churn event of the target customer, represents the exponential function with base , the customer feature vector, , the customer feature vector, , the weight vector of the customer feature vector, , the transpose of the weight vector, , the bias term, , the behavior change rate, , the weight of the behavior change rate, , the sentiment tendency factor, , the weight of the sentiment tendency factor.

[0060] The customer feature vector refers to a numerical vector representation formed after multi-dimensional feature quantization of the target customer, such as age, gender, region, occupation, login frequency, page dwell time, etc. The potential demand type refers to the product and service category that the target customer may be interested in but has not yet explicitly expressed or purchased. The potential demand value refers to the quantified score of the target customer's interest or purchase tendency for a certain potential demand type. The activity feature refers to the interaction intensity of the target customer with the platform product within a certain time period. The behavior change rate refers to the fluctuation degree of the customer's recent behavior relative to the historical behavior. The sentiment tendency factor refers to the emotional tendency exhibited by the customer during interaction. In this application, the value range is [-1, 1].

[0061] Optionally, the potential demand value can be quantified by reinforcement learning techniques such as DQN, PPO, etc.

[0062] Optionally, the behavior change rate can be analyzed by sliding window time series change detection, and the sentiment tendency factor can be analyzed by language models such as BERT, RoBERTa, etc.

[0063] The customer portrait construction module 104 is configured to generate a multi-dimensional customer portrait of the target customer based on the basic information, the enterprise feature, the purchase cycle fluctuation coefficient, the service response preference, the potential demand probability, and the churn risk probability.

[0064] The embodiment of the present application can enable the enterprise to more comprehensively and deeply understand each customer by generating the multi-dimensional customer portrait of the target customer according to the basic information, the enterprise characteristics, the purchase cycle fluctuation coefficient, the service response preference, the potential demand probability and the churn risk probability, so as to formulate more accurate marketing strategies, provide personalized product recommendations and services, and thus improve marketing effect and customer satisfaction. The multi-dimensional customer portrait refers to a comprehensive, dynamic and structured customer description model constructed by integrating multi-dimensional information of the customer.

[0065] As an embodiment of the present application, the multi-dimensional customer portrait of the target customer is generated according to the basic information, the enterprise characteristics, the purchase cycle fluctuation coefficient, the service response preference, the potential demand probability and the churn risk probability, and includes: According to the basic information, the enterprise characteristics, the purchase cycle fluctuation coefficient, the service response preference, the potential demand probability and the churn risk probability, the portrait dimension and the portrait label of the target customer are determined; According to the portrait dimension, the customer characteristic radar chart of the target customer is fitted; The portrait label is mapped to the customer characteristic radar chart to obtain a customer label radar chart; Based on the customer label radar chart, the customer distribution state, the label proportion and the customer change state of the target customer are analyzed; According to the customer distribution state, the label proportion and the customer change state, the customer portrait dashboard of the target customer is generated; Based on the customer label radar chart and the customer portrait dashboard, the multi-dimensional customer portrait of the target customer is constructed.

[0066] The portrait dimension refers to different analysis perspectives for constructing the customer portrait, such as basic information dimension, behavior characteristic dimension, predictive dimension, etc. The portrait label refers to specific identification of customer attributes or behavior characteristics extracted based on the portrait dimension, such as high-value customer, high-frequency purchaser, etc. The customer characteristic radar chart refers to a graph for displaying the standardized score distribution of the target customer in multiple portrait dimensions. The customer label radar chart refers to an enhanced visualization chart formed by mapping the portrait label to the customer characteristic radar chart. The customer distribution state refers to the distribution of the target customer in multiple portrait dimensions or labels. The label proportion refers to the proportion of the portrait label in the target customer group. The customer change state refers to the change trend of the portrait label and the characteristic of the target customer in the time dimension. The customer portrait dashboard refers to a visualization interface integrating information such as customer distribution state, label proportion, behavior change state, etc., for intuitively displaying the icon of the overall customer situation.

[0067] Optionally, the customer distribution status, tag proportion, and customer change status can be analyzed using clustering analysis algorithms, such as K-means, DBSCAN, Gaussian mixture model, etc.

[0068] Optionally, the multi-dimensional customer profile of the target customer can be constructed through multi-task learning and joint modeling, such as Transformer, LSTM, etc.

[0069] The customer analysis report module 105 is used to generate a customer analysis report for the target customer based on the multi-dimensional customer profile.

[0070] This invention, through generating a customer analysis report for the target customer based on the multi-dimensional customer profile, achieves a closed-loop process from data to insight, and from insight to decision-making, thereby enhancing an enterprise's customer management, marketing strategy formulation, and risk control capabilities. The customer analysis report refers to a structured, visual, and interpretable document or data presentation format generated based on the multi-dimensional customer profile.

[0071] This invention extracts structured transaction data and unstructured interaction data from CRM systems, processes the data to generate a customer master data view, and constructs a customer segmentation feature set based on this view. Combined with RFM value indicators and behavioral sequence characteristics, cluster analysis is performed to form multi-level value customer groups. Furthermore, by constructing a multi-layered customer analysis network, basic customer information, enterprise characteristics, purchase cycle fluctuation coefficients, service response preferences, potential demand probability, and churn risk probability are extracted from the static attribute layer, dynamic behavior layer, and tendency analysis layer, respectively, ultimately forming a multi-dimensional customer profile and generating a customer analysis report. Its beneficial effects include: significantly improving the comprehensiveness and accuracy of customer insights; achieving accurate prediction of customer behavior, needs, and risks through multi-dimensional data fusion and in-depth analysis; supporting enterprises in formulating personalized marketing strategies and risk intervention measures, effectively reducing customer churn rate and increasing customer value; optimizing resource allocation and improving operational efficiency; and providing a scientific basis for enterprise decision-making through automated report generation, enhancing market competitiveness. Therefore, this invention can improve the comprehensiveness and accuracy of customer data analysis.

[0072] like Figure 2 The diagram shown is a flowchart illustrating a CRM-based customer data analysis method according to an embodiment of the present invention. In this embodiment, the CRM-based customer data analysis method includes: S1. Extract multi-source customer data of the target customer from the CRM system, wherein the multi-source customer data includes structured transaction data and unstructured interaction data, process the multi-source customer data to obtain processed data, and generate a customer master data view of the target customer. S2, constructing a customer cluster feature set of the target customer according to the customer master data view, wherein the customer cluster feature set comprises an RFM value index and a behavior sequence feature, performing cluster analysis on the target customer based on the customer cluster feature set to obtain a multi-level value customer group; S3, constructing a multi-layer customer analysis network of the target customer according to the multi-level value customer group and the customer master data view, extracting basic information and enterprise features of the target customer through a static attribute layer in the multi-layer customer analysis network, analyzing a purchase cycle fluctuation coefficient and a service response preference of the target customer through a dynamic behavior layer in the multi-layer customer analysis network, and calculating a potential demand probability and a churn risk probability of the target customer through a tendency analysis layer in the multi-layer customer analysis network; S4, generating a multi-dimensional customer portrait of the target customer according to the basic information, the enterprise features, the purchase cycle fluctuation coefficient, the service response preference, the potential demand probability and the churn risk probability; S5, generating a customer analysis report of the target customer according to the multi-dimensional customer portrait.

[0073] In several embodiments provided by the present application, it should be understood that the provided system and method can be implemented in other manners. For example, the system embodiments described above are merely schematic. For example, the division of the modules is merely a logical function division. In actual implementation, another division manner can be adopted.

[0074] In addition, each function module in each embodiment of the present application can be integrated in a processing unit, or each unit can exist physically, or two or more units can be integrated in one unit. The integrated unit can be realized in the form of hardware, or in the form of hardware plus software function module.

[0075] Finally, it should be noted that each of the above embodiments can be combined or independent, and deleting any one of them does not affect the technical implementation of other embodiments. The above embodiments are only used to illustrate the technical solutions of the present application and are not limiting. Although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present application.

Claims

1. A customer data analysis system based on CRM, characterized in that, The CRM-based customer data analysis system includes: The master data view construction module is used to extract multi-source customer data of target customers from the CRM system. The multi-source customer data includes structured transaction data and unstructured interaction data. The multi-source customer data is processed to obtain processed data to generate the customer master data view of the target customer. The customer value segmentation module is used to construct a customer segmentation feature set for the target customers based on the customer master data view. The customer segmentation feature set includes: RFM value indicators and behavioral sequence features. Based on the customer segmentation feature set, cluster analysis is performed on the target customers to obtain multi-level value customer groups. The customer data analysis module is used to construct a multi-level customer analysis network for the target customer based on the multi-level value customer group and the customer master data view. Through the static attribute layer of the multi-level customer analysis network, it extracts the basic information and enterprise characteristics of the target customer. Through the dynamic behavior layer of the multi-level customer analysis network, it analyzes the purchase cycle fluctuation coefficient and service response preference of the target customer. Through the tendency analysis layer of the multi-level customer analysis network, it calculates the potential demand probability and churn risk probability of the target customer. The customer profile building module is used to generate a multi-dimensional customer profile of the target customer based on the basic information, the enterprise characteristics, the purchase cycle fluctuation coefficient, the service response preference, the potential demand probability, and the churn risk probability. The customer analysis report module is used to generate a customer analysis report for the target customer based on the multi-dimensional customer profile.

2. The CRM-based customer data analysis system as described in claim 1, characterized in that, The step of calculating the potential demand probability and churn risk probability of the target customer through the propensity analysis layer in the multi-layer customer analysis network includes: The customer feature vector of the target customer is extracted through the tendency analysis layer in the multi-layer customer analysis network. Based on the service response preferences of the target customers, identify the potential demand types of the target customers; Based on the customer feature vector, quantify the potential demand value of the target customer for the potential demand type; Based on the potential demand value, calculate the probability of potential demand from the target customer; Extract the activity characteristics of the target customer; Based on the activity characteristics and the behavioral sequence characteristics corresponding to the target customers, analyze the behavioral change rate and sentiment tendency factor of the target customers; The probability of churn for the target customer is calculated based on the rate of change in behavior and the sentiment tendency factor.

3. The CRM-based customer data analysis system as described in claim 1, characterized in that, Based on the basic information, enterprise characteristics, purchase cycle fluctuation coefficient, service response preference, potential demand probability, and churn risk probability, a multi-dimensional customer profile of the target customer is generated, including: Based on the basic information, the enterprise characteristics, the purchase cycle fluctuation coefficient, the service response preference, the potential demand probability, and the churn risk probability, the profile dimensions and profile tags of the target customer are determined. Based on the profile dimensions, fit a customer feature radar chart of the target customer; The profile tags are mapped onto the customer feature radar map to obtain the customer tag radar map; Based on the customer tag radar chart, analyze the customer distribution status, tag percentage, and customer change status of the target customers; Based on the customer distribution status, the tag proportion, and the customer change status, a customer profile dashboard for the target customer is generated. Based on the customer tag radar chart and the customer profile dashboard, a multi-dimensional customer profile of the target customer is constructed.

4. The CRM-based customer data analysis system as described in claim 1, characterized in that, The process of processing the multi-source customer data to obtain processed data includes: Identify duplicate, erroneous, and outlier values ​​in the multi-source customer data. Based on the duplicate values, the error values, and the outliers, the multi-source customer data is cleaned to obtain cleaned multi-source customer data. The cleaned multi-source customer data is standardized to obtain standardized multi-source customer data. Extract the multi-source data features from the standardized multi-source customer data; Based on the characteristics of the multi-source data, the standardized multi-source customer data is fused to obtain fused customer data; Sensitive factors in the fused customer data are identified to desensitize the fused customer data, resulting in processed data.

5. The CRM-based customer data analysis system as described in claim 1, characterized in that, The process of generating the customer master data view for the target customer includes: Extract key features from the processed data corresponding to the target customer, wherein the key features include: identity features, transaction features, and behavioral features; Based on the key features, the similarity score between the key fields in the processed data and the target customer is calculated; Generate the customer primary key of the processed data based on the similarity score; Based on the customer primary key, the processed data is clustered to obtain a wide table of customer master data; Based on the customer master data wide table, construct the customer master data view of the target customer.

6. The CRM-based customer data analysis system as described in claim 1, characterized in that, The step of constructing the customer segmentation feature set of the target customer based on the customer master data view includes: Extract the interest chart set from the customer master data view, wherein the interest chart set includes: transaction table, behavior log table, member information table and order table; Based on the interest chart set, calculate the most recent purchase time interval, purchase frequency, and purchase amount of the target customer; Based on the most recent purchase time interval, purchase frequency, and purchase amount, integrate the RFM value index of the target customer; Based on the interest chart set, a sequence of behavioral events for the target customer is generated; Extract the behavioral sequence features from the behavioral event sequence; Based on the behavioral sequence characteristics and the RFM value index, the customer segmentation feature set of the target customer is integrated.

7. The CRM-based customer data analysis system as described in claim 1, characterized in that, Based on the customer segmentation feature set, cluster analysis is performed on the target customers to obtain multi-level value customer groups, including: Calculate the customer value score of the target customer based on the RFM value index corresponding to the customer segmentation feature set; Based on the behavioral sequence features corresponding to the customer segmentation feature set, the anomaly coefficient of the target customer is calculated; Based on the anomaly coefficient, the abnormal behavior status of the target customer is identified; Based on the customer value score and the abnormal behavior status, the target customer is classified into multiple levels to obtain a multi-level value customer group.

8. The CRM-based customer data analysis system as described in claim 1, characterized in that, The step of constructing a multi-level customer analysis network for the target customer based on the multi-level value customer group and the customer master data view includes: Based on the multi-level value customer group and the customer master data view, a multi-layer network architecture for the target customer is constructed, wherein the multi-layer network architecture includes: a static attribute layer, a dynamic behavior layer, and a tendency analysis layer; Analyze the customer relationship network of the target customers to construct the network environment of the multi-layered network architecture; Determine the communication protocol and operating logic of the multi-layer network architecture; Based on the communication protocol, the operating logic, and the network environment, the multi-layer network architecture is integrated to obtain a multi-layer client analysis network.

9. The CRM-based customer data analysis system as described in claim 1, characterized in that, The step of analyzing the purchase cycle fluctuation coefficient and service response preferences of the target customer through the dynamic behavior layer in the multi-layer customer analysis network includes: The purchase interval sequence of the target customer is constructed through the dynamic behavior layer in the multi-layer customer analysis network. Analyze the time decay factor of the purchase interval sequence; Calculate the purchase cycle volatility coefficient of the target customer based on the time decay factor and the purchase interval sequence; Identify the target customer's response method, response time, and number of responses; Based on the customer response method, the response time, and the number of responses, calculate the response rate of the target customer to construct a response rate distribution map of the target customer; Based on the response rate distribution map, the service response preferences of the target customer are determined.

10. A customer data analysis method based on CRM, characterized in that, The method includes: Extract multi-source customer data of target customers from the CRM system. The multi-source customer data includes structured transaction data and unstructured interaction data. Perform data processing on the multi-source customer data to obtain processed data, and generate a customer master data view of the target customer. Based on the customer master data view, a customer segmentation feature set for the target customers is constructed, wherein the customer segmentation feature set includes: RFM value indicators and behavioral sequence features. Based on the customer segmentation feature set, cluster analysis is performed on the target customers to obtain multi-level value customer groups. Based on the multi-level value customer group and the customer master data view, a multi-level customer analysis network for the target customer is constructed. Through the static attribute layer of the multi-level customer analysis network, the basic information and enterprise characteristics of the target customer are extracted. Through the dynamic behavior layer of the multi-level customer analysis network, the purchase cycle fluctuation coefficient and service response preference of the target customer are analyzed. Through the tendency analysis layer of the multi-level customer analysis network, the potential demand probability and churn risk probability of the target customer are calculated. Based on the basic information, the enterprise characteristics, the purchase cycle fluctuation coefficient, the service response preference, the potential demand probability, and the churn risk probability, a multi-dimensional customer profile of the target customer is generated. Based on the multi-dimensional customer profile, a customer analysis report for the target customer is generated.

Images