KR102162765B1 - System and method for automated management of customer churn based on artificial intelligence and computer program for the same - Google Patents

Abstract

The customer churn management automation system includes: a data collection unit configured to receive customer information and customer-related operation system information of a plurality of customers from a client device, and convert the customer information and the operation system information into an analysis data set; A customer segmentation unit configured to extract customer relationship information from the analysis data set and classify the plurality of customers into one or more churn pattern groups based on the customer relationship information; A departure prediction unit configured to calculate a departure risk of each of the one or more departure pattern groups from the analysis data set using a machine learning based classification prediction algorithm; And a personalized recommendation unit configured to execute a personalized recommendation for preventing churn to potential churn customers determined based on the churn risk. According to the above system, a series of customer churn management processes ranging from customer segmentation, churn prediction, and churn prevention can be implemented as an automated service based on machine learning.

Description

Artificial intelligence-based customer churn management automation system and method, and a computer program for it {SYSTEM AND METHOD FOR AUTOMATED MANAGEMENT OF CUSTOMER CHURN BASED ON ARTIFICIAL INTELLIGENCE AND COMPUTER PROGRAM FOR THE SAME}

The embodiments relate to an artificial intelligence-based customer churn management automation system and method, and a computer program therefor. More specifically, the embodiments relate to a technology for implementing a series of customer churn management processes ranging from customer segmentation, churn prediction, and churn prevention as an automated service based on machine learning.

Most companies have not yet performed customer churn management, and even companies that perform churn management are quite limited, or work in a form that is disconnected from other marketing tasks. Technically, a machine learning algorithm for binary prediction is applied to all customers to measure the risk of churn, and spamming marketing activities to induce simple purchases for customers with high churn risk are mainstream.

However, the customer churn management method according to the prior art includes (i) applying an existing machine learning model that does not reflect business characteristics that vary depending on the industry, and (ii) all targets without considering customer transaction patterns. The same model is applied to the customer, (iii) the business process between the churn prediction activity and the marketing activity is disconnected, (iv) only the churn prevention activity in the form of a simple purchase promotion proposal that does not take into account the cause of the churn, And (v) continuous maintenance is not possible due to the burden of a machine learning project that requires high costs.

Registered Patent Publication No. 10-1726404

According to an aspect of the present invention, a customer who overcomes the limitations of the prior art and enables even small and medium-sized enterprises that lack data science capabilities to systematically perform automated customer churn management based on machine learning. It is possible to provide a departure management automation system and method, and a computer program for the same.

The customer churn management automation system according to an embodiment is configured to receive customer information of a plurality of customers and customer-related operation system information from a client device, and convert the customer information and the operation system information into an analysis data set. Collection unit; A customer segmentation unit configured to extract customer relationship information from the analysis data set and classify the plurality of customers into one or more churn pattern groups based on the customer relationship information; A departure prediction unit configured to calculate a departure risk of each of the one or more departure pattern groups from the analysis data set using a machine learning based classification prediction algorithm; And a personalized recommendation unit configured to perform a personalized recommendation for preventing churn to potential churn customers determined based on the churn risk.

The automated customer churn management system according to an embodiment further includes a churn-cause estimating unit configured to determine the churn cause of the potential churning customer using a machine learning-based estimation algorithm. At this time, the type of personalized recommendation executed by the personalized recommendation unit is determined based on the reason for departure determined by the reason for departure estimation unit.

The automated customer churn management system according to an embodiment further includes a personal behavior statistics database configured to store the results of the personalized recommendation executed to the potential churn customer into a database.

In one embodiment, the operating system information includes at least one of customer transaction amount information, transaction frequency information, cross-purchase index information, communication index information, and whether or not to receive marketing information.

A method for automating customer churn management according to an embodiment includes the steps of, by a customer churn management automation system, receiving customer information of a plurality of customers and customer-related operating system information from a client device; Converting, by the customer churn management automation system, the customer information and the operating system information into an analysis data set; Extracting, by the customer churn management automation system, customer relationship information from the analysis data set; Classifying, by the customer churn management automation system, the plurality of customers into one or more churn pattern groups based on the customer relationship information; Calculating, by the customer churn management automation system, a churn risk of each of the one or more churn pattern groups from the analysis data set using a machine learning-based classification prediction algorithm; And executing, by the customer churn management automation system, a personalized recommendation to prevent churn to potential churn customers determined based on the churn risk.

The automated customer churn management method according to an embodiment further includes the step of determining, by the customer churn management automation system, a churn cause of the potential churning customer using a machine learning-based estimation algorithm. In this case, the type of personalized recommendation executed in the step of executing the personalized recommendation is determined based on the reason for departure determined in the step of determining the cause of departure.

The customer churn management automation method according to an embodiment further includes the step of storing, by the customer churn management automation system, a result of the personalized recommendation executed to the potential churn customer into a database.

A computer program according to an embodiment is for executing the method for automating customer churn management according to the above-described embodiments by being combined with hardware and may be stored in a computer-readable medium.

According to an automated system and method for customer churn management according to an aspect of the present invention, a big data analyst within an organization by automating a series of processes for effective customer churn management of a company, such as customer segmentation, churn prediction, and personalized recommendations for churn prevention. There is an advantage of being able to perform a machine learning-based churn management task even in an organization that does not have it.

In addition, the customer churn management automation system and method according to an aspect of the present invention accumulates information of an efficient machine learning model for customer churn management by industry and churn management results according to customer behavior in the form of big data, and is based on this. By performing customer churn management, there is an advantage in that it can provide customized customer churn management optimized for the characteristics of the target company and potential churn customers.

Through the automated system and method for customer churn management according to an aspect of the present invention, a company can achieve effective effects in increasing sales and maintaining customers by accommodating the main business area of customer management called customer churn management.

1 is a schematic block diagram of a customer churn management automation system according to an embodiment of the present invention.
2 is a conceptual diagram showing a service flow by an automated customer churn management system according to an embodiment.
3 is a flow chart showing each step of a method for automating customer churn management according to an embodiment.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the drawings.

1 is a schematic block diagram of a customer churn management automation system according to an embodiment.

Referring to FIG. 1, the customer churn management automation system 2 according to the present embodiment is configured to operate while communicating with the client device 1 through a wired and/or wireless network. The communication method between the customer churn management automation system 2 and the client device 1 may include all communication methods in which an object and an object can network through wired and/or wireless networks, wired communication, wireless communication, It is not limited to 3G, 4G, or other methods.

For example, a wired and/or wireless network between the customer churn management automation system 2 and the client device 1 is a local area network (LAN), a metropolitan area network (MAN), a global system for mobile network (GSM), EDGE (Enhanced Data GSM Environment), HSDPA (High Speed Downlink Packet Access), W-CDMA (Wideband Code Division Multiple Access), CDMA (Code Division Multiple Access), TDMA (Time Division Multiple Access), Bluetooth (Bluetooth), Zigbee (Zigbee), Wi-Fi, VoIP (Voice over Internet Protocol), LTE Advanced, IEEE802.16m, WirelessMAN-Advanced, HSPA+, 3GPP Long Term Evolution (LTE), Mobile WiMAX (IEEE 802.16e), UMB (formerly EV-DO Rev. C), Flash-OFDM, iBurst and MBWA (IEEE 802.20) systems, HIPERMAN, Beam-Division Multiple Access (BDMA), Wi-MAX (World Interoperability for Microwave Access) and ultrasonic communication It may refer to a communication network using one or more communication methods selected from the group consisting of, but is not limited thereto.

The client device 1 is a device of a user who wants to enjoy the customer churn management service provided by the customer churn management automation system 2. At this time, the user refers to an individual or corporate business that corresponds to a company that provides products and/or services to one or more customers. At this time, the business of the company may be in any form requiring churn management for customers, and It is not limited to those related to products or services.

In FIG. 1, the client device 1 is shown as a single device, but this is for convenience of explanation, and the number of companies using the customer churn management automation system 2 according to the embodiments or the number of devices used by the company It will be readily understood that the number is not limited.

The client device 1 may include an operating system database (DB) 12 in which operating system information related to a customer of a company using the client device 1 is stored. Operational information refers to arbitrary information necessary to realize customer churn management by the customer churn management automation system (2), and may include, for example, transaction information of each customer, customer behavior information, etc., but is limited thereto. It is not. In addition, the customer's transaction information may include information such as product or service information, transaction amount, transaction frequency, and transaction date and time. Furthermore, the customer's behavior information may include any quantitative and/or qualitative information defining a relationship between the customer and the user of the client device 1, such as a customer's past communication history or whether or not to receive marketing information.

In addition, the client device 1 may include a data linking unit 11 that provides operating system information to the customer churn management automation system 2. The data linkage unit 11 serves to transmit part or all of the operating system information stored in the operating system DB 12 to the customer churn management automation system 2. The transmission of operating system information may be performed by the data linking unit 11 in response to a request from the customer churn management automation system 2, or in the absence of a separate request from the customer churn management automation system 2 It may be performed periodically by the data linking unit 11. In addition, the operating system information transmitted to the customer churn management automation system 2 may be data of a specific item designated by the customer churn management automation system 2 among all the operating system information, or may be the entire operating system information.

In one embodiment, the customer churn management automation system 2 includes a data collection unit 21, a customer segmentation unit 22, a departure prediction unit 23, and a personalized recommendation unit 25. In an embodiment, the customer churn management automation system 2 further includes a churn cause estimation unit 24. Further, in one embodiment, the customer churn management automation system 2 further includes a service interface unit 26 and/or a management interface unit 27.

The customer churn management automation system 2 according to the embodiments and each unit 21-27 included therein may have an aspect that is entirely hardware, or partially hardware and partially software. For example, the customer churn management automation system (2) and each unit (21-27) included therein may collectively refer to hardware and related software for processing data in a specific format and content or/or sending and receiving data in an electronic communication method. have. In this specification, terms such as "unit", "module", "device", "terminal", "server" or "system" are intended to refer to a combination of hardware and software driven by the hardware. For example, the hardware may be a data processing device including a CPU or other processor. In addition, software driven by hardware may refer to an executing process, an object, an executable file, a thread of execution, a program, and the like.

In addition, each element constituting the customer churn management automation system 2 is not necessarily intended to refer to separate devices that are physically separated from each other. That is, the data collection unit 21, the customer segmentation unit 22, the departure prediction unit 23, the departure cause estimation unit 24, the personalized recommendation unit 25, the service interface unit 26, and the management interface of FIG. 1 The unit 27 is only functionally divided hardware constituting the customer churn management automation system 2 according to an operation performed by the corresponding hardware, and each unit is not necessarily provided independently of each other.

Of course, depending on the embodiment, the data collection unit 21, the customer segmentation unit 22, the departure prediction unit 23, the departure cause estimation unit 24, the personalized recommendation unit 25, the service interface unit 26, and One or more of the management interface units 27 may be implemented as separate devices that are physically separated from each other.

2 is a conceptual diagram showing a service flow by the customer churn management automation system 2 according to an embodiment. 1 and 2, functions of each part of the automated customer churn management system 2 according to the present embodiment and an automated service process through the same will be described.

First, the data collection unit 21 may receive customer information and customer-related operating system information from the data linking unit 11 of the client device 1. At this time, customer information refers to basic information related to each customer, such as each customer's gender, age, and subscription period. In addition, operational information refers to customer transaction information and customer behavior information. In one embodiment, the data collection unit 21 may further receive corporate information of a company corresponding to the client device 1.

The data collection unit 21 may generate an analysis data set for analysis based on a machine learning model from the collected customer information and operating system information. The analysis data set may simply be a collection of information necessary for machine learning, or data of a specific item according to a data format previously designated from customer information and operation system information by the data conversion engine 210 of the data collection unit 21 May be extracted, transformed, and loaded.

The customer segmentation unit 22 may classify customers into a plurality of groups having at least partially different characteristics by determining a segmentation criterion suitable for customer churn management and performing multidimensional segmentation based on this criterion. In this case, the segmentation criterion for grouping customers may be determined according to the characteristics of the company corresponding to the client device 1.

In one embodiment, the customer segmentation unit 22 extracts the relationship information of each customer from the analysis data set obtained by the data collection unit 21, and based on the customer relationship information, a plurality of customers are grouped into one or more churn patterns. It can be classified as In the present specification, customer relationship information refers to a predetermined degree of closeness of a customer to a company's product or service by using the transaction amount, transaction frequency, transaction date, communication history, whether or not to receive marketing information, or other appropriate information included in the analysis data set. It is information calculated based on the criteria, and the calculation criteria may vary. In addition, the customer relationship information may be information quantified as one or a plurality of values, or information obtained by subdividing customers into ratings or stages based on these values or numbers, and is not limited to a specific form.

In one embodiment, the customer relationship information may be loyalty information that evaluates a degree of utilization of a customer for a company's product or service financially or in terms of interaction. Also, in an embodiment, the loyalty information may include quantitative loyalty information and qualitative loyalty information. For example, when the quantitative royalty information is L _qnt and the qualitative royalty information is L _qlt , the quantitative royalty and qualitative royalty information may be calculated as in Equations 1 and 2 below.

[Equation 1]

[Equation 2]

및

는 미리 설정된 조정 상수이며,

,

,

,

는 미리 설정된 비례 상수로서, 이들 상수의 값은 양적로열티 및 질적로열티의 산출 수식에 따라 적절히 결정될 수 있다. In Equations 1 and 2, Monetary represents the customer's transaction amount, and Frequency represents the customer's transaction frequency. That is, in Equation 1, the transaction amount and the transaction frequency are used to define the quantitative level of the transaction relationship. In addition,

And

Is a preset adjustment constant,

,

,

,

Is a preset proportionality constant, and the values of these constants can be appropriately determined according to calculation formulas of quantitative and qualitative royalty.

In addition, in Equations 1 and 2, Communication represents a communication index between a customer and a company, and Openness represents a customer's openness index. The communication index is an index of the degree of communication between customers and companies through a number of counters such as A/S inquiries, complaints reception, and marketing guidance, and the openness index is based on whether or not customers receive marketing information. It is an index of customer openness. For example, the higher the communication index is, the more customers who communicate various requirements to companies and receive feedback on them. In addition, if a customer is receiving company announcements or marketing information without rejecting it, the customer's openness index will be determined to be high.

The customer segmentation unit 22 categorizes the numerical values of the quantitative royalty and qualitative royalty information calculated by Equations 1 and 2 into a predetermined range (eg, manageable scale), and each customer segment divided by royalty Areas where segments intersect can be established as individual customer segments.

For example, if an individual customer segment corresponding to the i-th segment based on quantitative loyalty and the j-th segment based on qualitative loyalty among m customer segments is Seg _ij , the individual customer segment may be defined as in Equation 3 below. . That is, when three first segments based on quantitative royalty and three second segments based on qualitative royalty are obtained, the total number of customer segments may be 9 according to Equation 3.

[Equation 3]

However, the above-described formula is merely exemplary, and the types of data for calculating quantitative royalty and qualitative royalty and a method of calculating the same may be different from those described above. In addition, in another embodiment, loyalty information is divided into quantitative and qualitative loyalty to define loyalty in a different form other than a method of dividing customer segments, or it is possible to group customers based on differently defined relationship information other than loyalty. Do.

Next, the customer segmentation unit 22 analyzes the churn pattern indicators such as churn rate and churn loss amount for customer segments by the customer segmentation engine 220, and combines the segments showing similar churn patterns to provide a plurality of customer information. May be formed as one or more separation pattern groups. In this case, each departure pattern group may include only one customer segment, or may include a plurality of customer segments having a similarity in which the reference value of the departure pattern is preset.

In one embodiment, the customer segmentation engine 220 performs a test for the difference between each customer segment for a churn pattern index such as churn rate, churn loss amount, etc. through a mutivariate analysis of variance (ANOVA) technique. Thereafter, similar segments in which the difference between the departure pattern index is less than a predetermined threshold value may be grouped into one group to form a departure pattern group.

For example, assume a multidimensional segment in which a number of customers are divided into 9 customer segments by the intersection of 3 segments based on quantitative royalty and 3 segments based on qualitative royalty. At this time, the heterogeneity between segments is determined using the Wilk's Lambda value obtained by dividing the deviation pattern variance value within a specific segment by the total deviation pattern variance of all segments, and combining segments with similar Wilks lambda values to form one deviation pattern group. Can be formed by

는 하기 수학식 4와 같이 산출될 수 있다.For example, Wilkes Lambda value

Can be calculated as in Equation 4 below.

[Equation 4]

은 다변량분산분석(MANOVA)을 통해 검증하고자 하는 p개 차원의 람다(

)값을 의미하며, 람다(

) 값은 일반적으로 오차변량의 행렬식 값(det(Σ_res))을 전체 변량의 행렬식 값(det(Σ_res + Σ_seg))으로 나누어 구하게 된다. 여기에서 res는 잔차(residual)를 의미하고, seg는 다차원 세그먼트 기준으로 분리된 각 세그먼트(segment)를 의미한다.In Equation 4 above

Is a lambda of p dimensions to be verified through multivariate variance analysis (MANOVA) (

) Means the value, and lambda (

) Is generally obtained by dividing the determinant value (det(Σ _res )) of the error variable by the determinant value (det(Σ _res + Σ _seg )) of the entire variable. Here, res means residual, and seg means each segment separated by a multidimensional segment.

The departure prediction unit 23 may calculate a churn risk for each customer by applying a churn prediction model to the churn pattern group generated as described above. To this end, the departure prediction engine 230 of the departure prediction unit 23 is equipped with a machine learning algorithm that can be used for classification prediction. For example, the departure prediction engine 230 is an artificial neural network or a deep learning model, a logistic regression model, a decision tree model, and a support vector machine. The risk of departure can be calculated by applying various machine learning models such as a model, a random forest, or an ensemble model such as gradient boosting.

In one embodiment, the churn prediction engine 230 calculates churn risk by applying various machine learning models using customer information and operating system information about customers with known churn as a training dataset, and calculates the churn risk level calculated by each model. The performance of the model can be evaluated by comparing the deviation risk with the actual deviation. For example, the performance evaluation of the model can be calculated through a confusion matrix defined as shown in Table 1 below.

Actual / Forecast Prediction result Positive (departure) Negative (not churn) Actual result Positive (departure) Real positive
(True Positive; TP) False voice
(False Negative; FN) Negative (not churn) False positive
(False Positive; FP) Real voice
(True Negative; TN)

In one embodiment, the performance of a specific machine learning model can be evaluated using an F-measure, and the F-score is Equation 5 to Equation 5 below using precision and recall. It can be calculated as Equation 7.

[Equation 5]

[Equation 6]

[Equation 7]

However, the above-described method for evaluating the performance of a machine learning model using the F-score is exemplary, and the machine learning model may be evaluated according to other different criteria not described herein.

When the machine learning model to be applied to the churn prediction engine 230 is determined through the above process, the churn pattern group obtained from the unknown customer information and operating system information that does not know whether to churn using the determined machine learning model Risk can be calculated.

In one embodiment, the churn cause estimating unit 24 selects potential churn customers whose churn risk is equal to or greater than a preset threshold, and estimates possible churn causes for potential churn customers by using a machine learning-based churn cause estimation algorithm. You can decide.

Similar to the above-described churn prediction engine 230, the churn cause estimation unit 24 may include a churn cause estimation engine 240 that estimates the churn cause of a potential churn customer by applying a machine learning model. The reason of departure estimation engine 240 uses customer information and operating system information whose reason of departure is known as a learning dataset, and uses machine learning about what causes of departure of customers with patterns of specific personal information, transaction information, and/or behavioral information. Based on prediction, it is possible to analyze the cause of churn for potential churn customers obtained from unknown datasets that do not know the cause of churn based on the machine learning model determined to have excellent prediction results.

Next, the personalized recommendation unit 25 may execute personalized recommendations effective to prevent churn for potential churn customers. In this specification, the execution of a personalized recommendation refers to sending an offer for a specific product or service to a potential churning customer, encouraging a potential churning customer to participate in a specific event and/or event, or encouraging a potential churning customer to a specific event and/or Selecting as a beneficiary for the event (e.g., discounting transaction amount, giving points or coupons, etc.), or conducting communication for customer relationship management to potential churn customers (e.g., customer contact via phone, text, SNS, etc.) Etc.) and the like.

Similar to the above-described departure prediction engine 230 and departure cause estimation engine 240, the personalized recommendation unit 25 applies a machine learning model to determine the personalized recommendation content to be presented to potential churn customers ( 250). The personalized recommendation engine 250 performs machine learning model-based learning using past data with known actions and results (ie, final churn or transaction continuation) as a training dataset, thereby It is possible to determine which personalized recommendation is suitable for customers with patterns of information and/or behavioral information, and based on this, to determine a personalized recommendation suitable for unknown potential customers.

In addition, in an embodiment, the content of the personalized recommendation executed by the personalization recommendation unit 25 may be determined according to the reason of departure estimated by the reason of departure estimation unit 24. At this time, the personalized recommendation engine 250 performs machine learning model-based learning using past data for which the cause of departure and the execution result of personalized recommendation are known, as a training dataset, so that any personalized recommendation is suitable for customers with a certain cause of departure. You can also decide. The above process may be performed in the same process as the personalization recommendation reflecting the aforementioned personal information, transaction information, and/or behavior information, or may be performed in parallel therewith.

Meanwhile, the service interface unit 26 is a part that provides a user interface (UI) for the person in charge of the client device 1 to use the customer churn management automation system 2 according to the embodiments. For example, a customer relationship management (CRM) person in charge of a company corresponding to the client device (1), or a marketing person in charge, through the UI provided by the service interface unit (26), the customer churn management automation system (2) The information provided by this can be received, searched, viewed and/or printed.

In addition, the management interface unit 27 manages the interlocking of data necessary for the machine learning-based customer churn management operation by the customer churn management automation system 2, loading a machine learning model, and adjusting parameters related to machine learning algorithms. It is a part that provides a UI to perform a function. For example, a person in charge of a company corresponding to the operator of the customer churn management automation system 2 may define, monitor, or modify the customer churn management process using a UI provided by the management interface unit 27.

In one embodiment, the customer churn management automation system 2 further includes a storage module 3 communicatively connected with other modules of the customer churn management automation system 2. The storage module 3 may include an industry standard machine learning model 31 and/or a personal behavior statistics DB 32.

The industry standard machine learning model 31 refers to a machine learning model that is standardized by reflecting the characteristics of each industry of a company that performs customer churn management. In this embodiment, the customer churn management automation system 2 receives the business type of the company corresponding to the client device 1 from the client device 1, and the machine learning corresponding to the business type of the industry standard machine learning model 31 The above-described customer churn management can be performed using the model. For example, when a company is a department store, a distribution business customer churn management machine learning model that is specialized for the distribution industry may be used among the industry standard machine learning models 31 to implement churn management services.

Alternatively, the machine learning model stored in the industry standard machine learning model 31 may be a machine learning model that is generated based on data received in the past from the client device 1 and reflects the characteristics of the company. You can perform customized customer churn management.

In one embodiment, the machine learning model 31 stored in the storage module 3 is not limited to the classification by industry of the company corresponding to the client device 1, but detailed characteristics of the company (eg, sales, number of employees, capital, It may be classified according to operating profit, etc.). For example, when taken by a large company belonging to the top 10 companies, the effective churn pattern management and personalized recommendation method, and when taken by a start-up or small business, effective churn pattern management and personalized recommendation method may be different. By applying a machine learning model 31 suitable for this according to characteristics, machine learning-based analysis optimized for the size and/or financial situation of the company can be performed.

To this end, in one embodiment, the data collection unit 21 receives corporate customer information and operating system information from the client device 1, as well as the client device 1 from an external server (not shown) such as an IRS server. Company information (eg, number of employees, company size classification, financial statements, etc.) corresponding to the company can be received. The above information reception method may be performed by a method such as screen scraping in which the data collection unit 21 accesses a web page provided by an external server to collect necessary information, but is limited thereto. no.

At this time, the customer segmentation unit 22, the departure prediction unit 23, the departure cause estimation unit 24 and/or the personalized recommendation unit 25 collects data from among the machine learning models 31 of the storage module 3 By applying the machine learning model 31 corresponding to the company information collected in the unit 21, customer churn management can be performed based on the machine learning model 31 optimized for the size and financial status of the company.

In addition, the personal behavior statistics DB 32 is a big data storage in which data is accumulated as the service experience of the churn management service through the customer churn management automation system 2 increases. At this time, the customer behavior information stored in the personal behavior statistics DB 32 may be de-identified so that the customer information of which customer of a certain company cannot be identified. Since the accuracy of the churn prediction and personalized recommendation may not meet expectations only by analyzing the data inside the company, the data of a large number of customers of a number of companies are accumulated in the personal behavior statistics DB 32, based on this analysis and prediction shown in FIG. By performing the operations of the engines 210, 220, 230, 240, 250, it is possible to improve the accuracy of departure prediction and personalized recommendation.

3 is a flow chart showing each step of a method for automating customer churn management according to an embodiment.

Referring to FIG. 3, according to the automated method for customer churn management according to the present embodiment, customer information and operating system information may be first received from a client device (S1), and then converted into an analysis data set (S2).

Next, a plurality of customers may be classified into one or more churn pattern groups based on the analysis data set generated from customer information and operating system information. For example, relationship information of each customer may be extracted from customer information and operation system information (S3), and customers may be grouped using the relationship information (S4). At this time, a plurality of customers are separated into customer segments by two or more different variables, and the end customer segment is determined through intersections between heterogeneous customer segments determined by different variables, and one or more of these customer segments. It is also possible to determine an escape pattern group to include.

Next, by calculating the churn risk of each grouped churn pattern group, it is possible to determine the potential churn customer (S5). The deviation risk may be calculated by applying a machine learning model that has been evaluated to have high accuracy among machine learning models that have been learned by a training dataset with known departure or not.

In an embodiment, the cause of the churn may be estimated based on a machine learning model for a customer determined as a potential churn customer (S6). Through this process, it is possible to analyze which personal information and the characteristics of operating system information are mainly caused by which customers leave.

Next, it is possible to perform personalized recommendations to potential churn customers having a high churn risk (S7). Personalized recommendation means offering a product and/or service to a customer, sending information that encourages the customer to participate in an event and/or event, selecting a customer as a beneficiary of the event and/or event, or contacting a customer with an agent. It may include conducting communication, etc.

In one embodiment, a personalized recommendation method suitable for each cause of departure may be stored in advance according to the cause of the customer's churn, and personalized recommendations corresponding to the individual cause of departure may be executed to potential churn customers. In this case, a personalized recommendation method suitable for each cause of departure may be determined based on a result previously learned by a machine learning model, similar to the above-described process of grouping the departure pattern and estimating the cause of departure.

In one embodiment, the customer churn management result (i.e., churn risk calculation result, churn cause estimation result, personalized recommendation execution result) for one or more customers having specific customer information and operating system information may be stored in the personal behavior statistics DB. Yes (S7). The personal behavior statistics stored in this way can be used to provide customized customer management suitable for the business type or special situation to the company in the future.

The operation by the automated method for customer churn management according to the above-described embodiments may be implemented at least partially as a computer program and recorded on a computer-readable recording medium. A computer-readable recording medium in which a program for implementing an operation by the automated method for customer churn management according to the embodiments is recorded and includes all types of recording devices in which data that can be read by a computer are stored. Examples of computer-readable recording media include ROM, RAM, CD-ROM, magnetic tapes, floppy disks, and optical data storage devices. In addition, the computer-readable recording medium may be distributed over a computer system connected through a network, and computer-readable codes may be stored and executed in a distributed manner. In addition, functional programs, codes, and code segments for implementing the present embodiment may be easily understood by those skilled in the art to which the present embodiment belongs.

The present invention described above has been described with reference to the embodiments shown in the drawings, but these are merely exemplary, and those of ordinary skill in the art will understand that various modifications and variations of the embodiments are possible therefrom. However, such modifications should be considered to be within the technical protection scope of the present invention. Accordingly, the true technical scope of the present invention should be determined by the technical spirit of the appended claims.

Claims (9)

A data collection unit configured to receive customer information of a plurality of customers and customer-related operation system information from a client device, and convert the customer information and the operation system information into an analysis data set;
A customer segmentation unit configured to extract customer relationship information from the analysis data set and classify the plurality of customers into one or more churn pattern groups based on the customer relationship information;
A departure prediction unit configured to calculate a departure risk of each of the one or more departure pattern groups from the analysis data set using a machine learning-based classification prediction algorithm; And
Including a personalized recommendation unit configured to execute a personalized recommendation for preventing churn to a potential churning customer determined based on the churn risk,
The customer segmentation unit,
Extracting quantitative royalty information from the quantitative relationship information of the analysis data set that defines the quantitative level of the customer's transaction relationship,
Extracting qualitative royalty information from the qualitative relationship information of the above analysis dataset that indexes the degree of customer communication or the degree of customer openness,
A plurality of the customer segments are created by defining an area where customers within a numerical range defined by each of the quantitative loyalty information and the qualitative loyalty information intersect as a customer segment,
It is further configured to analyze the churn pattern index for the customer segment and combine the one or more customer segments determined based on the similarity of the churn pattern indicator to generate the churn pattern group,
The departure prediction unit is further configured to determine a machine learning-based classification prediction algorithm to be applied to each of the one or more departure pattern groups, and calculate the departure risk of a customer belonging to the departure pattern group using the determined classification prediction algorithm. Customer churn management automation system.
The method of claim 1,
Further comprising a churn cause estimation unit configured to determine the churn cause of the potential churn customer using a machine learning-based estimation algorithm,
The type of personalized recommendation executed by the personalized recommendation unit is determined based on the reason for departure determined by the reason for the departure estimation unit.
The method of claim 1,
The customer churn management automation system further comprises a personal behavior statistics database configured to store the result of the personalized recommendation executed to the potential churn customers into a database.
The method of claim 1,
The operating system information is customer churn management automation system including at least one of the customer's transaction amount information, transaction frequency information, cross-purchase index information, communication index information, and whether or not to receive marketing information.
Receiving, by the customer churn management automation system, customer information of a plurality of customers and customer-related operating system information from the client device;
Converting, by the customer churn management automation system, the customer information and the operating system information into an analysis data set;
Extracting, by the customer churn management automation system, customer relationship information from the analysis data set;
Classifying, by the customer churn management automation system, the plurality of customers into one or more churn pattern groups based on the customer relationship information;
Calculating, by the customer churn management automation system, a churn risk of each of the one or more churn pattern groups from the analysis data set using a machine learning-based classification prediction algorithm; And
The customer churn management automation system, comprising the step of executing a personalized recommendation to prevent churn to a potential churn customer determined based on the churn risk,
Classifying into the one or more departure pattern groups,
Extracting, by the customer churn management automation system, quantitative loyalty information from the quantitative relationship information of the analysis data set that defines the quantitative level of the customer's transaction relationship;
Extracting, by the customer churn management automation system, qualitative loyalty information from the qualitative relationship information of the analysis dataset obtained by indexing the degree of customer communication or the degree of customer openness;
Generating a plurality of the customer segments by defining, by the customer churn management automation system, a region where customers belonging to a numerical range defined by the quantitative loyalty information and the qualitative loyalty information intersect, as a customer segment;
Analyzing, by the customer churn management automation system, a churn pattern index for the customer segment; And
The customer churn management automation system comprises the step of creating the churn pattern group by combining one or more of the customer segments determined based on the similarity of the churn pattern index,
The step of calculating the departure risk,
Determining, by the customer churn management automation system, a machine learning-based classification prediction algorithm to be applied to each of the one or more churn pattern groups; And
And calculating, by the customer churn management automation system, the churn risk of a customer belonging to the churn pattern group using the determined classification prediction algorithm.
The method of claim 5,
The customer churn management automation system further comprises the step of determining a churn cause of the potential churn customer using a machine learning-based estimation algorithm,
The type of personalized recommendation executed in the step of executing the personalized recommendation is determined based on the reason for departure determined in the step of determining the cause of departure.
The method of claim 5,
The customer churn management automation system further comprises the step of storing the result of the personalized recommendation executed to the potential churn customer into a database and storing.
The method of claim 5,
The operating system information is customer churn management automation method including one or more of the customer's transaction amount information, transaction frequency information, cross-purchase index information, communication index information, and whether or not to receive marketing information.
A computer program stored in a computer-readable recording medium in combination with hardware to execute the automated method for customer churn management according to any one of claims 5 to 8.

Images