A marketing strategy optimization processing method and system based on user clustering

Abstract

The application provides a marketing strategy optimization processing method and system based on user clustering. The marketing strategy optimization processing method based on user clustering comprises the following steps: accessing the business data of users and user vehicles of a plurality of business systems into a customer service operation platform, and constructing an ID graph of the users; constructing a label of the users by using the related business data of the users and the user vehicles; performing clustering processing on the users according to the label of the users, and obtaining a target clustering package corresponding to the users; obtaining a multi-dimensional portrait of a target group corresponding to the target clustering package by deep analysis and insight based on the label corresponding to the users; formulating a differentiated operation strategy for groups with different value layers according to the target clustering package corresponding to the users, reaching the users through multiple channels and recovering effect data, and optimizing the strategy by taking the effect data as a guide. The system comprises modules corresponding to the method steps.

Description

TECHNICAL FIELD

[0001] The application provides a marketing strategy optimization processing method and system based on user clustering, and belongs to the technical field of strategy optimization processing. BACKGROUND

[0002] The external environment is more complex in the economic downturn cycle, and the competition in the automobile industry is intensifying, facing elimination and clearing; digitalization brings efficiency improvement of the real industry, and traditional industries are facing accelerated transformation. High-quality demand promotes the expansion and optimization of the service industry structure, and the integration of digitalization and the service industry brings new growth opportunities. The "14th Five-Year Plan" proposes to "deepen the digital application of research and development, design, production and manufacturing, operation management, and market services", and "promote the digital transformation of the industry". Digital transformation has become one of the core strategies of automobile industry enterprises, and more and more automobile enterprises expect to realize the transformation, innovation and growth of enterprise business through digital transformation. With the demand for digital transformation of automobile enterprises, the role of software system platforms in the operation of automobile industry enterprises has changed from a tool for improving enterprise operation efficiency to a driving force for driving enterprise innovation. The software in the field of automobile marketing and after-sales service will further penetrate the whole life cycle of car purchase, car use and car maintenance, and integrate with many software and application scenarios. Through software, automobile enterprises can better understand consumer demand, improve consumer experience, and improve consumer satisfaction. According to statistics, the software market in the automobile after-sales field will continue to expand in the future, with an annual growth rate of more than 25%, and automobile enterprises will increase investment in the digital transformation of the automobile after-sales field.

[0003] The customer service operation system has undergone many changes, from the past single and passive service to the present fine and proactive service mode, and in the future, enterprises will face more challenges in the field of customer service. In the past: single product or service, limited customer touchpoints and sales channels, more passive response, and lack of proactive service awareness. Now: three-dimensional and diversified service journey, richer consumer channels, shorter customer touch link, fine and meticulous work, and proactive occupation of user mind. In the future: deepen service depth, broaden service breadth, VR&AR technology application and popularization, and data analysis empowerment of customer service.

[0004] The digital economy and the real economy accelerate integration, the digitalization investment of the automobile after-sales market continues to increase, and the customer service operation system is facing profound changes. As the marketing environment is changing, the traditional marketing mode is facing the following challenges: (1) The number of online contacts increases. Influenced by the mobile Internet, users are migrating to the online at an accelerated pace. At the same time, as user attention is becoming more fragmented, it is more difficult to integrate resources and achieve effective dissemination. (2) The marketing cost increases. With the disappearance of the traffic dividend, the increase of market saturation, and the increasingly fierce market competition, the difficulty and cost of acquiring new customers at the traffic end are gradually increasing. (3) The conversion effect is poor. With the continuous influx of diversified content, it is difficult to continue to expand customers and implement one-size-fits-all user operation. Undifferentiated marketing can no longer easily impress target customers. SUMMARY

[0005] The application provides a marketing strategy optimization processing method and system based on user grouping, to solve the technical problems in the prior art. The technical solutions adopted are as follows:

[0006] A marketing strategy optimization processing method based on user grouping, the marketing strategy optimization processing method based on user grouping comprises:

[0007] Accessing the user and user vehicle related business data of multiple business systems to a customer service operation platform, and constructing an ID graph corresponding to the user;

[0008] Constructing a label corresponding to the user by using the user and user vehicle related business data;

[0009] Grouping the user according to the label corresponding to the user, obtaining a target grouping package corresponding to the user, and displaying the user and the data information having an upstream and downstream dependency relationship of the user according to the target grouping package;

[0010] Based on the label corresponding to the user, a multi-dimensional portrait of a target group corresponding to the target grouping package is obtained through deep analysis and insight;

[0011] According to the target grouping package corresponding to the user, a differentiated operation strategy is formulated for different value stratified groups, the user is reached through multiple channels, and effect data is recovered, and the effect data is used as a guide to optimize the strategy.

[0012] Further, accessing the user and user vehicle related business data of multiple business systems to a customer service operation platform, and constructing an ID graph corresponding to the user, comprises:

[0013] Accessing the system business original data corresponding to the user and the user vehicle in multiple business systems to the customer service operation platform; wherein the multiple business systems include but are not limited to a marketing system, an ERP system and a vehicle system;

[0014] The system performs data cleaning on the raw business data and then transmits the cleaned raw business data to the target source; wherein, the data cleaning process includes, but is not limited to, filtering, deduplication, and replacement.

[0015] Configure IDmapping logic, and set the priority in the multi-source data matching process according to the source and data integrity of the original business data of the system and the actual business rules. At the same time, establish a unique identifier for the user and its corresponding user vehicle, wherein the unique identifier is OneID.

[0016] Fragmented data from multiple accounts of the same user are linked based on user ID information, and the fragmented data is integrated to generate integrated data corresponding to the user ID information; wherein, the user ID information includes, but is not limited to, user ID, mobile phone number, and device number.

[0017] Furthermore, data quality monitoring is performed on the integrated data corresponding to user ID information, including:

[0018] Scan the integrated data corresponding to the user ID information to obtain the initial abnormal feature elements contained in the integrated data; wherein, the initial abnormal feature elements include the number of missing key fields and the weight value corresponding to the missing data, and the data garbled rate contained in the integrated data and the weight value corresponding to the garbled data.

[0019] A row matrix X is formed based on the number of missing key fields and their corresponding weight values, corresponding to the key fields. c ;

[0020] A column matrix Y is formed based on the number of garbled fields and their corresponding weight values, corresponding to the garbled fields. c ;

[0021] According to row matrix X c Sum column matrix Y c Using the number of elements in the matrix with more elements as a baseline, the matrix with fewer elements is padded with the number 1 to increase the number of dimensions, generating the padded row matrix X and column matrix Y; the structures of the padded row matrix X and column matrix Y are as follows:

[0022]

[0023] Where X represents the row matrix after padding operations corresponding to the key fields; x 01 x 02 , ..., x α These represent the weight values ​​corresponding to the elements contained in the row matrix after the padding operation;

[0024]

[0025] wherein Y represents the column matrix after the padding operation corresponding to the garbled field; y 01 , y 02 , …, y β respectively represent the weight values corresponding to the elements contained in the column matrix after the padding operation;

[0026] An initial anomaly factor is obtained by using the row matrix X and the column matrix Y after the padding, in combination with the key field missing rate and the garbled rate;

[0027] wherein the initial anomaly factor is obtained by the following formula:

[0028]

[0029] wherein K represents the initial anomaly factor; σ represents the number of elements contained in the row matrix X after the padding; μ represents the number of elements contained in the column matrix Y after the padding; x i represents the weight value corresponding to the i-th element contained in the row matrix X after the padding; y i represents the weight value corresponding to the i-th element contained in the column matrix Y after the padding; P 01 and P 02 respectively represent the data missing rate and the data garbled rate corresponding to the integrated data; represents the norm operator corresponding to the matrix;

[0030] The initial anomaly factor is compared with a preset factor threshold value;

[0031] When the initial anomaly factor exceeds the preset factor threshold value, data anomaly alarm is performed;

[0032] When the initial anomaly factor does not exceed the preset factor threshold value, the time difference between the integrated data update time and the fragmented data update time, the number of missing key fields after data update, the weight values of the missing data, and the data garbled rate and the weight values of the garbled data contained in the integrated data are monitored in real time;

[0033] An anomaly factor corresponding to each data update is obtained by using the initial anomaly factor in combination with the time difference between the integrated data update time and the fragmented data update time, the number of missing key fields after data update, the weight values of the missing data, and the data garbled rate and the weight values of the garbled data contained in the integrated data;

[0034] wherein the anomaly factor is obtained by the following formula:

[0035]

[0036] Wherein, Q represents an abnormality factor; w represents the number of updated fragmented data corresponding to the current data update; T ci represents the time difference between the time when the i-th updated fragmented data completes the update and the time when the data corresponding to the fragmented data in the integrated data completes the update; K represents an initial abnormality factor; K g represents the initial abnormality factor corresponding to the current integrated data after completing the update;

[0037] comparing the abnormality factor with a preset abnormality factor threshold;

[0038] when the abnormality factor exceeds the preset abnormality factor threshold, data abnormality alarm is performed.

[0039] Further, the label corresponding to the user is constructed by using the related service data of the user and the user vehicle, including:

[0040] the behavior data and the attribute data corresponding to the user are called, and the label corresponding to the user and the user vehicle is created according to a preset service logic combined with different model categories;

[0041] the label content is presented and managed according to the label structural mode for each label corresponding to the user and the user vehicle, wherein the label structure includes the number, classification and hierarchical relationship of the label, etc.

[0042] Further, the different model categories include an AIPL model, a 5A model and an RFM model; wherein the structures of the AIPL model, the 5A model and the RFM model are as follows:

[0043] In the AIPL model, A represents a brand awareness population; I represents a brand interest population; P represents a brand purchase population; and L represents a brand loyalty population;

[0044] In the 5A model, the parameters A1, A2, A3, A4 and A5 are included, wherein A1 represents a customer who passively accepts information; A2 represents a customer whose brand impression is increased; A3 represents a customer who actively searches information driven by curiosity; A4 represents a customer who takes action; and A5 represents a customer who has loyalty to the brand and propagates the brand;

[0045] The RFM model evaluates the R value, the F value and the M value of each user, and corresponds them to different intervals, so as to divide the users into eight user value types; wherein the R value represents the last consumption, reflecting the activity degree of a customer; the F value represents the consumption frequency, reflecting the loyalty of a customer; the M value represents the consumption amount, reflecting the contribution degree of a customer; and the eight user value types include important value customers, important exchange customers, important deep cultivation customers, important retention customers, potential customers, new customers, general maintenance customers and lost customers.

[0046] Further, the user is grouped according to the label corresponding to the user, a target group package corresponding to the user is obtained, and the user and data information having an upstream and downstream dependency relationship with the user are displayed according to the target group package, including:

[0047] A target group package corresponding to the user is created through a preset group rule or an uploaded list; the target group package is used for insight or pushing to a downstream marketing channel, so as to realize accurate reach of a target customer;

[0048] A group movement record is automatically generated according to the target group package; the group movement record is used for displaying running records, historical performances and trends of the group, so as to timely understand the latest situation of the group task;

[0049] The blood relationship details corresponding to the target group package are viewed according to the target group package, and data sources, people group packages and label resources having an upstream and downstream dependency relationship with the target group package are displayed with the target group package as the center.

[0050] Further, the target group package corresponding to the user is created in the following manner:

[0051] Rule creation: a group is created by using labels, behaviors, people group packages, detailed data and user attribute data through a visual component;

[0052] Upload creation: a target group package is quickly created by uploading a local seed people group file.

[0053] Further, based on the label corresponding to the user, a multidimensional portrait of a target group corresponding to the target group package is obtained through deep analysis and insight, including:

[0054] The label corresponding to the user is called, and the label corresponding to the user is displayed for the main body information of the user and the vehicle of the user, wherein the main body information includes basic information, covering labels and behavior timelines of the user, and a group to which the user belongs; key information of a user is quickly understood through an individual portrait, so as to further perform a marketing action on the user;

[0055] The target group package corresponding to the user and the vehicle of the user is called, an insight report of a target group corresponding to the main body of the user and the vehicle of the user is generated, and group characteristics are mined, and meanwhile, each dimension of the user is deeply understood through cross-listening and drilling analysis, wherein the group characteristics are used for guiding enterprise operation decision;

[0056] Extract the explicit and implicit features of the target cluster package corresponding to the user and the user vehicle, and mine the label combination matching the target cluster package according to the explicit and implicit features, and generate a cluster package meeting the preset demand according to the saliency, coverage and cluster number, etc., so as to further analyze and decide;

[0057] Create a life cycle label, and use the life cycle label to gain insight into the user life cycle, and export a target cluster package according to the user stage, carry out fine operation, promote user purchase decision, and continuously improve user loyalty to the brand.

[0058] Further, according to the target cluster package corresponding to the user, a differentiated operation strategy is formulated for different value stratified groups, the user is reached through multiple channels and effect data is recovered, and the effect data is used to guide the optimization of the strategy, including:

[0059] According to the cluster attribute and behavior prediction result of the target cluster package corresponding to the user, an operation strategy is formulated;

[0060] Connect to CRM and other systems to obtain a customer panoramic image, and execute a marketing strategy according to the customer panoramic image;

[0061] By constructing a churn prediction model and gaining insight into the churned population, an individualized operation strategy is executed to maximize the extension of the user life cycle.

[0062] Further, the user cluster-based marketing strategy optimization processing method further comprises:

[0063] Real-time monitoring of data transmission operation parameters between each business system and the customer service operation platform, wherein the data transmission operation parameters include the standard deviation of data call response time per unit time, the ratio between the process and available CPU per unit time, the API request frequency per unit time and the security event occurrence rate per unit time; and the unit time is 3-8 minutes;

[0064] Using the data transmission operation parameters to form a feature vector corresponding to each business system per unit time; wherein the feature vector structure is as follows:

[0065]

[0066] Wherein, A represents the feature vector corresponding to each unit time; T b represents the standard deviation of data call response time per unit time; P c represents the ratio between the process and available CPU per unit time; F represents the API request frequency per unit time; P arepresents the security event occurrence rate per unit time;

[0067] The standard deviation of the feature vector of each unit time corresponding to each business system is processed, the standardized processed feature vector is generated, and the standardized processed feature vector of each unit time corresponding to each business system is obtained by the following formula:

[0068]

[0069] wherein A b represents the standardized processed feature vector; δ(T b ), δ(P c ), δ(F) and δ(P a ) respectively represent the data call response time standard deviation per unit time, the ratio between the process and the available CPU per unit time, the API request frequency per unit time and the security event occurrence rate per unit time corresponding to the standardized parameters;

[0070] The feature coefficient is obtained by using the standardized processed feature vector corresponding to all the unit times experienced by each business system, wherein the feature coefficient is obtained by the following formula:

[0071]

[0072] wherein ξ represents the feature coefficient corresponding to each business system; n represents the number of all the unit times experienced by each business system; represents the Euclidean norm of the standardized processed feature vector corresponding to the i-th unit time, the cumulative effect of the overall load; represents the Manhattan norm of the standardized processed feature vector corresponding to the i-th unit time, reflecting the overall strength or importance of the feature vector; ε represents a preset minimum constant, used to prevent from being 0; δ(T b ) i , δ(P c ) i , δ(F) i and δ(P a ) i respectively represent the data call response time standard deviation corresponding to the i-th unit time, the ratio between the process and the available CPU, the API request frequency and the security event occurrence rate corresponding to the standardized parameters; α and β respectively represent the first adjustment coefficient and the second adjustment coefficient, and the value range of the first adjustment coefficient and the second adjustment coefficient is 0.53-0.72, 0.47-0.79;

[0073] Obtain a comprehensive load coefficient by using the characteristic coefficient corresponding to each business system;

[0074] The comprehensive load coefficient is obtained by the following formula:

[0075]

[0076] Wherein, S represents the comprehensive load coefficient; m represents the number of business systems; ξ i represents the characteristic coefficient corresponding to the i th business coefficient; ξ b represents the characteristic coefficient standard deviation corresponding to m business coefficients; ξ bi represents the characteristic coefficient standard deviation corresponding to n unit time corresponding to the i th business coefficient;

[0077] Compare the comprehensive load coefficient with the preset comprehensive coefficient threshold value;

[0078] When the comprehensive load coefficient exceeds the preset comprehensive coefficient threshold value, it is determined that the communication operation load of the customer service operation platform is overloaded, and load overload early warning is performed.

[0079] A marketing strategy optimization processing system based on user grouping, the marketing strategy optimization processing system based on user grouping comprises:

[0080] An ID map construction module is configured to access the related business data of users and user vehicles of multiple business systems to a customer service operation platform, and construct an ID map corresponding to the users;

[0081] A label construction module is configured to construct labels corresponding to the users by using the related business data of the users and user vehicles; a target grouping package acquisition module is configured to group the users according to the labels corresponding to the users, acquire a target grouping package corresponding to the users, and display the users and data information having an upstream and downstream dependency relationship with the users according to the target grouping package;

[0082] A multi-dimensional portrait acquisition module is configured to acquire a multi-dimensional portrait of a target group corresponding to the target grouping package by deep analysis and insight based on the labels corresponding to the users;

[0083] An operation strategy acquisition and optimization module is configured to formulate differentiated operation strategies for different value stratification groups according to the target grouping package corresponding to the users, reach the users through multiple channels and recover effect data, and optimize the strategies by using the effect data as a guide.

[0084] The present application has the following advantages:

[0085] The marketing strategy optimization processing method and system based on user grouping provided by the application help enterprises break data silos, establish unified people and vehicle archives, empower enterprises to drive full-link marketing and deep operation with data, and realize enterprise digital transformation and growth. Enterprises can build user tags and portraits through a customer service operation platform, stratify and group users, and through group insight capabilities, they can mine group characteristics and deeply understand target user groups. For users with different characteristics, they can develop targeted operation strategies. Through fine operation of customers, they can change from traffic thinking to user thinking, thereby mastering the initiative of marketing. At the same time, by establishing a customer service operation platform and integrating with marketing-related systems, all people and vehicle data can be imported into the lake. Through the connection of different system people and vehicle IDs, the uniqueness of individuals such as users and vehicles can be identified. Tags and groups are established to analyze user groups and individuals, enabling quick selection of specific groups. Through machine learning and model algorithms, the application provides modeling of segmented groups for analysis and decision-making. According to the attributes and consumption prediction of different groups, the application establishes precise matching marketing strategies to achieve precise marketing of target groups.

[0086] On the other hand, the marketing strategy optimization processing method and system based on user grouping also has the following beneficial technical effects:

[0087] (1) The marketing strategy optimization processing method and system based on user grouping provided by the application has a data fusion function. Different business systems and identity tags from different sources are identified as the same subject through data technology, thereby breaking down data silos and realizing data integration.

[0088] (2) The marketing strategy optimization processing method and system based on user grouping provided by the application has a tag setting and tag system function. Through the tag full life cycle + tag application effect, a tag optimization full-process closed loop is built. Through one-stop tag construction and visual interface interaction, tags are self-created and managed, and 360° tag system construction is completed efficiently by man and machine.

[0089] (3) The marketing strategy optimization processing method and system based on user grouping provided by the application has a user grouping function. Based on tags and data, target groups can be self-selected, and people groups can be accurately and quickly selected to meet diverse analysis and operation needs.

[0090] (4) The customer operation platform algorithm model function provided by the application can quickly obtain prediction results according to configured data content, thereby supporting decision-making, optimization, prediction, and other business goals, and achieving greater business value.

[0091] (5), the marketing strategy optimization processing method and system insight analysis function based on user group provided by the application can be used for the sales to view the detailed customer file for the key customers, and more targeted service can be provided according to the customer condition. According to the portrait difference between different user groups, the significant difference characteristics can be compared and analyzed, and the marketing strategy can be continuously optimized. BRIEF DESCRIPTION OF DRAWINGS

[0092] Figure 1 The flow chart of the marketing strategy optimization processing method is provided in the application.

[0093] Figure 2 The system principle diagram of the marketing strategy optimization processing system is provided in the application. EMBODIMENT

[0094] The preferred embodiments of the application are described below in combination with the drawings, and it should be understood that the preferred embodiments described herein are only used to illustrate and explain the application, and are not used to limit the application.

[0095] The embodiment of the application provides a marketing strategy optimization processing method based on user group, as shown in the figure, the marketing strategy optimization processing method based on user group comprises the following steps: Figure 1

[0096] S1, the user and the related business data of the user vehicle of a plurality of business systems are connected to a customer service operation platform, and an ID graph corresponding to the user is constructed;

[0097] S2, the user corresponding label is constructed by using the related business data of the user and the user vehicle;

[0098] S3, the user is grouped according to the user corresponding label, the target group package corresponding to the user is obtained, and the user and the data information with upstream and downstream dependent relationship are displayed according to the target group package;

[0099] S4, the multidimensional portrait of the target group corresponding to the target group package is obtained based on the user corresponding label through deep analysis and insight mode;

[0100] S5, according to the target group package corresponding to the user, different value stratification groups are formulated, the user is reached through multiple channels, and the effect data is recovered, and the effect data is used as the guidance to optimize the strategy.

[0101] ​The working principle of the technical solution is that the technical solution of the embodiment establishes unified person and vehicle archives, enables enterprises to drive full-link marketing and deep operation with data, and realizes digital transformation and growth of enterprises. Enterprises can build user tags and portraits through a customer service operation platform, stratify and group users, and through group insight capabilities, dig out group characteristics, deeply understand target user groups, develop targeted operation strategies for users with different characteristics, and through fine operation of customers, change from traffic thinking to user thinking, thereby mastering the initiative of marketing. At the same time, by establishing a customer service operation platform, the platform is integrated with marketing-related systems to pass through all people and vehicle data, and through the connection of different system people and vehicle IDs, the uniqueness of individuals such as users and vehicles is realized. Tags and groups are established to analyze the portraits of user groups and individuals, and the rapid selection of specific groups is realized. Through machine learning and model algorithms, the modeling of segmented groups is provided for analysis and decision-making. According to the attributes and consumption prediction of different groups, precise matching marketing strategies are established to realize precise marketing of target groups.

[0102] First, data integration and fusion, the data of the business system related to people and vehicles is fully accessed, through data cleaning and processing, the connection of people and vehicle OneID is realized, and the unique identification is recommended. Then, build tags and groups, tags are the basis of the customer service operation platform, through the establishment of the tag system, the user groups, group insight and marketing application are established. Subsequently, select user groups, according to the obtained tags or tag system, select user groups, then, user insight and analysis, through deep analysis and insight of the multidimensional portrait of the target group, realize global insight and analysis. Finally, marketing application, develop differentiated operation strategies for groups with different value stratification, through multi-channel user touch and effect data recovery, optimize strategies with data as a guide, and continuously improve the conversion effect.

[0103] The effect of the technical solution is that the marketing strategy optimization processing method based on user grouping helps enterprises break data silos, establish unified people and vehicle archives, empower enterprises to drive all-link marketing and deep operation with data, and realize enterprise digital transformation and growth. Enterprises can build user tags and portraits through a customer service operation platform, stratify and group users, and through group insight capabilities, dig out group characteristics, deeply understand target user groups, develop targeted operation strategies for users with different characteristics, and through fine operation of customers, change from traffic thinking to user thinking, thereby mastering the initiative of marketing. At the same time, through the integration of various marketing-related systems, all people and vehicle data are fully entered into the lake; through the connection of different system people and vehicle IDs, the uniqueness of individual users and vehicles is realized; tags and groups are established to analyze user groups and individuals, and specific groups are quickly selected; through machine learning and model algorithms, the modeling of subgroups is provided for analysis and decision-making; according to the attributes and consumption prediction of different groups, accurate marketing strategies are established to realize accurate marketing of target groups.

[0104] In an embodiment of the present application, the user and user vehicle related business data of multiple business systems are connected to the customer service operation platform, and the ID graph corresponding to the user is constructed, including:

[0105] S101, the system business original data corresponding to the user and the user vehicle in the multiple business systems is connected to the customer service operation platform; wherein the multiple business systems include but are not limited to marketing system, ERP system and vehicle system;

[0106] S102, the system business original data is processed by data cleaning, and the system business original data processed by data cleaning is transmitted to the target source; wherein the data cleaning processing includes but is not limited to screening, deduplication and replacement;

[0107] S103, the ID mapping logic is configured, the priority in the multi-source data matching process is set according to the source and data integrity of the system business original data combined with the actual business rules, and the unique identification of the user and the corresponding user vehicle is established, wherein the unique identification is OneID;

[0108] S104, the fragmented data of multiple accounts of the same user is associated based on the user ID information, and the fragmented data is integrated to generate the integrated data corresponding to the user ID information; wherein the user ID information includes but is not limited to user ID, mobile phone number and device number.

[0109] The working principle of the above technical solution is as follows: original data access: access the original data of all people and vehicle related data systems such as marketing system, ERP system, vehicle system, etc. to the customer service operation platform.

[0110] Data cleaning and processing: perform cleaning operations such as screening, deduplication, and replacement on the data. After data cleaning is completed, the processed data can be output to the target source.

[0111] Data connection to build OneID system: configure ID mapping logic, set priorities in multi-source data matching process according to data source, data completeness, and actual business rules, and establish a unique identifier (i.e. OneID) for people and vehicles.

[0112] Build ID map: based on user ID, mobile phone number, device number, etc. ID information, associate fragmented data of multiple accounts of the same user, and integrate multi-party data of enterprises.

[0113] The effect of the above technical solution is as follows: by accessing the user and user vehicle related original data in multiple business systems to the customer service operation platform, cross-system data integration is achieved. This provides a comprehensive and unified data view for users, enabling enterprises to better understand users and their vehicles. Data cleaning and processing steps (such as screening, deduplication, and replacement) effectively improve data quality. This ensures the accuracy of subsequent analysis and application, avoiding decision-making errors caused by data errors or duplication. Configure ID mapping logic and set priorities according to data source and data completeness, enabling enterprises to more effectively manage multi-source data. This helps ensure data accuracy and consistency, while improving data processing efficiency. By establishing a unique identifier (OneID) for users and their corresponding user vehicles, accurate identification of users and their vehicles is achieved. This provides more accurate user profiles for enterprises, helping them better provide personalized services and precision marketing. Based on user ID information (such as user ID, mobile phone number, device number, etc.), fragmented data of multiple accounts of the same user is associated and integrated. This helps enterprises better understand user behavior and preferences, providing more personalized services and products. The integrated data provides strong support for business decision-making for enterprises. Through data analysis and mining, enterprises can discover potential market opportunities, optimize business processes, improve user experience, and enhance enterprise competitiveness.

[0114] In summary, this technical solution integrates data from multiple business systems, improves data quality, establishes a unique identifier, and integrates fragmented data, providing a comprehensive, accurate, and efficient data management solution for enterprises, helping them better understand users, optimize business decisions, and enhance competitiveness.

[0115] Specifically, the data quality monitoring is performed on the integrated data corresponding to the user ID information, including:

[0116] The integrated data corresponding to the user ID information is scanned to obtain initial abnormal feature elements contained in the integrated data, wherein the initial abnormal feature elements include a number of missing key fields, weight values corresponding to missing data, a data garbled rate contained in the integrated data, and weight values corresponding to garbled data.

[0117] A row matrix X corresponding to the key fields is formed according to the number of missing key fields in combination with the corresponding weight values c .

[0118] A column matrix Y corresponding to the garbled fields is formed according to the number of garbled fields in combination with the corresponding weight values c .

[0119] According to the number of elements contained in the row matrix X c and the column matrix Y c , the number of elements with fewer elements is used as a reference to perform dimension number padding on the matrix with fewer elements using the number 1 to generate the padded row matrix X and the column matrix Y, wherein the structure of the padded row matrix X and the column matrix Y is as follows:

[0120]

[0121] X represents the row matrix after the padding operation corresponding to the key fields; x 01 , x 02 , ……, x α respectively represent weight values corresponding to the elements contained in the row matrix after the padding operation.

[0122]

[0123] Y represents the column matrix after the padding operation corresponding to the garbled fields; y 01 , y 02 , ……, y β respectively represent weight values corresponding to the elements contained in the column matrix after the padding operation.

[0124] The initial abnormal factor is obtained by using the padded row matrix X and the column matrix Y in combination with the key field missing rate and the garbled rate.

[0125] The initial abnormal factor is obtained by using the padded row matrix X and the column matrix Y in combination with the key field missing rate and the garbled rate.

[0126]

[0127] K represents the initial abnormal factor; σ represents the number of elements contained in the padded row matrix X; μ represents the number of elements contained in the padded column matrix Y; xi represents the weight value corresponding to the i-th element contained in the post-padding row matrix X; y i represents the weight value corresponding to the i-th element contained in the post-padding column matrix Y; P 01 and P 02 respectively represent the data missing rate and the data garbled rate corresponding to the integrated data; represents the norm operator corresponding to the matrix;

[0128] comparing the initial anomaly factor with a preset factor threshold;

[0129] when the initial anomaly factor exceeds the preset factor threshold, data anomaly alarm is performed;

[0130] when the initial anomaly factor does not exceed the preset factor threshold, the time difference between the integrated data update time and the fragmented data update time, the number of missing key fields after data update, the weight value of the missing data, and the data garbled rate and the weight value of the garbled data contained in the integrated data are monitored in real time each time the fragmented data is updated;

[0131] an anomaly factor corresponding to each data update is obtained by using the initial anomaly factor in combination with the time difference between the integrated data update time and the fragmented data update time, the number of missing key fields after data update, the weight value of the missing data, and the data garbled rate and the weight value of the garbled data contained in the integrated data;

[0132] wherein the anomaly factor is obtained by the following formula:

[0133]

[0134] wherein Q represents the anomaly factor; w represents the number of updated fragmented data corresponding to the current data update; T ci represents the time difference between the time when the i-th updated fragmented data completes the update and the time when the data corresponding to the fragmented data in the integrated data completes the update; K represents the initial anomaly factor; K g represents the initial anomaly factor corresponding to the current integrated data after completing the update;

[0135] comparing the anomaly factor with a preset anomaly factor threshold;

[0136] when the anomaly factor exceeds the preset anomaly factor threshold, data anomaly alarm is performed.

[0137] The working principle of the above technical solution is: scanning the integrated data, identifying and extracting initial abnormal feature elements, including the number of missing key fields and their weight values, the data garbled rate and its weight value. According to the number of missing key fields and the corresponding weight value, a row matrix X is constructed c . According to the number of garbled fields and the corresponding weight value, a column matrix Y is constructed c . According to the number of elements in the row matrix X c and the column matrix Y c , the matrix with fewer elements is filled in the number of dimensions using the number 1, and the filled row matrix X and column matrix Y are generated.

[0138] Using the filled row matrix X and column matrix Y, combining the data missing rate and the garbled rate, the initial abnormal factor K is calculated through a specific formula. Compare the initial abnormal factor K with the preset factor threshold value. If K exceeds the threshold value, data abnormality alarm is performed; otherwise, enter the real-time monitoring stage.

[0139] Real-time monitoring of fragmented data update, recording the time difference between the integrated data update time and the fragmented data update time, and the number of missing key fields and the garbled rate after updating. Combine the initial abnormal factor K, the time difference, the number of missing key fields after updating and the garbled rate, and calculate the abnormal factor Q corresponding to each data update through a specific formula. Compare the abnormal factor Q with the preset abnormal factor threshold value. If Q exceeds the threshold value, data abnormality alarm is performed.

[0140] The technical effect of the above technical solution is: through real-time monitoring of fragmented data update, the data quality problem can be found in time, and the real-time performance of data monitoring is improved. By considering multiple factors such as the number of missing key fields, the garbled rate and the time difference, the abnormal factor is calculated, which improves the accuracy of data quality monitoring. The number 1 is used to fill in the number of dimensions of the matrix with fewer elements, so that the row matrix X and the column matrix Y are consistent in structure, which is convenient for subsequent calculation. This filling method is simple and flexible, and does not increase the additional calculation complexity. By setting the factor threshold value and the abnormal factor threshold value, when the data quality exceeds the acceptable range, the alarm mechanism can be automatically triggered to improve the reliability of data processing. The technical solution can adjust the identification rules of key fields and garbled fields and the setting of weight values according to actual business needs, and has good scalability. At the same time, the technical solution is suitable for various types of data integration scenarios and has high applicability.

[0141] The technical solution can more accurately identify abnormal situations in the data by comprehensively considering multiple dimensions such as the number of missing key fields, the rate of garbled codes, and the time difference. Compared with single-dimensional identification methods, this multi-dimensional abnormality identification method can more comprehensively reflect the quality problems of the data. In calculating the initial abnormality factor and the abnormality factor, the technical solution uses simple and clear formulas, avoiding complex iterative calculations or optimization algorithms, thereby improving the calculation efficiency. This enables the technical solution to process and analyze a large amount of data in a short time, meeting the real-time requirements. The technical solution fully utilizes existing data and computing resources when constructing matrices and calculating abnormality factors, avoiding unnecessary waste of resources. For example, by using the number 1 to fill the dimensions of the matrix, the consistency of the matrix structure is ensured, and the introduction of additional data or computational load is avoided. Once data anomalies are detected, the technical solution can immediately trigger an alarm mechanism to notify relevant personnel for processing. This timely abnormality processing mechanism helps to reduce the impact of data errors on business decisions and improves the reliability and efficiency of data processing.

[0142] By scanning and analyzing the initial abnormal characteristic elements such as the number of missing key fields in the integrated data, the weight value of missing data, the rate of garbled data, and the weight value of garbled data, various abnormal situations existing in the data can be accurately captured, key problems can be avoided, and the accuracy of data abnormality judgment can be improved. At the same time, by calculating the initial abnormality factor and the abnormality factor through the above formula, the data abnormality is quantified, the evaluation of data quality is more accurate, and the actual quality status of the data can be more accurately reflected, providing a more reliable basis for subsequent decision-making. At the same time, when the initial abnormality factor does not exceed the threshold value, the time difference between the integration data update time and the fragmented data update time can be monitored in real time each time the fragmented data is updated, potential problems in the data update process can be found in time, and the data quality monitoring can be ensured to be real-time and can quickly respond to data changes. When the initial abnormality factor or the abnormality factor corresponding to each data update exceeds the preset threshold value, data abnormality alarm can be performed in time, so that relevant personnel can understand data quality problems at the first time, quickly take measures for processing, and reduce the influence time of data abnormality on business. In addition, the above technical solution not only considers the key field missing and data garbled, but also combines the corresponding weight value and the time difference in the data update process, and evaluates the data quality from multiple dimensions to avoid the limitations of single-dimensional evaluation, making the data quality monitoring more comprehensive. By continuously obtaining the abnormality factor corresponding to each data update, the change of data quality can be dynamically tracked, and the data quality can be continuously monitored whether in the initial integration stage or in the subsequent update process, so as to comprehensively ensure the stability of data quality. In the data update process, the abnormality factor can be dynamically calculated according to the number of updated fragmented data, the update time difference, and the updated abnormality, and the quality of data update of different scales and frequencies can be effectively monitored, which has strong adaptability and flexibility. By setting the preset factor threshold and the abnormality factor threshold, the sensitivity to data abnormality can be flexibly adjusted according to different business needs and data quality standards, so that the data quality monitoring is more in line with the requirements of actual business scenarios, and the adaptability and configurability of the system are improved.

[0143] In summary, the technical effects of the above technical solution on performance indicators not only include real-time and accuracy of data quality monitoring, flexibility of matrix dimension completion, reliability of alarm mechanism, scalability and applicability, but also further include accuracy of abnormality identification, improvement of calculation efficiency, optimization of resource utilization, timeliness of abnormality processing, and potential technical effects such as visualization and interpretability. These technical effects together constitute the overall advantages of the technical solution in performance indicators. At the same time, the technical solution comprehensively considers multiple factors to realize comprehensive, accurate, and real-time data quality monitoring of the integrated data corresponding to the user ID information, and improves the reliability and efficiency of data processing.

[0144] In one embodiment of the application, a user's corresponding label is constructed using relevant business data of the user and the user's vehicle, including:

[0145] S201, retrieve the behavior data and attribute data corresponding to the user, and create labels corresponding to the user and the user's vehicle according to a preset business logic and different model categories;

[0146] S202, present and manage the label content in a structural manner according to the label structure corresponding to each user and the user's vehicle, wherein the label structure includes the number, classification, and hierarchical relationship of the labels.

[0147] The different model categories include AIPL model, 5A model, and RFM model; the structures of the AIPL model, 5A model, and RFM model are as follows:

[0148] In the AIPL model, A represents the brand awareness population, I represents the brand interest population, P represents the brand purchase population, and L represents the brand loyalty population.

[0149] The 5A model includes parameters A1, A2, A3, A4, and A5, wherein A1 represents a customer who passively accepts information, A2 represents a customer whose brand impression increases, A3 represents a customer who actively searches for information driven by curiosity, A4 represents a customer who takes action, and A5 represents a customer who has loyalty to the brand and promotes it.

[0150] The RFM model assesses the R value, F value, and M value of each user to determine their corresponding intervals, thereby dividing users into eight types of user value, including important value customers, important exchange customers, important deep cultivation customers, important retention customers, potential customers, new customers, general maintenance customers, and lost customers. The R value represents the most recent consumption and reflects the activity level of a customer; the F value represents the consumption frequency and reflects the loyalty of a customer; and the M value represents the consumption amount and reflects the contribution of a customer.

[0151] The working principle of the above technical solution is as follows: label setting: based on behavior, attribute, and other data, a person and vehicle corresponding label is created based on business logic or model capability.

[0152] Label system: the label system is composed of labels, and the label content is presented and managed in a structural manner, including the number, classification, and hierarchical relationship of the labels.

[0153] Based on business logic or model capability, a label model is established, as follows:

[0154] AIPL model: a means of quantifying and linking brand audience assets. Specifically: A (Awareness) represents brand awareness audience; I (Interest) represents brand interest audience; P (Purchase) represents brand purchase audience; L (Loyalty) represents brand loyalty audience

[0155] 5A model: a marketing model proposed by Philip Kotler in Marketing Revolution 4.0. Specifically: A1 Understanding (Aware) refers to customers passively receiving information; A2 Appeal refers to customers with increased brand image; A3 Ask refers to customers who actively search for information driven by curiosity; A4 Action refers to customers who take action; A5 Advocate refers to customers who have loyalty to the brand and spread the word.

[0156] RFM model: by evaluating the R, F and M values of each user, it can be mapped to different intervals, thereby dividing users into 8 types of user value. Specifically: important value customers, important exchange customers, important deepening customers, important retention customers, potential customers, new customers, general maintenance customers, and lost customers. Among them, R represents the last time of consumption (Recency), which reflects the activity level of a customer; F represents the frequency of consumption (Frequency), which reflects the loyalty of a customer; M represents the consumption amount (Monetary), which reflects the contribution of a customer.

[0157] The technical scheme has the effects that: by calling the behavior data and attribute data corresponding to the user, and creating labels corresponding to the user and the user vehicle according to the preset business logic and different model types (such as AIPL model, 5A model and RFM model), the fine construction of the user portrait can be realized. These labels can more accurately reflect the state, interest, behavior pattern and the like of the user, which helps the enterprise to understand the user more deeply. The labels corresponding to each user and the user vehicle are presented and managed in a structural manner (such as the number, classification and hierarchical relationship of the labels), so that the label system is more clear and orderly. This helps the enterprise to more efficiently use the labels for user analysis, precision marketing and the like, and improves the work efficiency. With the help of different model types such as the AIPL model, the 5A model and the RFM model, the enterprise can divide the users more carefully, so as to formulate more accurate marketing strategies. For example, for brand loyal people (L) or A5-level customers, the enterprise can launch more loyalty reward programs or high-end services; for potential customers or new customers, the enterprise can adopt more attractive promotion strategies to guide them to become loyal users. The RFM model divides the user value types by evaluating the R value (the last time of consumption), the F value (the consumption frequency) and the M value (the consumption amount) of the user, which provides a scientific and objective user value evaluation method for the enterprise. This helps the enterprise to identify high-value users, potential users and lost users, so as to formulate targeted user maintenance and recovery strategies.

[0158] The technical scheme realizes the deep mining and analysis of user data by constructing the user label system. This provides rich data support for the enterprise, so that the enterprise can make decisions based on data, and improves the accuracy and scientificity of the decisions. By carefully dividing the users and carrying out precision marketing, the enterprise can better meet the needs and expectations of the users, and improve the user experience. At the same time, by identifying and focusing on the high-risk groups such as lost users, the enterprise can take timely measures to recover these users, so as to maintain user satisfaction and loyalty.

[0159] In summary, the technical scheme constructs the user label system, realizes efficient label management, accurate marketing strategies, scientific user value evaluation and data-driven decision-making, and provides comprehensive, scientific and efficient user management and marketing strategy support for the enterprise, which helps the enterprise to improve the competitiveness and realize sustainable development.

[0160] In an embodiment of the present application, the user is grouped according to the labels corresponding to the user, a target grouping package corresponding to the user is obtained, and the data information of the user and the data information having an upstream and downstream dependency relationship with the user are displayed according to the target grouping package, comprising:

[0161] S301, create a target group package corresponding to a user through a preset group rule or an uploaded list; wherein the target group package is used for insight or pushing to a downstream marketing channel, so as to realize accurate reach of target customers;

[0162] S302, automatically generate a group movement record according to the target group package, wherein the group movement record is used to show the running record, historical performance and trend of the group, so as to timely understand the latest situation of the group task;

[0163] S303, view blood relationship details corresponding to the target group package according to the target group package, and display data sources, people group packages and label resources having upstream and downstream dependent relationship with the target group package.

[0164] The way of creating a target group package corresponding to a user is as follows:

[0165] Rule creation: create a group by using labels, behaviors, people group packages, detailed data and user attribute data through a visual component;

[0166] Upload creation: quickly create a target group package by uploading a local seed people group file.

[0167] The working principle of the above technical solution is as follows: first, create a group package: create a group package by a certain rule or an uploaded list, and the target group package after creation can be used for insight or pushing to a downstream marketing channel, so as to realize accurate reach of target customers. The way of creating a group package is as follows: 1) rule creation: create a group by using labels, behaviors, people group packages, detailed data, user attribute data through a visual component.

[0168] 2) upload creation: quickly create a target group package by uploading a local seed people group file.

[0169] Seed people group file: according to the needs of a specific business scenario, people who have the same demand and interest for goods and services are called seed people group, which generally includes basic attributes of users such as gender, age, occupation, region, and behavior preference information.

[0170] Extended people group / similar people group: people who have the same characteristics as the seed people group are called extended people group.

[0171] Target group acquisition: find the extended people group by uploading the seed people group, and then take the intersection of the extended people groups of each seed people group as the final target user for delivery.

[0172] Then, the group operation record is recorded: based on the user corresponding to the above-mentioned people group package, the group operation record can be automatically generated, the running record, historical performance and trend of the group are displayed, and the latest situation of the group task is understood in time.

[0173] Finally, the group data blood relationship: based on the user corresponding to the above-mentioned people group package, the blood relationship details of the people group package are viewed, and the data source, people group package and label resource having upstream and downstream dependent relationship with the people group package are displayed as the center of the people group package.

[0174] The effect of the above technical solution is that: through the preset group rule or the uploaded list, the target group package corresponding to the user can be flexibly created. This way not only supports the combination group based on labels, behaviors, people group packages, detailed data and user attribute data, but also allows quick creation of groups by uploading local files, thereby realizing accurate division of target customers. This helps enterprises more accurately identify target user groups and improve the targeting and effectiveness of marketing activities. The automatically generated group movement record displays the running record, historical performance and trend of the group. This enables enterprises to timely understand the latest progress of the group task, monitor the group effect, and adjust and optimize the group strategy. This improves the transparency and efficiency of group management. According to the target group package, the corresponding blood relationship details are viewed, and the data source, people group package and label resource having upstream and downstream dependent relationship with the target group package are displayed as the center of the target group package. The visualization of data blood relationship and dependent relationship helps enterprises understand the source, flow and conversion process of data, ensuring the accuracy and consistency of data. At the same time, it also helps enterprises better manage and utilize data resources and improve data governance. The target group package can be used to understand user behavior, preferences and needs, and provide accurate user portraits and segmentation strategies for downstream marketing channels. This helps enterprises develop more targeted marketing strategies and improve the conversion rate and ROI of marketing activities. At the same time, through automation and optimization of group management, enterprises can more efficiently manage and operate user groups and reduce marketing costs. The group management, data blood relationship visualization and marketing effect monitoring functions provided by the technical solution provide enterprises with rich data support and analysis tools. This helps enterprises better understand users and market dynamics and develop more scientific decision-making and strategic planning.

[0175] In summary, the technical solution provides comprehensive and efficient user management and marketing strategy support for enterprises through precise user grouping, group task monitoring and management, data blood relationship and dependent relationship visualization, and marketing efficiency improvement. This helps enterprises better understand and utilize user data, improve the targeting and effectiveness of marketing activities, and achieve business growth and sustainable development.

[0176] In an embodiment of the present application, based on the labels corresponding to the user, the multi-dimensional portrait of the target group corresponding to the target group package is obtained through deep analysis and insight, including:

[0177] S401, retrieve the label corresponding to the user, display the main information of the user and the user vehicle corresponding to the label, wherein the main information includes basic information corresponding to the user, cover label and behavior timeline, and the group; key information of a user is quickly understood through individual portrait, so as to further marketing action for the user;

[0178] S402, retrieve the target group package corresponding to the user and the user vehicle, generate insight report for the target group corresponding to the main body of the user and the user vehicle, and mine group characteristics, and meanwhile, cross listening and drilling analysis and other methods are used to deeply understand the characteristics of each dimension of the user, wherein the group characteristics are used to guide enterprise operation decision;

[0179] S403, extract the explicit characteristics and implicit characteristics of the target group package corresponding to the user and the user vehicle, and mine the label combination matched with the target group package according to the explicit characteristics and implicit characteristics, and meanwhile, generate the group package meeting the preset demand according to the indexes such as saliency, coverage and group number, so as to further analyze and decide;

[0180] S404, create a life cycle label, and insight the life cycle of the user through the life cycle label, and meanwhile, export the target group package according to the user stage, carry out fine operation, promote the purchase decision of the user, and continuously improve the loyalty of the user to the brand.

[0181] The working principle of the above technical solution is: individual portrait analysis: based on label setting, detailed display of main information of a person, vehicle and other subjects is generated, including basic information of the subject, cover label, behavior timeline, and the group, key information of a user is quickly understood through individual portrait, so as to further marketing action for the user.

[0182] Group portrait insight: based on the target group package corresponding to the user, the group of the main body of the user and the vehicle is selected, the target group is generated, the insight report is generated, the group salient characteristics are mined, and the enterprise marketing decision is guided. Meanwhile, cross listening, drilling analysis and other methods can be used to deeply understand the characteristics of each dimension. Multi-feature analysis: based on model and algorithm, the explicit and implicit characteristics of the group are understood, and the label combination most suitable for the group is mined. Meanwhile, according to the indexes such as saliency, coverage and group number, the group package meeting the specific demand is generated, so as to further analyze and decide.

[0183] Life cycle analysis: through creating a life cycle label, the life cycle of the user is understood based on the model label. Meanwhile, the group package is exported according to the user stage, fine operation is carried out, the purchase decision of the user is promoted, and the loyalty of the user to the brand is continuously improved.

[0184] The effect of the above technical solution is: by calling the corresponding label of the user and displaying the main information of the user and the user's vehicle (such as basic information, covering label, behavior timeline and the group the user belongs to), the enterprise can quickly understand the key information of the user, so as to carry out more accurate and personalized marketing actions for the user. This helps to improve user experience, increase user stickiness, and promote user purchase decision. The target group package corresponding to the user and the user's vehicle generates an insight report, excavates group characteristics, and uses cross analysis, drilling analysis and other methods for deep insight, which helps the enterprise better understand the behavior patterns, preferences and needs of the target group. These information can provide strong support for the enterprise to make business decisions, optimize product design and improve marketing strategies. Extracting the explicit and implicit features of the target group package and mining matching label combinations according to these features helps the enterprise more accurately identify target user groups and optimize group strategy. At the same time, by generating group packages that meet the preset requirements according to indicators such as saliency, coverage and group number, the accuracy and effectiveness of the group can be further improved, providing strong support for subsequent marketing and operation. Creating a life cycle label and using the life cycle label to understand the user life cycle helps the enterprise understand the needs and behavior changes of users at different stages. Exporting the target group package according to the user stage and carrying out fine operation can provide more appropriate services and products for users at different stages, thereby promoting user purchase decision and improving user loyalty to the brand. The technical solution realizes the fine user portrait, group feature mining, label combination optimization and user life cycle management through deep analysis and insight of user data. These measures help the enterprise make more scientific decisions, improve operational efficiency, reduce marketing costs and achieve sustainable development.

[0185] In summary, the technical solution provides comprehensive and efficient data analysis and marketing strategy support for enterprises through fine user portrait, deep insight into group characteristics, optimization of label combination and group strategy, and implementation of user life cycle management. This helps enterprises better understand and utilize user data, improve marketing effectiveness, enhance user loyalty, and achieve business growth and competitive advantage.

[0186] In one embodiment of the present application, different operation strategies are developed for different value stratified groups according to the target group package corresponding to the user, the user is reached through multiple channels, and effect data is collected, and the effect data is used to guide the optimization of the strategy, including:

[0187] S501, according to the group attribute and behavior prediction result of the target group package corresponding to the user, develop an operation strategy;

[0188] S502, connect to CRM and other systems to obtain a customer panoramic portrait, and execute a marketing strategy according to the customer panoramic portrait;

[0189] S503, execute personalized operation strategy for maximizing the extension of user life cycle by constructing churn prediction model and insightting churn population.

[0190] The working principle of the above technical solution is: fine operation: based on the target group package corresponding to the user, specific operation strategies are formulated according to the group attribute and behavior prediction, realizing more accurate, intelligent and personalized touch, promoting conversion and repurchase.

[0191] Sales scenario empowerment: interface with CRM and other systems to assist sales personnel in obtaining customer panoramic portrait and formulating marketing strategies to effectively improve lead conversion rate.

[0192] Churned user recovery: by constructing churn prediction model, insightting churn population and formulating personalized operation strategy, the user life cycle is maximized.

[0193] The effect of the above technical solution is: according to the group attribute and behavior prediction result of the target group package corresponding to the user, different operation strategies can be formulated for different value stratified groups. This precise strategy formulation helps to improve the pertinence and effectiveness of marketing activities and avoid resource waste. By interfacing with CRM and other systems, customer panoramic portrait is obtained, which provides enterprises with more comprehensive and in-depth user information. Based on these information, enterprises can execute more accurate and personalized marketing strategies to meet the individual needs of users, improve user experience and satisfaction. Constructing churn prediction model and insightting churn population helps enterprises to discover potential churned users in time and take personalized operation strategies to retain them. This not only maximizes the extension of user life cycle, but also improves user retention rate and loyalty, bringing sustainable value to enterprises. The technical solution supports multi-channel touch of users, including social media, email, SMS, telephone and other ways. This helps to ensure that information can be accurately and timely conveyed to users, improving the arrival rate and reading rate of information. At the same time, by recycling effect data, enterprises can evaluate the execution effect of marketing strategies and provide data support for subsequent strategy optimization. Guided by recycled effect data, enterprises can continuously optimize operation strategies. This data-driven decision-making approach helps to improve the accuracy and scientificity of decision-making, reduce trial and error cost, and improve enterprise operation efficiency and market competitiveness. Through the implementation of the above technical solution, enterprises can more accurately target user groups, formulate effective marketing strategies and touch users through multiple channels. This helps to improve overall operation efficiency, reduce marketing cost and improve return on investment (ROI).

[0194] In summary, the technical scheme provides comprehensive and efficient operation support for enterprises through precise formulation of operation strategies, acquisition of customer panoramic portraits, construction of loss prediction models, multi-channel user reach, recovery of effect data, and data-driven strategy optimization. This helps enterprises better understand and utilize user data, improve operational efficiency and market competitiveness, and achieve sustainable development.

[0195] In one embodiment of the present application, the marketing strategy optimization processing method based on user grouping further comprises:

[0196] Step 1, real-time monitoring of data transmission operation parameters between each business system and the customer service operation platform, wherein the data transmission operation parameters include data call response time standard deviation per unit time, process to available CPU ratio per unit time, API request frequency per unit time, and security event occurrence rate per unit time; and the unit time is 3-8 minutes.

[0197] Step 2, using the data transmission operation parameters to form a feature vector corresponding to each business system per unit time; wherein the feature vector structure is as follows:

[0198]

[0199] Wherein, A represents the feature vector corresponding to each unit time; T b represents the data call response time standard deviation per unit time; P c represents the process to available CPU ratio per unit time; F represents the API request frequency per unit time; P a represents the security event occurrence rate per unit time;

[0200] Step 3, standard deviation processing of the feature vector corresponding to each business system per unit time, generating a standardized feature vector, and the standardized feature vector corresponding to each business system per unit time is obtained by the following formula:

[0201]

[0202] Wherein, A b represents the standardized feature vector; δ(T b ), δ(P c ), δ(F), and δ(P arespectively represent the normalized parameters corresponding to the standard deviation of the data call response time per unit time, the ratio between the process and the available CPU per unit time, the API request frequency per unit time, and the security event occurrence rate per unit time;

[0203] Step 4, obtaining a feature coefficient corresponding to each business system by using the normalized feature vectors corresponding to all the unit times experienced by each business system, wherein the feature coefficient is obtained by the following formula:

[0204]

[0205] wherein ξ represents the feature coefficient corresponding to each business system; n represents the number of all the unit times experienced by each business system; represents the Euclidean norm corresponding to the normalized feature vector corresponding to the i-th unit time, the cumulative effect of the overall load; represents the Manhattan norm corresponding to the normalized feature vector corresponding to the i-th unit time, reflecting the overall strength or importance of the feature vector; ε represents a preset minimum constant, used to prevent from being 0; δ(T b ) i , δ(P c ) i , δ(F) i , and δ(P a ) i respectively represent the normalized parameters corresponding to the standard deviation of the data call response time per unit time, the ratio between the process and the available CPU per unit time, the API request frequency per unit time, and the security event occurrence rate per unit time; α and β respectively represent a first adjustment coefficient and a second adjustment coefficient, and the value range of the first adjustment coefficient and the second adjustment coefficient is 0.53-0.72, 0.47-0.79;

[0206] Step 5, obtaining a comprehensive load coefficient by using the feature coefficient corresponding to each business system;

[0207] wherein the comprehensive load coefficient is obtained by the following formula:

[0208]

[0209] wherein S represents the comprehensive load coefficient; m represents the number of business systems; ξ i represents the feature coefficient corresponding to the i-th business coefficient; ξ b represents the standard deviation of the feature coefficient corresponding to the m business coefficients; ξ bi represents the standard deviation of the feature coefficient corresponding to the n unit times corresponding to the i-th business coefficient;

[0210] Step 6, compare the comprehensive load coefficient with a preset comprehensive coefficient threshold value;

[0211] Step 7, when the comprehensive load coefficient exceeds the preset comprehensive coefficient threshold value, it is determined that the communication running load of the customer service operation platform is overloaded, and a load overload warning is performed.

[0212] The working principle of the above technical solution is: real-time collection of data transmission running parameters between each business system and the customer service operation platform, including data retrieval response time standard deviation, process and available CPU ratio, API request frequency and security event occurrence rate. These parameters are collected in units of 3-8 minutes to ensure the real-time and accuracy of the data. Using the collected data transmission running parameters, a feature vector is constructed for each business system in each unit time. The feature vector contains four dimensions of information: data retrieval response time standard deviation, process and available CPU ratio, API request frequency and security event occurrence rate. The feature vector of each business system is processed by standard deviation to generate a standardized feature vector. Standardization processing helps to eliminate the dimensional differences between different parameters, making subsequent calculations more accurate. Using all the standardized feature vectors corresponding to each unit time experienced by each business system, a feature coefficient is calculated. The feature coefficient considers the Manhattan norm and Euclidean norm of the feature vector, as well as a preset minimum constant, to reflect the overall load of the business system. According to the feature coefficient of each business system, a comprehensive load coefficient is calculated, which reflects the overall communication running load of all business systems. Compare the comprehensive load coefficient with a preset comprehensive coefficient threshold value. When the comprehensive load coefficient exceeds the preset comprehensive coefficient threshold value, it is determined that the communication running load of the customer service operation platform is overloaded, and a load overload warning is performed.

[0213] The effect of the above technical solution is: by real-time monitoring of data transmission running parameters between each business system and the customer service operation platform, potential communication running load problems can be discovered in time. When the load is overloaded, the system can automatically perform a warning, which helps enterprises to take timely measures to avoid service interruption or performance degradation. By constructing a feature vector, performing standardization processing and calculating a feature coefficient, the communication running load of each business system can be accurately evaluated. The calculation of the comprehensive load coefficient further reflects the overall load of all business systems, providing strong support for enterprise decision-making. Through real-time monitoring and warning, enterprises can discover and solve communication running load problems in time, thereby improving the stability and reliability of the system. This helps to improve user experience and enhance the market competitiveness of enterprises.

[0214] On the other hand, by monitoring a plurality of key data transmission operation parameters (such as data retrieval response time standard deviation, process-to-available CPU ratio, API request frequency, and security event occurrence rate) in real time, the technical solution can comprehensively reflect the communication operation state between the business system and the customer service operation platform. Standard deviation processing of the feature vector eliminates the dimensional differences between different parameters, making the calculation process more unified and accurate. By calculating the comprehensive load coefficient, the technical solution can comprehensively consider the overall load of all business systems, thereby more accurately evaluating the communication operation load of the customer service operation platform. The technical solution can monitor data transmission operation parameters in real time, and can immediately issue a warning once an abnormal situation is found, thereby improving the timeliness of the warning. When the comprehensive load coefficient exceeds the preset comprehensive coefficient threshold, the system will automatically issue a warning without human intervention, further improving the efficiency and accuracy of the warning.

[0215] Through real-time monitoring and warning, enterprises can timely discover and solve potential communication operation load problems, thereby avoiding service interruption or performance degradation. According to the monitoring and warning results, enterprises can reasonably adjust resource allocation, such as increasing servers and optimizing network architecture, to improve the stability and reliability of the system. By ensuring the stable operation of the customer service operation platform, the technical solution helps to improve user experience and enhance the market competitiveness of enterprises. The data and calculation results collected by the technical solution can provide reference for the optimization of marketing strategies. For example, by analyzing the load of different business systems, enterprises can adjust marketing strategies to reduce the pressure on the customer service operation platform. Combined with user segmentation technology, enterprises can develop more accurate marketing strategies for different user groups, thereby improving marketing effectiveness and conversion rates.

[0216] In summary, the technical effects of the technical solution and calculation process on performance indicators mainly include improving evaluation accuracy, improving warning timeliness, enhancing system stability and reliability, and supporting marketing strategy optimization. These technical effects help enterprises better monitor and manage the communication operation load of the customer service operation platform, improve user experience and market competitiveness. At the same time, the technical solution realizes precise evaluation and warning of the communication operation load of the customer service operation platform by monitoring data transmission operation parameters in real time, constructing a feature vector, calculating a feature coefficient, and calculating a comprehensive load coefficient. This helps enterprises timely discover and solve potential problems, improve the stability and reliability of the system, and provide strong support for the optimization of marketing strategies.

[0217] An embodiment of the present application proposes a marketing strategy optimization processing system based on user segmentation, as shown in Figure 2 The marketing strategy optimization processing system based on user segmentation includes:

[0218] An ID graph construction module is configured to access user and user vehicle related business data of multiple business systems into a customer service operation platform, and construct an ID graph corresponding to the user;

[0219] A label construction module is configured to construct a label corresponding to the user by using the user and user vehicle related business data;

[0220] A target group package acquisition module is configured to perform group processing on the user according to the label corresponding to the user, acquire a target group package corresponding to the user, and display the user and data information having an upstream and downstream dependency relationship of the user according to the target group package;

[0221] A multi-dimensional portrait acquisition module is configured to acquire a multi-dimensional portrait of a target group corresponding to the target group package by deep analysis and insight based on the label corresponding to the user;

[0222] An operation strategy acquisition and optimization module is configured to formulate a differentiated operation strategy for different value stratified groups according to the target group package corresponding to the user, reach the user through multiple channels and collect effect data, and optimize the strategy by using the effect data as a guide.

[0223] The working principle of the above technical solution is that the above technical solution of the embodiment establishes a unified person and vehicle file, enables enterprises to drive full-link marketing and deep operation with data, and realizes digital transformation and growth of the enterprise. The enterprise can construct a user label and portrait through the customer service operation platform, stratify and group the user, and mine significant features of the group through group insight ability, deeply understand the target user group, formulate a targeted operation strategy for users with different characteristics, change from a traffic thinking to a user thinking through fine operation of the customer, thereby master the initiative of marketing. At the same time, by establishing a customer service operation platform, the fusion of marketing related systems is realized, and all person and vehicle data is fully entered into the lake. By connecting the person and vehicle IDs of different systems, the user and vehicle are uniquely identified. The label and group are established, the user group and individual are analyzed, the specific group is quickly selected, the modeling of the sub-group is provided by machine learning and model algorithm, the modeling is used for analysis and decision, the precise marketing strategy is established according to the attributes and consumption prediction of different groups, and the precise marketing of the target group is realized.

[0224] First, data integration and fusion, the business system data related to people and vehicles is fully accessed, and the unique identification of people and vehicles is realized through data cleaning and processing. Then, the label and group are constructed, the label is the basis of the customer service operation platform, and the establishment of the label system is used to group users, group insight and marketing application. Subsequently, the user group is selected according to the obtained label or label system, and then the user insight and analysis are carried out through deep analysis and insight of the multidimensional portrait of the target group to realize global insight and analysis. Finally, the marketing application is carried out for different value stratified groups to develop differentiated operation strategies, and the effect data is collected through multi-channel user touch to optimize the strategy guided by data and continuously improve the conversion effect.

[0225] The effect of the above technical scheme is that the marketing strategy optimization processing system based on user group proposed in the embodiment helps enterprises to break the data island, establish unified people and vehicle archives, empower enterprises to drive all-link marketing and deep operation with data, and realize enterprise digital transformation and growth. Enterprises can build user labels and portraits through the customer service operation platform, stratify and group users, and through group insight ability, dig out group characteristics, deeply understand target user groups, develop targeted operation strategies for users with different characteristics, change from traffic thinking to user thinking through fine operation of customers, thereby master the initiative of marketing, at the same time, through the fusion of marketing related systems, all people and vehicle data are fully accessed; through the connection of different system people and vehicle ID, the unique identification of users and vehicles is realized; the label and group are established, and the portrait analysis of user groups and individuals is carried out to realize the rapid selection of specific groups; through machine learning and model algorithm, the modeling of the sub-group is provided for analysis and decision; according to the attributes and consumption prediction of different groups, the precise marketing strategy is established to realize the precise marketing of target groups.

[0226] Obviously, those skilled in the art can make various modifications and variations to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application belong to the scope of the claims of the present application and their equivalent technologies, the present application also intends to include these modifications and variations.

Claims

1. A marketing strategy optimization method based on user segmentation, characterized in that, The marketing strategy optimization method based on user segmentation includes: The platform integrates relevant business data on users and their vehicles from multiple business systems and constructs an ID graph corresponding to each user. Utilize relevant business data about users and their vehicles to construct user-specific tags; Users are segmented according to their corresponding tags, target segmentation packages are obtained for each user, and data information of the user and its upstream and downstream dependencies are displayed based on the target segmentation packages. Based on user-specific tags, multi-dimensional profiles of the target groups corresponding to the target segmentation packages are obtained through in-depth analysis and insights. Based on the target segmentation packages corresponding to the users, differentiated operation strategies are formulated for groups with different value levels. Users are reached through multiple channels and performance data is collected. The performance data is used to guide the optimization of the strategies. This involves cleaning and integrating user and vehicle-related business data from multiple business systems to obtain integrated data corresponding to user ID information. Data quality monitoring is then performed on this integrated data, including: The integrated data is used to form a row matrix corresponding to the key fields by combining the number of missing key fields with their corresponding weight values, and a column matrix corresponding to the garbled fields is formed by combining the number of garbled fields with their corresponding weight values. The row and column matrices are then used to obtain the initial anomaly factor. The initial anomaly factor is compared with a preset factor threshold. When the initial anomaly factor exceeds the preset factor threshold, a data anomaly alarm is triggered. When the initial anomaly factor does not exceed the preset factor threshold, the anomaly factor corresponding to each data update is obtained by combining the initial anomaly factor with the time difference between the integrated data update time and the fragmented data update time. When the anomaly factor exceeds the preset anomaly factor threshold, a data anomaly alarm is triggered. Meanwhile, the marketing strategy optimization method based on user segmentation also includes: Real-time monitoring of data transmission operation parameters between various business systems and the customer service operation platform; using the data transmission operation parameters to form a feature vector for each business system at each unit of time. The standard deviation of the feature vector for each unit of time corresponding to each business system is processed to generate a standardized feature vector; the feature coefficients are obtained by using the standardized feature vectors corresponding to all units of time that have been experienced for each business system. The characteristic coefficients are obtained using the following formula: Where ξ represents the characteristic coefficient corresponding to each business system; n represents the number of all time units that have been experienced for each business system; Let Euclidean norm be the Euclidean norm of the standardized eigenvector corresponding to the i-th unit of time. Let represent the Manhattan norm of the standardized eigenvector corresponding to the i-th unit time; ε represents a preset minimum constant used to prevent... =0; δ(T) b ) i δ(P) c ) i δ(F) i and δ(P) a ) i These represent the standardized parameters corresponding to the standard deviation of data retrieval response time, the ratio between processes and available CPUs, the API request frequency, and the security event occurrence rate for the i-th unit of time, respectively; α and β represent the first adjustment coefficient and the second adjustment coefficient, respectively, and the values ​​of the first adjustment coefficient and the second adjustment coefficient are in the range of 0.53-0.72 and 0.47-0.79, respectively. The overall load factor is obtained using the characteristic coefficients corresponding to each business system; wherein, the overall load factor is obtained by the following formula: Where S represents the overall load factor; m represents the number of business systems; ξ i ξ represents the characteristic coefficient corresponding to the i-th business coefficient; b ξ represents the standard deviation of the characteristic coefficients corresponding to m business coefficients; bi This represents the standard deviation of the characteristic coefficient corresponding to the i-th business coefficient over n units of time; The comprehensive load factor is compared with a preset comprehensive load factor threshold; when the comprehensive load factor exceeds the preset comprehensive load factor threshold, it is determined that the communication operation load of the customer service operation platform is overloaded, and an overload warning is issued.

2. The marketing strategy optimization method based on user segmentation according to claim 1, characterized in that, The platform integrates user and vehicle-related business data from multiple business systems into the customer service operations platform, and constructs an ID graph corresponding to each user, including: The system business raw data related to the user and user vehicle from multiple business systems are connected to the customer service operation platform; wherein, the multiple business systems include the marketing system, ERP system and vehicle system; The system's original business data is cleaned and then transmitted to the target source; the data cleaning process includes filtering, deduplication, and replacement. Configure ID mapping logic, and set the priority in the multi-source data matching process according to the source and data integrity of the original data of the system business and the actual business rules. At the same time, establish a unique identifier for the user and its corresponding user vehicle, wherein the unique identifier is OneID. Fragmented data from multiple accounts of the same user are linked based on user ID information, and the fragmented data is integrated to generate integrated data corresponding to the user ID information; wherein, the user ID information includes user ID, mobile phone number and device number.

3. The marketing strategy optimization method based on user segmentation according to claim 1, characterized in that, Utilize relevant business data about users and their vehicles to construct user-specific tags, including: Retrieve user behavior and attribute data, and create user-vehicle tags based on preset business logic and different model types; The tags corresponding to each user and their vehicle are presented and managed in a structured manner, whereby the tag structure includes the number of tags, their classification, and their hierarchical relationship.

4. The marketing strategy optimization method based on user segmentation according to claim 3, characterized in that, The different model types include AIPL model, 5A model, and RFM model; the structures of the AIPL model, 5A model, and RFM model are as follows: In the AIPL model, A represents the brand-aware audience; I represents the brand-interested audience; P represents the brand-purchasing audience; and L represents the brand-loyal audience. The 5A model includes parameters A1, A2, A3, A4, and A5, where A1 represents customers passively receiving information; A2 represents customers whose brand impression increases; A3 represents customers who actively search for information driven by curiosity; A4 represents customers who take action; and A5 represents customers who are loyal to the brand and promote it. The RFM model categorizes users into eight user value types by evaluating their R, F, and M values ​​and assigning them to different intervals. The R value represents the most recent purchase, reflecting a customer's activity level; the F value represents purchase frequency, reflecting a customer's loyalty; and the M value represents the purchase amount, reflecting a customer's contribution. These eight user value types include: key value customers, key customer retention customers, key customer retention customers, key customer repurchase customers, potential customers, new customers, general customer retention customers, and churned customers.

5. The marketing strategy optimization method based on user segmentation according to claim 1, characterized in that, Users are segmented based on their corresponding tags to obtain target segmentation packages for each user. Data information about the user and their upstream / downstream dependencies is then displayed based on these target segmentation packages, including: By using preset grouping rules or upload lists, target grouping packages corresponding to users are created; wherein, the target grouping packages are used to gain insights or push to downstream marketing channels, thereby achieving precise reach to target customers; Automatically generate cluster motion records based on the target cluster package, wherein the cluster motion records are used to display the operation record, historical performance and trend of the cluster; View the lineage details corresponding to the target segment package based on the target segment package, and display the data sources, audience packages and tag resources that have upstream and downstream dependencies with the target segment package as the center.

6. The marketing strategy optimization method based on user segmentation according to claim 5, characterized in that, The method for creating a target group package for a user is as follows: Rule creation: Create segments by combining tags, behaviors, audience packages, detailed data, and user attribute data using visual components; Upload and create: Quickly create a target subgroup package by uploading a local seed population file.

7. The marketing strategy optimization method based on user segmentation according to claim 1, characterized in that, Based on user-specific tags, a multi-dimensional profile of the target group corresponding to the target segment package is obtained through in-depth analysis and insight, including: Retrieve the user's corresponding tags and display the main information of the user and the user's vehicle based on the tags. The main information includes the user's basic information, covering tags and behavior timeline, as well as the user's group. The system retrieves target segmentation packages corresponding to users and user vehicles, generates insight reports for the target groups corresponding to the main body of users and user vehicles, and mines group characteristics. At the same time, it uses cross-listening and drill-down analysis to conduct in-depth insights into the characteristics of users in various dimensions. The group characteristics are used to guide business decision-making. Extract the explicit and implicit features of the target group packages corresponding to users and user vehicles, and mine the tag combinations that match the target group packages based on the explicit and implicit features. At the same time, generate group packages that meet the preset requirements according to saliency, coverage and number of groups. Create lifecycle tags and gain insights into the user lifecycle through these tags. At the same time, export target segment packages according to user stages for refined operations.

8. The marketing strategy optimization method based on user segmentation according to claim 1, characterized in that, Based on the target segmentation packages corresponding to the users, differentiated operational strategies are formulated for groups with different value stratifications. Users are reached through multiple channels, and performance data is collected. The performance data is used to guide the optimization of the strategies, including: Develop operational strategies based on the segmentation attributes and behavioral prediction results of the user's corresponding target segmentation package; Integrate with the CRM system to obtain a comprehensive customer profile and execute marketing strategies based on that profile. By building churn prediction models and gaining insights into churned users, we can execute personalized operational strategies to maximize user lifecycle extension.

9. The marketing strategy optimization method based on user segmentation according to claim 1, characterized in that, The marketing strategy optimization method based on user segmentation also... include: The data transmission operation parameters include the standard deviation of data retrieval response time per unit time, the ratio between processes and available CPUs per unit time, the API request frequency per unit time, and the security event occurrence rate per unit time; and the value of the unit time ranges from 3 min to 8 min.

10. A marketing strategy optimization processing system based on user segmentation for executing the marketing strategy optimization processing method of any one of claims 1-9, characterized in that, The marketing strategy optimization processing system based on user segmentation includes: The ID graph construction module is used to connect relevant business data of users and user vehicles from multiple business systems to the customer service operation platform and build an ID graph corresponding to the user. The tag building module is used to build tags corresponding to users using relevant business data about users and their vehicles. The target segmentation package acquisition module is used to segment users according to the tags corresponding to the users, acquire the target segmentation package corresponding to the users, and display the data information of the users and their upstream and downstream dependencies according to the target segmentation package. The multi-dimensional profile acquisition module is used to acquire multi-dimensional profiles of the target group corresponding to the target segment package based on the user's corresponding tags, through in-depth analysis and insight. The operation strategy acquisition and optimization module is used to formulate differentiated operation strategies for different value stratification groups based on the target group packages corresponding to the users, reach users through multiple channels and collect performance data, and use the performance data to guide the optimization of the strategy.

Images