A CRM-based information data analysis method and system

Abstract

The application relates to the technical field of customer relationship management, and discloses an information data analysis method and system based on CRM. The information data analysis method based on CRM comprises the following steps: S1, collecting user original information data based on CRM, and screening target users with loss risks according to the user original information data; S2, extracting customer loss risk features from the original information data of the target users, and inputting the customer loss risk features into a trained risk assessment calculation model to output a result indicating the loss risk grades of the target users; and S3, generating a retention scheme according to the loss risk grades of the target users and the corresponding original information data and outputting the retention scheme to the corresponding target users. The application classifies and intervenes the target users based on different risk grades, avoids adopting a unified retention strategy for all users, and improves the customer retention rate.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of customer relationship management, and particularly relates to a CRM-based information data analysis method. BACKGROUND

[0002] Customer churn refers to the behavior of existing customers of an enterprise stopping purchasing or using its products or services due to various reasons, which is usually manifested as contract termination, no longer renewing or turning to competitors. Churn may be triggered by factors such as price sensitivity, poor service experience, demand change or market competition, and directly affects the enterprise's income and market share, while increasing the cost of acquiring customers. High churn rate often exposes the shortcomings of the enterprise in customer satisfaction, product value or loyalty management. In view of the problem of customer churn, implementing reasonable retention measures is the key for the enterprise to reduce the churn rate and improve the lifetime value of customers.

[0003] The existing CRM system usually relies on static rules to judge the customer churn risk, and the intervention strategy lacks a grading mechanism, which cannot be processed differently for users with different churn risk levels, and a unified retention scheme (such as increasing the amount of advertisement push) often not only fails to retain customers, but also easily prompts customers to have a negative emotion, in addition, without combining the feedback of users to dynamically exclude customers with high churn tendency, it also leads to waste of marketing resources. SUMMARY

[0004] The present application aims to provide a CRM-based information data analysis method to solve the above technical problems.

[0005] The purpose of the present application can be achieved by the following technical solutions.

[0006] A CRM-based information data analysis method, the method comprising the following steps:

[0007] S1, collecting user original information data based on CRM, and screening target users with churn risk according to the user original information data;

[0008] S2, extracting customer churn risk features from the original information data of the target users, and importing the customer churn risk features into a trained risk assessment calculation model to output a result indicating the churn risk level of each target user;

[0009] S3, generating a retention scheme according to the churn risk level of the target user and the corresponding original information data and outputting it to the corresponding target user;

[0010] S4, collecting and analyzing feedback information of the target user on the retention scheme, and excluding completely churned customers according to the analysis result.

[0011] As a further technical solution, the original information data related to the user includes:

[0012] User basic information data, including identity information and occupation information;

[0013] Interaction information data, including customer service records and platform interaction records;

[0014] Behavioral information data; including login behavior data and transaction behavior data;

[0015] Contract information data, including contract terms and renewal dynamics.

[0016] As a further technical solution, the process of screening out target users at risk of churn includes:

[0017] Based on the user's basic information data, exclude blacklisted users and test accounts;

[0018] Filter out the user set C from the remaining users whose login time exceeds the first preset time. a The user set C whose contract expires and is not renewed for more than the second preset time b ;

[0019] By C target =C a ∪C b Get user set C a and user set C b The union C of target , and marked as the target user set;

[0020] If the target user set C target If the number of users included exceeds the preset value, step S2 is continued to be executed; otherwise, steps S2 to S4 are not executed.

[0021] As a further technical solution, the process of extracting customer churn risk features from the original information data of the target user is to clean the original information data of the target user and then extract the following customer churn risk features:

[0022] Time series features, including the target user's login times and login duration within a unit time period;

[0023] Transaction characteristics, including historical transaction times and total spending amount;

[0024] Sentimental characteristics, including user satisfaction with historical transactions;

[0025] Characteristics of the agreement conditions, including the validity period and effective date of the contract reached between the user and the platform.

[0026] As a further technical solution, the process of importing the customer churn risk features of the target user into the trained risk assessment calculation model and outputting the result indicating the churn risk level of each target user is as follows:

[0027] The customer churn risk feature parameters of the target user are imported into the calculation model:

[0028]

[0029] The customer churn risk value R is output, and based on the preset threshold rule, the customer churn risk value R is mapped to the discrete risk level interval by linear segmentation method, including: high risk level interval, medium risk level interval and low risk level interval;

[0030] Wherein, n is the number of login times of the target user in the current stage per unit time; n last is the average value of the number of login times of all unit time periods in the historical stage; d is the duration parameter from the last contract effective date of the target user to the present; d sum is the last contract effective date of the target user; E is the total consumption amount of the target user in all historical transactions; m is the number of historical transactions of the target user; μ is the satisfaction parameter of the target user to the historical transactions; ω1-ω3 are preset weights; ε1-ε3 are preset normalization coefficients.

[0031] As a further technical solution, the process of generating a retention scheme according to the churn risk level of the target user and the corresponding original information data includes:

[0032] The input conditions include the churn risk level corresponding to the risk level interval of the target user, the user basic information data and the interaction information data;

[0033] A hierarchical intervention strategy is adopted to match a differentiated scheme according to the churn risk level of the target user;

[0034] According to the basic information data and the interaction information data of the target user, the push text of the differentiated scheme is output and sent to the client.

[0035] As a further technical solution, the process of collecting and analyzing the feedback information of the target user to the retention scheme includes:

[0036] After the push text of the differentiated scheme is sent out, the feedback information is obtained through the dialogue window monitoring;

[0037] The feature parameters in the feedback information are extracted, including the response delay duration and the dialogue round;

[0038] After the feature parameters in the feedback information and the churn risk level parameters of the target user are normalized, they are imported into a weighted operation model, and the output result is a churn tendency score.

[0039] The screening threshold is obtained by the churn tendency score of the completely lost users in history, the churn tendency score of the current target user is compared with the screening threshold, if the churn tendency score of the target user exceeds the screening threshold, the corresponding target user is marked as a completely lost customer and screened out, otherwise, the target user is continuously retained.

[0040] A CRM-based information data analysis system, the system comprises:

[0041] An information collection module collects user-related raw information data based on CRM;

[0042] A feature extraction module is used to extract customer churn risk features from the user-related raw information data;

[0043] A risk analysis module imports the customer churn risk features into a trained calculation model to obtain a customer churn risk level through calculation;

[0044] A dynamic intervention strategy generation module is used to output a retention plan according to the risk level and the raw information data;

[0045] A feedback analysis module is used to collect and analyze feedback information of the target user for the retention plan;

[0046] A churn determination module determines and screens out completely lost customers according to the analysis result.

[0047] The beneficial effects of the present application are:

[0048] The present application can reduce the data processing amount by screening target users, focus on customers with high churn possibility, quickly narrow the monitoring range, then calculate the customer churn risk level to accurately quantify the customer churn risk degree, the present application classifies and intervenes the target users based on different risk levels, avoids using a unified retention strategy for all users, and improves the customer retention rate. Based on user feedback, completely lost customers can be screened out, which can avoid excessive disturbance to customers, improve customer experience, and reduce waste of marketing resources. BRIEF DESCRIPTION OF DRAWINGS

[0049] The present application will be further described below with reference to the accompanying drawings.

[0050] Figure 1 The flowchart of the CRM-based information data analysis method in the present application;

[0051] Figure 2 The framework diagram of the CRM-based information data analysis system in the present application; DETAILED DESCRIPTION

[0052] With reference to the drawings of the embodiments of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described. Obviously, the described embodiments are only a part of the embodiments of the present application, but not all the embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the scope of the present application.

[0053] Please refer to Figure 1 As shown in the drawings, a CRM-based information data analysis method comprises the following steps:

[0054] S1, collecting user original information data based on CRM, screening target users with loss risk according to user original information data, which includes transaction records, service interaction records, user login or use frequency, etc. The above preliminary screening can reduce data processing amount, focus on customers with high loss possibility, and improve subsequent analysis efficiency. In addition, blacklisted users and test accounts can also be screened out.

[0055] S2, extracting customer loss risk features from the original information data of the target users, which specifically include transaction features, interaction features, user behavior features, etc., and then importing the customer loss risk features into the trained risk assessment calculation model to output the results indicating the loss risk level of each target user. Specifically, a trained machine learning model such as random forest or deep learning model is used to calculate the loss probability and divide the risk level;

[0056] S3, generating a retention plan according to the loss risk level of the target user and the corresponding original information data and outputting it to the corresponding target user. Differentiated plans are developed according to different risk levels, for example, for low-risk users, optimize recommended content, send satisfaction surveys; for medium-risk users, issue coupons; for high-risk customers, provide customer manager follow-up services. In addition, the push text in the differentiated plan can be automatically generated according to the user's original information, such as the way of addressing, the type of recommended products, etc.

[0057] S4, collecting and analyzing feedback information of the target users on the retention plan, and screening out completely lost customers according to the analysis results. Monitor the user's response to the retention plan. If the user does not respond or explicitly terminates the service, stop marketing resource investment to avoid wasting marketing costs.

[0058] Through the technical solution, the embodiment provides an information data analysis method based on CRM, can reduce data processing amount by screening target users, focuses on customers with high possibility of loss, quickly narrows the monitoring range, then calculates the customer loss risk level, accurately quantifies the customer loss risk degree, and then classifies and intervenes the target users based on different risk levels, avoids adopting a unified retention strategy for all users, and improves the customer retention rate. Based on user feedback, completely lost customers are screened out, which can avoid excessive disturbance to customers, improve customer experience, and reduce waste of marketing resources.

[0059] The user-related raw information data includes:

[0060] The user basic information data includes identity information and occupation information, and the identity information includes age, gender, and region, and the occupation information includes industry, position, and income level. The above information is collected to provide reference for intervention strategy and facilitate generation of pushed text, for example, high-income occupation groups may not be sensitive to price, but service requirements are higher.

[0061] The interaction information data includes customer service records and platform interaction records, and the customer service includes complaint, consultation, and solution time length, and the platform interaction record includes positive and negative evaluation.

[0062] The behavior information data includes login behavior data and transaction behavior data, and the login behavior data includes login frequency and time length, and the transaction service includes transaction times and amount.

[0063] The contract information data includes contract clause information and renewal dynamic information.

[0064] Through the technical solution, the embodiment provides specific content of the user-related raw information data.

[0065] The process of screening the target users with loss risk includes:

[0066] According to the user basic information data, the blacklisted users and test accounts are excluded first to ensure that the analysis object is an effective customer;

[0067] In the remaining users, a user set C is screened out, which is a user set whose login time length exceeds a first preset time length (for example, 20 days) a And a user set C is screened out, which is a user set whose contract expiration and non-renewal time length exceeds a second preset time length (for example, 15 days) b ;

[0068] The union set C of the user set C and the user set C is obtained target C a ∪C b C a , and the union set C of the user set C b C targetand mark as a target user set;

[0069] If the target user set C target If the number of users contained exceeds a preset value, continue to perform step S2; otherwise, do not perform steps S2-S4. Only when the number of target users reaches a preset threshold (such as 100 people), the subsequent deep analysis and intervention are triggered, avoiding resource waste.

[0070] Through the above technical solutions, the embodiment quickly locates high-risk users through preliminary screening, controls analysis cost by combining scale verification, ensures that only a sufficient number of potential loss customers are effective in subsequent steps, and balances efficiency and accuracy.

[0071] The process of extracting customer loss risk features from the original information data of the target user is to extract the following customer loss risk features after cleaning the original information data of the target user:

[0072] The time series feature is used to identify the active decay trend, including the number of logins and login duration of the target user in a unit time period;

[0073] Transaction features, including historical transaction frequency and total consumption amount; statistics of consumption frequency and total amount to judge user value and consumption willingness change.

[0074] Emotional features, including user historical transaction satisfaction, which can be obtained by quantifying the star rating (1-5 stars) submitted by the user after completing the transaction;

[0075] Protocol condition features, including the effective period and effective period of the contract reached by the user and the platform.

[0076] Through the above technical solutions, the embodiment provides a process of extracting customer loss risk features from the original information data of the target user. The process extracts four types of features of behavior, transaction, emotion, and contract from the cleaned data, quantifies user activity, consumption, satisfaction, and protocol constraint, and builds a multi-dimensional loss risk assessment system to provide structured input for subsequent model analysis.

[0077] The process of importing customer loss risk features into a trained risk assessment calculation model and outputting results indicating the loss risk level of each target user is:

[0078] Importing the customer loss risk feature parameters of the target user into the calculation model:

[0079]

[0080] Output the customer churn risk value R, and based on the preset threshold rule, map the customer churn risk value R to discrete risk level intervals, including a high risk level interval, a medium risk level interval, and a low risk level interval, by a linear segmentation method. In the formula, Reflects the contract state. If is a positive value, it indicates that the current contract is not renewed, and the corresponding churn risk is larger. If is a negative value, it indicates that the current contract is within the contract period, and the corresponding churn risk is smaller.

[0081] Where n is the number of logins of the target user in the current stage per unit time; n last is the average number of logins of all unit time periods in the historical stage; d is the duration parameter from the last contract effective period of the target user to the present time; d sum is the last contract effective period of the target user; E is the total consumption amount of the target user in all historical transactions; m is the historical transaction number of the target user; μ is the satisfaction parameter of the target user to the historical transactions, which can be specifically the average value of the star ratings after all historical transactions; ω1-ω3 are preset weights; and ε1-ε3 are preset normalization coefficients. Through the above technical solution, the process of calculating the user churn risk level is provided.

[0082] The process of generating a retention plan according to the churn risk level of the target user and the corresponding original information data includes:

[0083] Input conditions, including the churn risk level corresponding to the risk level interval of the target user, user basic information data, and interaction information data;

[0084] A hierarchical intervention strategy is adopted to match a differentiated plan according to the churn risk level of the target user. Specifically, for example:

[0085] For a low-risk level user, implement a mild reminder, such as a satisfaction survey.

[0086] For a medium-risk level user, implement active care, such as issuing a coupon.

[0087] For a high-risk level user, implement a strong retention strategy, such as free value-added services.

[0088] According to the basic information data and the interaction information data of the target user, output a push text of the differentiated plan, which is used to clarify the plan content, with a corresponding activity link, and is sent to the client. The basic information data of the user can be used as a reference for addressing when editing the text, and the interaction information data can be used as a product category limit in the push text.

[0089] Through the technical solution, the embodiment provides a process of generating a retention scheme according to a target user's risk level of loss and corresponding original information data.

[0090] The process of collecting and analyzing feedback information of the target user for the retention scheme includes:

[0091] After the push text of the differentiated scheme is sent out, feedback information is obtained through a dialogue window;

[0092] Feature parameters in the feedback information are extracted, including a response delay duration and a dialogue round (a number of times of initiating a dialogue by the customer after receiving the push text); after the feature parameters in the feedback information and the risk level parameter of the target user are normalized, the normalized parameters are input into a weighted operation model, and an output result is a loss tendency score;

[0093] For example, the risk level of loss is The response delay duration is x2.

[0094] The normalized processing is: The mean value of x1-x3 is x, and sigma is the standard deviation of x1-x3.

[0095] The loss tendency score S is calculated by the formula

[0096] A screening threshold is obtained by the loss tendency score of the historical completely lost user, for example, the average value of the loss tendency score of each completely lost user is taken as the screening threshold, the loss tendency score of the current target user is compared with the screening threshold, if the loss tendency score of the target user exceeds the screening threshold, the corresponding target user is marked as a completely lost customer and screened out; otherwise, the target user is retained.

[0097] Through the technical solution, the embodiment provides a process of collecting and analyzing feedback information of the target user for the retention scheme.

[0098] Referring to FIG. 1, Figure 2 a CRM-based information data analysis system is shown, the system includes:

[0099] An information collection module collects user-related original information data based on CRM, obtains basic data, transaction records, interactive behaviors and other original information of the user from a CRM system, and provides a data basis for subsequent analysis;

[0100] A feature extraction module is configured to extract customer loss risk features from the user-related original information data; key features are extracted from the original data, and are used to evaluate the loss risk.

[0101] ​The risk analysis module imports the customer churn risk characteristics into the trained calculation model, obtains the customer churn risk level through calculation, calculates the customer churn probability, and divides the risk level, such as low, medium, and high

[0102] The dynamic intervention strategy generation module is used for outputting a retention scheme according to the risk level and the original information data, and automatically generating a personalized retention scheme, such as a coupon or exclusive service, according to the risk level and the original information data of the user.

[0103] The feedback analysis module is used for collecting and analyzing the feedback information of the target user to the retention scheme, monitoring the response of the user to the intervention measure, and evaluating the effectiveness of the strategy.

[0104] The churn determination module determines and filters out the completely churned customers according to the analysis result. Based on the feedback data, the completely churned customers are identified and removed from the active recovery list, so as to optimize the allocation of resources.

[0105] Through the above technical solutions, the system provided by the embodiment can intervene in the target user based on different risk levels, avoid using a unified retention strategy for all users, and improve the customer retention rate. Based on the feedback of the user, the completely churned customers are filtered out, so as to avoid excessive disturbance to the customers, improve the customer experience, and reduce the waste of marketing resources.

[0106] The above describes one embodiment of the present application in detail, but the content described is only the preferred embodiment of the present application, and cannot be considered as limiting the scope of the present application. Any equivalent changes and improvements made within the scope of the present application should still belong to the patent coverage range of the present application.

Claims

1. A CRM-based information data analysis method, characterized in that: The method comprises the following steps: S1. Collect user original information data based on CRM and screen out target users with churn risk based on the user original information data; S2. Extract customer churn risk features from the original information data of target users, import the customer churn risk features into the trained risk assessment calculation model, and output the results indicating the churn risk level of each target user; S3. Generate a retention plan based on the target user's churn risk level and the corresponding original information data and output it to the corresponding target user; S4. Collect and analyze target users' feedback on the retention plan, and screen out completely lost customers based on the analysis results; The process of extracting customer churn risk features from the target user's original information data is to clean the target user's original information data and then extract the following customer churn risk features: Time series features, including the target user's login times and login duration within a unit time period; Transaction characteristics, including historical transaction times and total spending amount; Sentimental characteristics, including user satisfaction with historical transactions; Characteristics of the terms of the agreement, including the validity period and effective date of the contract reached between the user and the platform; The process of importing customer churn risk characteristics into the trained risk assessment calculation model and outputting the results indicating the churn risk level of each target user is as follows: Import the parameters corresponding to the target user's customer churn risk characteristics into the calculation model: ， Output customer churn risk value , and then based on the preset threshold rules, the customer churn risk value Mapping to discrete risk level intervals through linear segmentation method, including: high risk level interval, medium risk level interval and low risk level interval; in, The number of logins per unit time for the target user in the current stage; The average number of logins in all unit time periods in the historical stage; The duration parameter from the target user's most recent contract validity date to the present; The most recent contract validity period of the target user; The total amount of consumption of the target user in all historical transactions; The historical transaction count of the target user; is the target user's satisfaction parameter with historical transactions; ~ is the preset weight; ~ is the preset normalization coefficient; The process of collecting and analyzing target users' feedback on the retention plan includes: After the push text of the differentiation plan is sent, feedback information is obtained through the dialogue window monitoring; Extract characteristic parameters from feedback information, including response delay and conversation turns; After normalizing the characteristic parameters in the feedback information and the target user's churn risk level parameters, they are imported into the weighted calculation model, and the output result is the churn propensity score; The screening threshold is obtained by using the churn propensity scores of historical completely churned users. The churn propensity score of the current target user is compared with the screening threshold. If the churn propensity score of the target user exceeds the screening threshold, the corresponding target user is marked as a completely churned customer and screened out; otherwise, the target user is retained.

2. The CRM-based information data analysis method according to claim 1, characterized in that: User-related raw information data includes: User basic information data, including identity information and occupation information; Interaction information data, including customer service records and platform interaction records; Behavioral information data; including login behavior data and transaction behavior data; Contract information data, including contract terms and renewal dynamics.

3. The CRM-based information data analysis method according to claim 2, characterized in that: The process of screening out target users at risk of churn includes: Based on the user's basic information data, exclude blacklisted users and test accounts; Filter out the users who have not logged in for more than the first preset time from the remaining users The collection of users whose contracts have expired and not been renewed for more than the second preset period ; pass Get user collection and user collections The union of , and marked as the target user set; If the target user set If the number of users included exceeds the preset value, step S2 is continued to be executed; otherwise, steps S2 to S4 are not executed.

4. The CRM-based information data analysis method according to claim 3, characterized in that: The process of generating a retention plan based on the target user's churn risk level and the corresponding original information data includes: Input conditions include the target user's churn risk level corresponding to the risk level range, user basic information data, and interaction information data; Adopt a graded intervention strategy and match differentiated solutions based on the target user's churn risk level; According to the basic information data and interactive information data of the target user, the push text of the differentiated solution is output and sent to the client.

5. A CRM-based information data analysis system, characterized in that: The system is used to perform the CRM-based information data analysis method according to any one of claims 1 to 4, and the system includes: Information collection module, which collects user-related original information data based on CRM; Feature extraction module, used to extract customer churn risk features from user-related raw information data; The risk analysis module imports customer churn risk characteristics into the trained calculation model and obtains the customer churn risk level through calculation; Dynamic intervention strategy generation module, used to output retention plans based on risk levels and original information data; Feedback analysis module, used to collect and analyze target users' feedback on the retention plan; The churn determination module determines and screens out completely lost customers based on the analysis results.

Images