Customer relationship management method and system based on big data

Abstract

The invention discloses a customer relationship management method and system based on big data, and the method comprises the following steps: S1, data collection and integration, S2, data preprocessing, S3, customer portrait construction, S4, customer value evaluation, S5, personalized marketing strategy making, and S6, customer relationship dynamic monitoring and optimization. The customer relationship management system based on the big data comprises a data acquisition interface, a data preprocessing unit, an analysis and processing engine, a strategy execution module and a user interaction interface. Multi-channel, massive and rich data can be integrated through customer relationship management of big data, a detailed and comprehensive customer portrait is constructed through an advanced algorithm, customer values are scientifically evaluated, behavior characteristics, preferences and potential demands of different customer groups are deeply informed, and the customer experience is improved. And personalized marketing strategies and service experiences are customized for all levels of customers.

Description

Technical Field

[0001] The present invention relates to the field of big data technology, and in particular to a customer relationship management method and system based on big data. Background Art

[0002] In today's highly competitive business environment, companies have accumulated a massive amount of customer data, including basic customer information, purchase history, interaction records, etc. Traditional customer relationship management methods often find it difficult to effectively integrate and deeply analyze these large-scale and diverse data, resulting in an inability to accurately grasp customer needs, predict customer behavior, and optimize customer service experience. Therefore, there is an urgent need for a customer relationship management solution that leverages the powerful analytical capabilities of big data.

[0003] Based on this, the present invention proposes a customer relationship management method and system based on big data. Summary of the invention

[0004] The purpose of the present invention is to solve the shortcomings of the prior art and to propose a customer relationship management method and system based on big data.

[0005] In order to achieve the above object, the present invention adopts the following technical solutions:

[0006] A customer relationship management method and system based on big data, comprising the following steps:

[0007] S1. Data collection and integration: Collect customer-related data from multiple data sources, including but not limited to online business platforms, offline business locations, customer service systems, and social media platforms; online business platforms are used to obtain customer registration information, browsing history, and transaction order information; offline business locations collect customer on-site transaction records and store-related behavior data; customer service systems collect customer consultation, feedback, and complaint data; and social media platforms obtain customer interaction information;

[0008] S2. Data preprocessing: Clean the collected customer-related data, remove duplicate, erroneous and incomplete data records, and convert the data in different formats into a format suitable for subsequent analysis and processing, and then store the processed data in a big data storage warehouse;

[0009] S3. Customer portrait construction: Extract key features for building customer portraits from the data stored in the big data storage warehouse. The key features include demographic features, consumption behavior features, hobby features, and loyalty features. Use clustering algorithms to cluster customers to divide them into different customer segments.

[0010] S4. Customer value assessment: The analytic hierarchy process is used to determine the weights of various factors that affect customer value. The factors include customer current value, potential value, loyalty, etc. A hierarchical model is constructed, and a judgment matrix is ​​constructed using expert scoring. The maximum eigenvalue of the judgment matrix and its corresponding eigenvector are calculated. After normalization, the weight vector of each factor is obtained, and then the customer life cycle value is combined to comprehensively evaluate the customer value.

[0011] S5. Personalized marketing strategy formulation: Based on the customer portrait and customer value assessment results, personalized marketing strategies are formulated for customer groups at different value levels. For high-value customer groups, a collaborative filtering recommendation algorithm is used to recommend products. The collaborative filtering recommendation algorithm includes user-based collaborative filtering or item-based collaborative filtering.

[0012] S6. Dynamic monitoring and optimization of customer relationships: Collect customer feedback data and new interactive behavior data in real time or regularly, and re-enter them into the system using big data stream processing technology to update customer portraits, re-evaluate customer value, and adjust personalized marketing strategies based on actual conditions to ensure that the company can respond to changes in customer needs and market dynamics in a timely manner and continuously optimize customer relationship management results.

[0013] Preferably, the online business platform is an e-commerce platform, an online financial service platform or an online platform of a telecommunications operator, which collects e-commerce business-related data, financial business-related data or telecommunications business-related data accordingly; the offline business venue is a retail store, an offline branch of a financial institution or a business hall of a telecommunications operator, which collects offline retail transaction data, offline financial business processing data or offline telecommunications business processing data respectively.

[0014] Preferably, the demographic characteristics include age, gender, and region; the consumer behavior characteristics include purchase frequency, purchase amount, and purchase category preference; the interest and hobby characteristics are obtained by analyzing the customer's interactive behavior and browsing history on social media; and the loyalty characteristics are measured by repeat purchase rate and customer complaint rate.

[0015] Preferably, in the hierarchical structure model constructed by the analytic hierarchy process, the target layer is customer value assessment, the factors affecting customer value included in the criterion layer can be adjusted and increased or decreased according to the business characteristics of different industries, and the solution layer is specific individual customers.

[0016] Preferably, in the collaborative filtering recommendation algorithm adopted for high-value customer groups, item-based collaborative filtering is performed by calculating the similarity between items and making recommendations based on similar items that the target customers have purchased. The exclusive services provided to high-value customer groups also include customized product recommendations, priority participation in high-end membership activities, etc.; the promotional activities, point rewards, product upgrade guidance and other strategies provided for medium-value customer groups can be specifically designed and adjusted according to the characteristics of products or services in different industries; service optimization and product recommendations for low-value customer groups are also adapted according to the specific industry business content.

[0017] Preferably, the big data stream processing technology adopts Apache Flink or other technical frameworks with real-time stream processing capabilities, and dynamically optimizes customer relationship management by collecting actual feedback from customers after receiving marketing strategies, such as marketing email open rates, SMS reply rates, purchase behavior changes and other data.

[0018] A customer relationship management system based on big data, comprising:

[0019] Data collection interface: responsible for connecting with various external data sources, collecting the customer-related data described in claim 1 through technical means such as application programming interface and web crawler, and transmitting the data to the data preprocessing unit;

[0020] Data preprocessing unit: performs operations such as cleaning, denoising, and format conversion on the collected data, removes duplicate, erroneous, and incomplete data records, converts data in different formats into a format suitable for subsequent analysis and processing, and then stores the processed data in a big data storage warehouse;

[0021] Analysis and processing engine: It has built-in data mining and analysis algorithms such as the K-Means clustering algorithm, hierarchical analysis method, and collaborative filtering recommendation algorithm as described in claim 1, reads the integrated data from the big data storage warehouse, performs operations such as customer portrait construction, customer value assessment, and personalized marketing strategy generation, and outputs corresponding analysis results and decision recommendations to the strategy execution module;

[0022] Strategy execution module: Based on the personalized marketing strategy output by the analysis and processing engine, the corresponding marketing content is accurately pushed to different customer groups through different channels, including but not limited to the company's marketing automation platform, SMS mass messaging system, email marketing system, etc., to implement the marketing strategy. At the same time, feedback data during the execution process is collected and fed back to the analysis and processing engine to further optimize the strategy and customer relationship management effect;

[0023] User interaction interface: Provides a visual operation interface for the company's marketing personnel, customer service personnel and other relevant personnel, allowing them to view customer portraits, customer value assessment reports, marketing strategy implementation status and other information, and to perform some necessary parameter configurations.

[0024] Preferably, the big data storage warehouse uses the Hadoop distributed file system or other storage systems with large-scale data storage and management capabilities to store massive amounts of pre-processed customer-related data.

[0025] Preferably, when executing customer portrait construction, the analysis and processing engine can flexibly select and adjust the key features used for clustering according to the business needs of different industries; when executing customer value assessment, it can modify and improve the hierarchical model and the weight of each factor in the hierarchical analysis method according to actual conditions; when executing personalized marketing strategy generation, it can adapt to the marketing channels and specific marketing content forms of different industries and enterprises.

[0026] Preferably, when collecting feedback data, the strategy execution module can not only collect direct feedback data after the marketing content is pushed, but also integrate data on customer response to marketing strategies from other customer service related channels, so as to more comprehensively optimize the strategy and customer relationship management effects.

[0027] The present invention has the following beneficial effects:

[0028] 1. Achieve accurate customer insights and personalized services. Customer relationship management based on big data can integrate multi-channel, massive and rich data, build detailed and comprehensive customer portraits and scientifically evaluate customer value through advanced algorithms, and gain in-depth insights into the behavioral characteristics, preferences and potential needs of different customer groups, thereby tailoring personalized marketing strategies and service experiences for customers at all levels. This is in sharp contrast to the rough classification and general service model of ordinary management methods, greatly improving the accuracy of services and customer acceptance, and enhancing customer satisfaction and loyalty to the company.

[0029] 2. Help enterprises optimize resource allocation and business expansion. With the help of a scientific and reasonable analysis and evaluation system, the management method based on big data can accurately distinguish high, medium and low value customer groups, so that enterprises can reasonably allocate marketing, service and other resources according to customer value to avoid resource waste. At the same time, it can effectively tap potential valuable customers, plan business expansion in advance, guide customers to convert to higher value, optimize business income structure, and promote sustainable development of enterprises. However, ordinary management methods lack systematic value evaluation and resource allocation mechanisms, which are easy to miss business expansion opportunities and affect the long-term benefits of enterprises.

[0030] 3. Enhance the dynamic response capability and market competitiveness of enterprises. By collecting and analyzing customer feedback and behavior data in real time or regularly, and using big data stream processing technology to achieve dynamic monitoring and optimization of all aspects of customer relationship management, enterprises can quickly respond to changes in customer demand, market fluctuations, and service feedback, adjust marketing strategies in a timely manner, optimize product and service content, maintain close fit with customer needs, and enhance the company's adaptability and competitive advantage in the market. Ordinary management methods are often slow to respond to changes due to information lags and lack of dynamic adjustment mechanisms, which can easily lead to customer loss and make it difficult to gain a foothold in the fiercely competitive market. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 A module step diagram of a customer relationship management system based on big data proposed by the present invention;

[0032] Figure 2 This is a partial code display diagram of the K-Means clustering algorithm in the first embodiment of the customer relationship management method based on big data proposed by the present invention;

[0033] Figure 3 This is a partial code display diagram of the hierarchical analysis method in the first embodiment of the customer relationship management method based on big data proposed by the present invention;

[0034] Figure 4 This is a partial code display diagram of the collaborative filtering recommendation algorithm in the first embodiment of the customer relationship management method based on big data proposed by the present invention;

[0035] Figure 5 This is a partial code display diagram of the K-Means clustering algorithm in the second embodiment of the customer relationship management method based on big data proposed by the present invention;

[0036] Figure 6 This is a partial code display diagram of the hierarchical analysis method in the second embodiment of the customer relationship management method based on big data proposed by the present invention;

[0037] Figure 7 This is a partial code display diagram of the variance model example code in Embodiment 2 of the customer relationship management method based on big data proposed by the present invention;

[0038] Figure 8 This is a partial code display diagram of the K-Means clustering algorithm in Example 3 of a customer relationship management method based on big data proposed by the present invention;

[0039] Fig. 9 This is a partial code display diagram of the hierarchical analysis method in the third embodiment of the customer relationship management method based on big data proposed by the present invention;

[0040] Fig.10This is a partial code display diagram of the recommended package price increase plan in Example 3 of the customer relationship management method based on big data proposed by the present invention. DETAILED DESCRIPTION

[0041] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0042] A customer relationship management method based on big data, comprising the following steps:

[0043] S1. Data collection and integration: Collect customer-related data from multiple data sources, including but not limited to online business platforms, offline business locations, customer service systems, and social media platforms; online business platforms are used to obtain customer registration information, browsing history, and transaction order information; offline business locations collect customer on-site transaction records and store-related behavior data; customer service systems collect customer consultation, feedback, and complaint data; and social media platforms obtain customer interaction information;

[0044] S2. Data preprocessing: Clean the collected customer-related data, remove duplicate, erroneous and incomplete data records, and convert the data in different formats into a format suitable for subsequent analysis and processing, and then store the processed data in a big data storage warehouse;

[0045] S3. Customer portrait construction: Extract key features for building customer portraits from the data stored in the big data storage warehouse. The key features include demographic features, consumption behavior features, hobby features, and loyalty features. Use clustering algorithms to cluster customers to divide them into different customer segments.

[0046] S4. Customer value assessment: The analytic hierarchy process is used to determine the weights of various factors that affect customer value. The factors include customer current value, potential value, loyalty, etc. A hierarchical model is constructed, and a judgment matrix is ​​constructed using expert scoring. The maximum eigenvalue of the judgment matrix and its corresponding eigenvector are calculated. After normalization, the weight vector of each factor is obtained, and then the customer life cycle value is combined to comprehensively evaluate the customer value.

[0047] S5. Personalized marketing strategy formulation: Based on the customer portrait and customer value assessment results, personalized marketing strategies are formulated for customer groups at different value levels. For high-value customer groups, a collaborative filtering recommendation algorithm is used to recommend products. The collaborative filtering recommendation algorithm includes user-based collaborative filtering or item-based collaborative filtering.

[0048] S6. Dynamic monitoring and optimization of customer relationships: Collect customer feedback data and new interactive behavior data in real time or regularly, and re-enter them into the system using big data stream processing technology to update customer portraits, re-evaluate customer value, and adjust personalized marketing strategies based on actual conditions to ensure that the company can respond to changes in customer needs and market dynamics in a timely manner and continuously optimize customer relationship management results.

[0049] The online business platform is an e-commerce platform, an online financial service platform or an online platform of a telecommunications operator, which collects e-commerce business-related data, financial business-related data or telecommunications business-related data respectively; the offline business venue is a retail store, an offline outlet of a financial institution or a business hall of a telecommunications operator, which collects offline retail transaction data, offline financial business processing data or offline telecommunications business processing data respectively.

[0050] The demographic characteristics include age, gender, and region; the consumer behavior characteristics include purchase frequency, purchase amount, and purchase category preference; the interest and hobby characteristics are obtained by analyzing the customer's interactive behavior and browsing history on social media; the loyalty characteristics are measured by repeat purchase rate and customer complaint rate.

[0051] In the hierarchical structure model constructed by the analytic hierarchy process, the target layer is customer value assessment, the factors affecting customer value included in the criterion layer can be adjusted and increased or decreased according to the business characteristics of different industries, and the solution layer is the specific individual customers.

[0052] In the collaborative filtering recommendation algorithm used for high-value customer groups, item-based collaborative filtering calculates the similarity between items and makes recommendations based on similar items that the target customers have already purchased. Exclusive services provided to high-value customer groups also include customized product recommendations and priority participation in high-end membership activities. Strategies such as promotional activities, point rewards, and product upgrade guidance provided to medium-value customer groups can be specifically designed and adjusted according to the characteristics of products or services in different industries. Service optimization and product recommendations for low-value customer groups are also adapted based on the specific industry business content.

[0053] The big data stream processing technology adopts Apache Flink or other technical frameworks with real-time stream processing capabilities. By collecting actual feedback from customers after receiving marketing strategies, such as marketing email opening rate, SMS reply rate, purchase behavior changes and other data, customer relationship management is dynamically optimized.

[0054] A customer relationship management system based on big data, comprising:

[0055] Data collection interface: responsible for connecting with various external data sources, collecting the customer-related data described in claim 1 through technical means such as application programming interface and web crawler, and transmitting the data to the data preprocessing unit;

[0056] Data preprocessing unit: performs operations such as cleaning, denoising, and format conversion on the collected data, removes duplicate, erroneous, and incomplete data records, converts data in different formats into a format suitable for subsequent analysis and processing, and then stores the processed data in a big data storage warehouse;

[0057] Analysis and processing engine: It has built-in data mining and analysis algorithms such as the K-Means clustering algorithm, hierarchical analysis method, and collaborative filtering recommendation algorithm as described in claim 1, reads the integrated data from the big data storage warehouse, performs operations such as customer portrait construction, customer value assessment, and personalized marketing strategy generation, and outputs corresponding analysis results and decision recommendations to the strategy execution module;

[0058] Strategy execution module: Based on the personalized marketing strategy output by the analysis and processing engine, the corresponding marketing content is accurately pushed to different customer groups through different channels, including but not limited to the company's marketing automation platform, SMS mass messaging system, email marketing system, etc., to implement the marketing strategy. At the same time, feedback data during the execution process is collected and fed back to the analysis and processing engine to further optimize the strategy and customer relationship management effect;

[0059] User interaction interface: Provides a visual operation interface for the company's marketing personnel, customer service personnel and other relevant personnel, allowing them to view customer portraits, customer value assessment reports, marketing strategy implementation status and other information, and to perform some necessary parameter configurations.

[0060] The big data storage warehouse uses the Hadoop distributed file system or other storage systems with large-scale data storage and management capabilities to store massive amounts of pre-processed customer-related data.

[0061] When executing customer portrait construction, the analysis and processing engine can flexibly select and adjust the key features used for clustering according to the business needs of different industries; when executing customer value assessment, it can modify and improve the hierarchical model and the weight of each factor in the hierarchical analysis method according to actual conditions; when executing personalized marketing strategy generation, it can adapt to the marketing channels and specific marketing content forms of different industries and enterprises.

[0062] When collecting feedback data, the strategy execution module can not only collect direct feedback data after the marketing content is pushed, but also integrate data on customer response to marketing strategies from other customer service related channels, so as to more comprehensively optimize the strategy and customer relationship management effects.

[0063] Example 1: Customer relationship management application for retail enterprises;

[0064] Step 1: Enterprise background and data collection;

[0065] The online platform uses the API interface to obtain customer registration information (including name, age, gender, contact information, etc.), browsing history (browsing product categories, duration, frequency, etc.), purchase order information (purchased product details, amount, purchase time, etc.) and customer evaluation and scoring data from the e-commerce website. In offline stores, the point of sale system (POS) is used to collect the purchase transaction records of customers who come to the store (including purchased products, payment amount, purchase time, store location, etc.). At the same time, the Wi-Fi hotspots set up in the store are used to collect relevant data such as the connection time and frequency of customers' visits to the store (which can assist in analyzing customers' store habits). The collected structured and unstructured data are transmitted to the data preprocessing unit for integration and preprocessing.

[0066] Step 2: Data preprocessing and storage;

[0067] Data cleaning: remove duplicate transaction records (for example, the same online order is recorded repeatedly due to system failure), correct obviously erroneous data (such as age that is not filled in with a common sense value), and fill in some missing values ​​(through reasonable default values ​​or estimated values ​​based on other relevant data, such as inferring gender based on the customer's purchase history, etc.).

[0068] Format conversion: Convert data in different formats into a structured data format stored in columns, so that it can be stored in a big data storage warehouse (using Hadoop distributed file system HDFS). For example, parse the browsing history text on a web page to extract key information and convert it into standard table data.

[0069] Step 3: Build customer portrait;

[0070] Feature selection: select age, gender, purchase frequency in the past year, average purchase amount per time, and concentration of purchased product categories (measured by calculating the Herfindahl index, the formula is where x i is the amount of money that customers spend on the i-th category of goods, X is the total purchase amount, and n is the total number of product categories. The higher the value, the more concentrated the categories are) as clustering features.

[0071] Clustering execution: Set the number of clusters K = 5 (can be adjusted based on business experience and multiple experiments), randomly initialize 5 cluster centers, and for each customer data point x i , calculate the Euclidean distance between it and each cluster center, the calculation formula is:

[0072]

[0073] Here m is the number of features, such as m=5 corresponds to the 5 features selected above, and then the customer is assigned to the cluster with the nearest cluster center.

[0074] Recalculate the new cluster centers of each cluster (formula ), and continuously repeat the steps of allocating and updating cluster centers until the change of cluster centers is less than the set threshold (such as 0.001). Finally, 5 different customer portrait groups are obtained, such as "young high-frequency, low-amount, diversified purchasing group" and "middle-aged medium-frequency, high-amount, concentrated purchasing group".

[0075] Step 4: Customer value assessment;

[0076] Determine weights using the analytic hierarchy process: Build a hierarchical model, with the target layer being customer value assessment, and the criterion layer including four factors: current purchase amount contribution (total purchase amount in the past year), purchase frequency (number of purchases in the past year), potential purchasing power (based on inferences from browsing unpurchased products), and loyalty (measured by repeat purchase rate). Build a judgment matrix by inviting internal sales experts and market experts to compare and score two factors (using a 1-9 scale). For example, if the current purchase amount contribution is considered to be "slightly more important" than the purchase frequency, assign a value of 3 to the corresponding judgment matrix element, calculate the maximum eigenvalue of the judgment matrix and its corresponding eigenvector, and obtain the weight vector of each factor after normalization ω=[ω 1 ,ω 2 ,ω 3 ,ω 4 ].

[0077] Calculation of customer lifetime value: Assuming that the expected customer lifetime is T = 5 years, for customer i, its annual marginal revenue is M t (estimated based on commodity profit and purchase quantity), service cost C t The customer life cycle value is calculated based on the following formula:

[0078] Comprehensive value score calculation: The value score formula for customer i is:

[0079] V i =ω 1 ×f i1 +ω2 ×f i2 +ω 3 ×f i3 +ω 4 ×f i4

[0080] where f i1 The quantitative value of customer i's current purchase amount is contributed, and so on. Customers are divided into high-value, medium-value, and low-value customer groups based on their value scores.

[0081] Step 5: Personalized marketing strategy formulation and implementation;

[0082] High-value customer groups: Provide them with exclusive offline shopping experience event invitations (such as new product tasting sessions in advance), and push customized product recommendation lists online (through the collaborative filtering recommendation algorithm based on items, calculate the similarity between items, the formula is:

[0083]

[0084] Where i, j are different commodities, U is the user set, r ui The system recommends products based on the rating of product i by user u and the choices of other high-value customers who have purchased products similar to the target customer's preferences), and gives additional points rewards and priority delivery services, etc.

[0085] Medium-value customer group: Regularly send discount coupons and category discount coupons, recommend related products, and mine based on association rules, and calculate support and confidence at the same time. The formula for calculating support is:

[0086]

[0087] The formula for calculating confidence is: When the support and confidence meet a certain threshold, product Y is recommended to customers who have purchased product X, encouraging them to increase purchase frequency and amount and convert them into high-value customers.

[0088] Low-value customer groups: Improve their shopping experience by optimizing online shopping interface guidance, sending basic popular product recommendations, etc., while controlling marketing cost investment and observing whether their subsequent behavior tends to convert to higher value.

[0089] Step 6: Dynamic monitoring and optimization;

[0090] Through customer evaluation feedback after shopping, repurchase behavior data and real-time interaction data of online platforms, Apache Flink real-time stream processing technology is used to timely update to the big data storage warehouse, re-analyze customer portraits and value, and adjust corresponding marketing strategies. For example, if it is found that the purchase frequency of a low-value customer has increased significantly recently, after re-evaluating its value, it will be included in the marketing strategy scope of the medium-value customer group.

[0091] Embodiment 2: Customer relationship management application for financial service institutions;

[0092] Step 1: Data collection;

[0093] Obtain basic customer account information (name, age, occupation, account opening time, etc.), deposit business data (deposit balance of each account, deposit term, interest rate, etc.), loan business data (loan amount, loan term, repayment status, etc.) and credit card usage data (consumption amount, repayment record, credit limit, etc.) from the core banking system, collect customer operation records through online banking and mobile banking (login frequency, operation function usage, financial product pages browsed, etc.), as well as problem records of online customer service (reflecting customer focus and potential needs), record the type and frequency of business handled by customers at the branch, and the summary of the content of the interview with the account manager, etc.

[0094] Step 2: Data preprocessing and storage;

[0095] Remove invalid account operation records (such as abnormal records generated by system testing), check the consistency of customer information (such as the unification of key information such as customer age in different business systems), supplement missing occupation and other information (through third-party data verification or reasonable speculation based on other customer behaviors), and organize data in various formats into a unified structured format. For example, extract key topic tags from customer service consultation text content and merge them with other data and store them in a big data storage warehouse (based on Hadoop's HDFS) for subsequent analysis.

[0096] Step 3: Build customer portrait;

[0097] We select customer age, career stability (different levels can be set through comprehensive evaluation of industry attributes, years of work, etc.), asset size (calculated based on comprehensive deposits, investments, etc.), credit rating (assessed based on past loan repayment records and credit card usage), and diversity of financial product usage (the number of different types of financial products used, such as deposits, loans, and financial management) as clustering features, set K=4 (the number of customer groups is divided according to the characteristics and experience of financial services), and initialize 4 cluster centers.

[0098] For each customer data point, the distance to each cluster center is calculated according to the Euclidean distance formula (the same as the distance calculation formula in the above retail enterprise example), and the customer is assigned to the nearest cluster. Then the cluster center is updated, and it is iterated repeatedly until convergence (the cluster center changes very little, such as less than 0.0005), to obtain different customer portraits such as "young high-asset and high-credit diversified financial product users" and "middle-aged medium-asset, stable-credit, traditional business preference groups".

[0099] Step 4: Customer value assessment;

[0100] The estimated customer life cycle is T = 10 years (considering the long-term relationship characteristics of financial services). For customer j, the annual marginal revenue is M. t (including interest income, commission income, etc.), service cost C t (operating costs, risk management costs, etc.) are different, and the discount rate r = 0.08, then the customer life cycle value is calculated as:

[0101] Customer j’s value score V j =ω 1 ×f j1 +ω 2 ×f j2 +ω 3 ×f j3 +ω 4 ×f j4 , where f j1 Etc. are the quantitative values ​​corresponding to each factor, and customers are divided into high, medium and low value customers based on them.

[0102] Step 5: Personalized marketing strategy formulation and implementation;

[0103] High-value user groups: They are assigned exclusive financial advisors and provided with customized high-end financial solutions (based on analysis of customer risk preferences, asset size, etc., and determining the best asset allocation through portfolio optimization algorithms, such as the Markowitz mean-variance model, etc.). They are given priority in recommending limited edition high-yield financial products and are invited to participate in exclusive activities such as high-end financial forums to enhance customer stickiness and asset value-added services.

[0104] Medium-value customer groups: Regularly push recommendations for financial products that are suitable for their risk tolerance (determine the customer's risk level through a risk assessment model, and make recommendations based on products of corresponding levels on the market), provide fee discounts (such as transfer fee reductions, credit card annual fee discounts, etc.), and encourage them to increase their asset size and business scope and convert to high-value customers.

[0105] Low-value customer groups: send basic financial knowledge materials to guide them in the correct use of common financial services (such as online banking operation guides). At the same time, cultivate their habit of using banking services through some small incentives (such as cash back on consumption, points exchange for small gifts, etc.) to observe whether there is potential for value improvement.

[0106] Step 6: Dynamic monitoring and optimization;

[0107] Use the bank's real-time transaction monitoring system and customer feedback channels to collect new data, and use big data stream processing technology to update customer portraits and value assessments in real time. For example, if it is found that a medium-value customer's asset size has increased significantly after investing in a recommended wealth management product, adjust its value assessment in a timely manner, provide it with a higher level of service and product recommendations, and continuously optimize customer relationship management results.

[0108] Example 3: Telecom operator customer relationship management application;

[0109] Step 1: Data collection;

[0110] The system obtains data such as customer call duration, call charges, SMS usage, and traffic usage from the billing system; extracts information such as customer package processing status, contract term, and complaint records from the customer relationship management system (CRM); collects customer login status through the operator's official website and mobile business hall, business inquiry records, participation in online activities, and browsing and usage data of various value-added services (such as video ringtones, cloud disks, etc.); records the details of customers' business transactions at the business hall (such as package changes, new business activation, etc.), on-site consultation issues, and feedback from communication with staff, etc.; and summarizes these data and transmits them to the data preprocessing unit.

[0111] Step 2: Data preprocessing and storage;

[0112] Remove abnormal billing records caused by system failures, clean up duplicate business processing records, verify the accuracy of customer package information, etc., to ensure data quality, and convert data from different sources into a structured format that is easy to analyze. For example, perform natural language processing on the text content of on-site consultations in business halls, extract key topics (such as network signal problems, fee questions, etc.), and then integrate them with other business data and store them in a big data storage warehouse (using a suitable distributed storage system, such as HDFS).

[0113] Step 3: Build customer portrait;

[0114] Select customer age, package type (divided into voice-based, data-based, comprehensive packages, etc.), average monthly call duration, average monthly data usage, and type of value-added service usage (such as how many value-added services are used) as clustering features, set the number of clusters K = 3 (determined according to the main customer classification needs of telecommunications services), and initialize 3 cluster centers.

[0115] For each customer’s data point, the distance to each cluster center is calculated according to the Euclidean distance formula (such as the distance calculation formula in the previous example), and the customer is assigned to the cluster with the nearest distance. The cluster center is then updated, and this process is repeated until the change in the cluster center is less than the set threshold (such as 0.001). This results in different customer portraits, such as “young people with high traffic demand and active value-added services” and “middle-aged people with stable packages that mainly use voice calls”.

[0116] Step 4: Customer value assessment;

[0117] Determine weights through hierarchical analysis: Build a hierarchical model, with the target layer being customer value assessment, and the criterion layer including factors such as current consumption amount contribution (average monthly expenditure), service usage stability (reflected in low frequency of package changes, good contract performance, etc.), potential business needs (based on speculation based on browsing without value-added services, etc.), and loyalty (measured by online time).

[0118] By inviting market research experts and business operation experts within the operator to compare and score two factors, a judgment matrix is ​​constructed, and the weight vector of each factor is calculated as follows:

[0119] ω=[ω 1 ,ω 2 ,ω 3 ,ω 4 ]

[0120] Calculation of customer lifetime value: Estimated customer lifetime T = 3 years (considering the relatively short replacement cycle of telecom industry customers), for customer k, annual marginal revenue M t (estimated based on package profits, value-added service revenue, etc.), service cost C t (network maintenance costs, customer service costs, etc. are allocated to customers) and the discount rate is r = 0.12, then the customer lifetime value calculation formula is:

[0121] Comprehensive value score calculation: The value of customer k is divided into:

[0122] V k =ω 1 ×f k1 +ω 2 ×f k2 +ω 3 ×fk3 +ω 4 ×f k4

[0123] where f k1 etc. are the quantitative values ​​corresponding to each factor, which are used to distinguish high, medium and low value customers.

[0124] Step 5: Personalized marketing strategy formulation and implementation;

[0125] High-value customer groups: Provide exclusive high-speed network experience (such as improving network priority, increasing free data traffic, etc.), give priority to recommending the latest high-end value-added services (such as high-definition video calls, ultra-large cloud disk space, etc.), invite them to participate in high-end membership activities held by operators (such as celebrity meet-and-greets, technology experience activities, etc.), and provide benefits such as call fee discounts and doubling points to enhance customer stickiness.

[0126] Medium-value customer group: recommend package upgrade plans that suit their usage habits (determine better packages by analyzing their traffic and voice usage trends), send trial opportunities for value-added services (such as one month of free music membership, etc.), and hold targeted online and offline promotional activities (such as recharge and return of call fees, business application lucky draws, etc.) to promote their conversion to high-value customers.

[0127] Low-value customer groups: Provide basic network usage optimization.

[0128] From Example 1, Example 2, and Example 3, it can be seen that the specific implementation methods of different industries differ in terms of data characteristics of each link, focus, and implementation effects of marketing strategies, but they are all based on big data to achieve more refined and personalized management of customer relationships, in order to meet the business characteristics of their respective industries to enhance customer value and corporate operating efficiency.

[0129] In addition, customer relationship management based on big data has shown significant advantages in different fields such as retail enterprises, financial service institutions and telecom operators. It integrates massive, multi-channel data, uses a variety of data analysis methods such as clustering algorithms and hierarchical analysis methods, deeply analyzes customer characteristics, accurately builds customer portraits and scientifically evaluates customer value, and thus realizes customized precision marketing and services. Whether it is retail companies recommending suitable products for different consumer groups, financial institutions creating exclusive financial plans for customers, or telecom operators providing personalized services based on customer communication habits, they have effectively improved customer experience and deeply excavated customer value. At the same time, this management method helps companies to reasonably allocate resources based on customer value, avoid waste of resources, optimize operational efficiency and benefits, and comprehensively enhance the market competitiveness of enterprises, laying a solid foundation for the sustainable development of enterprises.

[0130] Furthermore, customer relationship management based on big data has demonstrated its important value in the practice of multiple industries. Taking retail enterprises, financial service institutions and telecom operators as examples, this management model, with the help of the powerful data collection and analysis capabilities of big data, can not only fully understand the customer's behavioral preferences, potential needs and value levels, but also monitor customer feedback and market changes in real time and dynamically. This allows enterprises to carry out highly personalized marketing activities and provide precise services for different customer groups, such as retail enterprises optimizing product recommendations, financial institutions flexibly adjusting financial services, and telecom operators timely upgrading packages, thereby improving customer satisfaction and loyalty and deeply tapping customer value. Moreover, in terms of resource allocation, it is more scientific and reasonable. Enterprises can invest resources in a targeted manner according to customer value and improve operational efficiency. At the same time, with the ability to quickly respond to market changes, they can better adapt to the complex and changing market environment, helping enterprises stand out in industry competition and achieve long-term development.

[0131] It should be noted that, in the comparative embodiment, the comparative example adopts a common management method to manage various user groups, adopts a universal marketing strategy, sends the same coupons or product recommendations to all customers, ignores the personalized needs of customers, and customers may feel annoyed with these irrelevant marketing messages, resulting in poor marketing effectiveness. Its unified service model is difficult to take care of the different needs of various customer groups, which can easily lead to customer dissatisfaction and affect customer retention rate. In addition, due to the lack of personalized guidance, it is difficult for customers to actively improve their consumption value. The growth of corporate business revenue may reach a bottleneck, making it difficult to optimize and upgrade the revenue structure.

[0132] Table 1: Data comparison of various embodiments and comparative examples

[0133]

[0134]

[0135]

[0136] From the comparison of the above tables, it can be clearly seen that the specific implementation methods of customer relationship management based on big data have advantages and characteristics in all aspects compared with ordinary management methods, which is more helpful for enterprises to deeply explore customer value and maintain good customer relationships. Whether it is the retail, finance or telecommunications industry, customer relationship management based on big data has shown strong advantages. Through the beneficial effects of deeply exploring customer value, optimizing resource allocation and real-time dynamic response, it helps enterprises improve their core competitiveness, better meet customer needs, and promote the sustainable and healthy development of enterprises in their respective fields.

[0137] Specifically, in each embodiment, key code examples and corresponding analysis are written in Python language.

[0138] Furthermore, if Figure 2 , Figure 5 and Figure 8 As shown in the figure, the algorithm aims to divide customer groups into different categories based on similarities through the distribution of intrinsic characteristics of the data, so as to build more targeted and representative customer portraits, help enterprises to deeply understand the characteristics and behavior patterns of different customer groups, and provide a basis for the subsequent accurate marketing strategy formulation. It can be applied in different scenarios of retail enterprises, financial service institutions, and telecom operators, but the specific features selected in each scenario are different according to the characteristics of the industry. For example, retail focuses on characteristics related to purchasing behavior, finance focuses on assets and credit, and telecommunications focuses on packages and service usage. This reflects the powerful function of the algorithm to flexibly adapt to different business fields to explore customer segment groups. The advantage is that its principle is relatively simple and easy to understand, and it is relatively intuitive to implement. It can quickly perform cluster analysis on large-scale data to obtain different customer segmentation results.

[0139] Furthermore, if Figure 3 , Figure 6 and Fig. 9 As shown in the figure, by constructing a hierarchical structure, we systematically sort out the multiple factors that affect customer value, and determine the weight of the relative importance of each factor based on expert experience and rigorous mathematical calculations, so that customer value assessment no longer relies on subjective and arbitrary judgment, but has a scientific and reasonable quantitative basis, which improves the accuracy and objectivity of value assessment. It is also applicable to the comprehensive consideration of customer value in different industry scenarios. Each industry selects different influencing factors according to its own business focus and puts them into the hierarchical structure. For example, retail focuses on purchase-related factors, and finance focuses on assets and business expansion factors. This fits the focus of various industries on customer value judgment. The advantage is that complex multi-factor decision-making problems can be organized and hierarchical, and more reliable weight results can be obtained by integrating the opinions of multiple experts. Moreover, the calculation process has clear mathematical logic support.

[0140] Furthermore, if Figure 4As shown in the figure, for high-value customer groups, personalized product recommendations are made based on the similarities between customers or items, thereby improving the accuracy of recommendations, meeting customers' personalized needs, and enhancing customers' acceptance of recommended content, thereby improving customer satisfaction and consumption likelihood, and helping companies optimize marketing effectiveness. This is very practical in scenarios such as retail where there are many products to recommend and where it is necessary to explore customers' personalized preferences. However, different industries need to make adjustments based on the characteristics of their own products or services. For example, the financial industry may be used to recommend financial products with similar risk-return characteristics, and the telecommunications industry may be used to recommend value-added services that suit customer usage habits. It has a certain degree of versatility, but it must be customized in combination with specific businesses when applied. The advantage is that it can make full use of existing customer behavior data (such as ratings, purchase records, etc.) to achieve personalized recommendations, without the need for in-depth analysis of the item content itself, and when the data volume is large enough and the customer base is rich, the recommendation effect is often better.

[0141] Furthermore, if Figure 7 As shown in the figure, it helps financial institutions to meet the asset allocation needs of high-value customers. On the basis of considering the balance between return and risk, it calculates the optimal asset weight combination through a scientific mathematical model, customizes financial management plans that meet their risk preferences for customers, realizes reasonable allocation and appreciation of assets, and enhances customers' trust and reliance on the professional services of financial institutions. It is mainly applicable to the field of financial investment, especially in planning diversified asset portfolios for customers. It determines the allocation plan based on the risk tolerance and return goals of different customers and the expected returns, covariance and other conditions of various assets. It is a commonly used and effective quantitative tool in the professional financial management services of financial institutions. Its advantage lies in the support of rigorous mathematical theory. It intuitively shows how to seek the optimal balance between risk and return from a quantitative perspective, provides a scientific basis for asset allocation decisions, and has significant application effects in mature financial markets and multi-asset environments.

[0142] Furthermore, if Fig.10 As shown, based on the actual daily usage of telecommunications customers, such as trends in traffic and voice usage, more suitable packages are recommended for medium-value customers, guiding customers to consume rationally and increase service usage, while improving customer satisfaction with packages, promoting customer conversion to high-value customers, and optimizing the business revenue structure of telecommunications operators. It is specifically suitable for the telecommunications operating industry, and formulates recommendation strategies based on its business characteristics and customer communication usage data. It has certain reference value for similar scenarios in other industries that need to upgrade business recommendations based on the actual situation of customers using products or services, but it needs to be adjusted and transformed according to the specific industry business logic. The advantage is that it is close to the actual customer experience and makes recommendations based on real data feedback, which is easy for customers to understand and accept, and it can respond to changes in customer usage habits in a timely manner and dynamically adjust the recommended content.

[0143] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A customer relationship management method based on big data, characterized in that: The following steps are involved: S1. Data collection and integration: Collect customer-related data from multiple data sources, including but not limited to online business platforms, offline business locations, customer service systems, and social media platforms; online business platforms are used to obtain customer registration information, browsing history, and transaction order information; offline business locations collect customer on-site transaction records and store-related behavior data; customer service systems collect customer consultation, feedback, and complaint data; and social media platforms obtain customer interaction information; S2. Data preprocessing: Clean the collected customer-related data, remove duplicate, erroneous and incomplete data records, and convert the data in different formats into a format suitable for subsequent analysis and processing, and then store the processed data in a big data storage warehouse; S3. Customer portrait construction: Extract key features for building customer portraits from the data stored in the big data storage warehouse. The key features include demographic features, consumption behavior features, hobby features, and loyalty features. Use clustering algorithms to cluster customers to divide them into different customer segments. S4. Customer value assessment: The analytic hierarchy process is used to determine the weights of various factors that affect customer value. The factors include customer current value, potential value, loyalty, etc. A hierarchical model is constructed, and a judgment matrix is ​​constructed using expert scoring. The maximum eigenvalue of the judgment matrix and its corresponding eigenvector are calculated. After normalization, the weight vector of each factor is obtained, and then the customer life cycle value is combined to comprehensively evaluate the customer value. S5. Personalized marketing strategy formulation: Based on the customer portrait and customer value assessment results, personalized marketing strategies are formulated for customer groups at different value levels. For high-value customer groups, a collaborative filtering recommendation algorithm is used to recommend products. The collaborative filtering recommendation algorithm includes user-based collaborative filtering or item-based collaborative filtering. S6. Dynamic monitoring and optimization of customer relationships: Collect customer feedback data and new interactive behavior data in real time or regularly, and re-enter them into the system using big data stream processing technology to update customer portraits, re-evaluate customer value, and adjust personalized marketing strategies based on actual conditions to ensure that the company can respond to changes in customer needs and market dynamics in a timely manner and continuously optimize customer relationship management results.

2. A customer relationship management method based on big data according to claim 1, characterized in that: In step S1, the online business platform is an e-commerce platform, an online financial service platform or an online platform of a telecommunications operator, which collects e-commerce business-related data, financial business-related data or telecommunications business-related data accordingly; the offline business venue is a retail store, an offline branch of a financial institution or a business hall of a telecommunications operator, which collects offline retail transaction data, offline financial business processing data or offline telecommunications business processing data respectively.

3. The customer relationship management method based on big data according to claim 1, characterized in that: In step S3, the demographic characteristics include age, gender, and region; the consumer behavior characteristics include purchase frequency, purchase amount, and purchase category preference; the interest and hobby characteristics are obtained by analyzing the customer's interactive behavior and browsing history on social media; and the loyalty characteristics are measured by repeat purchase rate and customer complaint rate.

4. The customer relationship management method based on big data according to claim 1, characterized in that: In step S4, in the hierarchical model constructed by the analytic hierarchy process, the target layer is customer value assessment, the factors affecting customer value included in the criterion layer can be adjusted and increased or decreased according to the business characteristics of different industries, and the solution layer is the specific individual customers.

5. The customer relationship management method based on big data according to claim 1, characterized in that: In step S5, in the collaborative filtering recommendation algorithm adopted for the high-value customer group, the item-based collaborative filtering is to make recommendations by calculating the similarity between items and referring to similar items that the target customer has purchased, and the exclusive services provided to the high-value customer group also include customized product recommendations, priority participation in high-end member activities, etc.; the promotional activities, point rewards, product upgrade guidance and other strategies provided for the medium-value customer group can be specifically designed and adjusted according to the characteristics of products or services in different industries; the service optimization and product recommendations for the low-value customer group are also adapted according to the specific industry business content.

6. The customer relationship management method based on big data according to claim 1, characterized in that: In step S6, the big data stream processing technology uses Apache Flink or other technical frameworks with real-time stream processing capabilities to dynamically optimize customer relationship management by collecting actual feedback from customers after receiving marketing strategies, such as marketing email open rates, SMS reply rates, purchase behavior changes and other data.

7. A customer relationship management system based on big data, characterized in that: include: Data collection interface: responsible for connecting with various external data sources, collecting the customer-related data described in claim 1 through technical means such as application programming interface and web crawler, and transmitting the data to the data preprocessing unit; Data preprocessing unit: performs operations such as cleaning, denoising, and format conversion on the collected data, removes duplicate, erroneous, and incomplete data records, converts data in different formats into a format suitable for subsequent analysis and processing, and then stores the processed data in a big data storage warehouse; Analysis and processing engine: It has built-in data mining and analysis algorithms such as the K-Means clustering algorithm, hierarchical analysis method, and collaborative filtering recommendation algorithm as described in claim 1, reads the integrated data from the big data storage warehouse, performs operations such as customer portrait construction, customer value assessment, and personalized marketing strategy generation, and outputs corresponding analysis results and decision recommendations to the strategy execution module; Strategy execution module: Based on the personalized marketing strategy output by the analysis and processing engine, the corresponding marketing content is accurately pushed to different customer groups through different channels, including but not limited to the company's marketing automation platform, SMS mass messaging system, email marketing system, etc., to implement the marketing strategy. At the same time, feedback data during the execution process is collected and fed back to the analysis and processing engine to further optimize the strategy and customer relationship management effect; User interaction interface: Provides a visual operation interface for the company's marketing personnel, customer service personnel and other relevant personnel, allowing them to view customer portraits, customer value assessment reports, marketing strategy implementation status and other information, and to perform some necessary parameter configurations.

8. A customer relationship management system based on big data according to claim 7, characterized in that: The big data storage warehouse uses the Hadoop distributed file system or other storage systems with large-scale data storage and management capabilities to store massive amounts of pre-processed customer-related data.

9. A customer relationship management system based on big data according to claim 7, characterized in that: When executing customer portrait construction, the analysis and processing engine can flexibly select and adjust the key features used for clustering according to the business needs of different industries; when executing customer value assessment, it can modify and improve the hierarchical model and the weight of each factor in the hierarchical analysis method according to actual conditions; when executing personalized marketing strategy generation, it can adapt to the marketing channels and specific marketing content forms of different industries and enterprises.

10. A customer relationship management system based on big data according to claim 7, characterized in that: When collecting feedback data, the strategy execution module can not only collect direct feedback data after the marketing content is pushed, but also integrate data on customer response to marketing strategies from other customer service related channels, so as to more comprehensively optimize the strategy and customer relationship management effects.

Images