Client data cabin-based analysis method

Abstract

The invention relates to the field of data analysis, in particular to an analysis method based on a customer data cabin, and the analysis method comprises the following steps: setting a lost customer screening condition; on the basis of a lost customer screening condition, screening out lost customers from the customer data cabin, and extracting historical data of the lost customers from the customer data cabin; according to historical data of the lost customers, acquiring lost customer characteristics in different expression dimensions, and according to the lost customer characteristics, creating lost customer portraits; extracting historical data of the target customer from the customer data cabin, obtaining customer characteristics of the target customer in different expression dimensions through the historical data of the target customer, and creating a target customer portrait according to the customer characteristics; and obtaining the similarity between the lost customer portrait and the target customer portrait, and generating a target customer loss early warning according to a comparison result of the similarity and a similarity threshold. According to the invention, through identifying the lost clients and generating the loss early warning, the enterprise is helped to improve the client retention rate.

Description

Technical Field

[0001] The present invention relates to the field of data analysis, and in particular to an analysis method based on a customer data cabin. Background Art

[0002] In today's highly competitive market environment, customer churn has become one of the main challenges facing companies. Customer churn rates in many industries remain high, and companies have to face the risk of losing long-term customers. Customer churn not only leads to direct loss of revenue, but also may weaken brand loyalty, affecting the competitiveness of companies in the market.

[0003] Although many companies have established customer data cabins that integrate multi-dimensional data such as customer behavior, feedback, and service requests, their application in customer churn warning is still insufficient. Currently, many companies still rely on traditional methods for churn warning, such as regular customer satisfaction surveys and manual monitoring, which are often lagging and unable to capture signals of customer churn in real time. Summary of the invention

[0004] The present invention provides an analysis method based on a customer data cabin, which aims to assist enterprises in improving customer retention rate by identifying lost customers and generating lost warnings.

[0005] The analysis method based on the customer data cabin provided by the present invention comprises the following steps:

[0006] Setting a lost customer screening condition, wherein the lost customer screening condition is used to screen out lost customers from the customer data cabin;

[0007] Based on the lost customer screening condition, screen out lost customers from the customer data cabin, and extract historical data of lost customers from the customer data cabin;

[0008] Obtain the lost customer characteristics in different performance dimensions through the historical data of lost customers, and create lost customer portraits based on the lost customer characteristics;

[0009] Extracting historical data of target customers from the customer data cabin, obtaining customer characteristics of target customers in different performance dimensions through the historical data of target customers, and creating a target customer profile based on the customer characteristics;

[0010] The similarity between the lost customer portrait and the target customer portrait is obtained, and a target customer lost warning is generated according to a comparison result between the similarity and a similarity threshold.

[0011] Optionally, setting the lost customer screening condition comprises the following steps:

[0012] Based on the time threshold, a first lost customer screening condition is set, and the first lost customer screening condition satisfies the following characterization model: ,in, Indicates the last transaction time of customer c, represents the time threshold, Indicates the current time.

[0013] Optionally, setting the lost customer screening condition comprises the following steps:

[0014] Based on the transaction frequency threshold, a second lost customer screening condition is set, and the second lost customer screening condition satisfies the following characterization model: ,in, Indicates that customer c is active during The number of transactions within Indicates the length of time that customer c is active. Indicates the transaction frequency threshold.

[0015] Optionally, obtaining lost customer characteristics in different performance dimensions through historical data of lost customers includes the following steps:

[0016] Set multiple performance dimensions, any one of which is used to represent an attribute characteristic of a customer, and any two performance dimensions represent different attribute characteristics;

[0017] Set the measurement indicators for each performance dimension separately. The number of measurement indicators for any performance dimension can be one or more.

[0018] The historical data of lost customers are used to obtain the representation value of each metric indicator, and the characteristics of lost customers in different performance dimensions are obtained through one or more metric indicator representation values ​​corresponding to each performance dimension.

[0019] Optionally, creating a lost customer profile according to the lost customer characteristics comprises the following steps:

[0020] Summarize the one or more metric representation values ​​corresponding to all lost customers in different performance dimensions to obtain lost customer portraits, including the first lost customer sub-portrait, the second lost customer sub-portrait, and the mth lost customer sub-portrait;

[0021] Among them, the first churned customer sub-portrait is a set of measurement indicator vectors corresponding to all churned customers on the first performance dimension, the second churned customer sub-portrait is a set of measurement indicator vectors corresponding to all churned customers on the second performance dimension, and the mth churned customer sub-portrait is a set of measurement indicator vectors corresponding to all churned customers on the mth performance dimension, where m is the number of performance dimensions.

[0022] Optionally, obtaining the customer characteristics of the target customer in the different performance dimensions through the historical data of the target customer includes the following steps:

[0023] Utilize the historical data of target customers to obtain the representation value of each metric indicator, and obtain customer characteristics on different performance dimensions through one or more metric indicator representation values ​​corresponding to each performance dimension.

[0024] Optionally, creating a target customer profile according to the customer characteristics comprises the following steps:

[0025] Summarize the representation values ​​of one or more metric indicators corresponding to the target customers in different performance dimensions respectively to obtain the lost customer portrait, including the first target customer sub-portrait, the second target customer sub-portrait, and the mth target customer sub-portrait;

[0026] Among them, the first target customer sub-portrait is a measurement indicator vector corresponding to the target customer in the first performance dimension, the second target customer sub-portrait is a measurement indicator vector corresponding to the target customer in the second performance dimension, and the mth target customer sub-portrait is a measurement indicator vector corresponding to the target customer in the mth performance dimension.

[0027] Optionally, obtaining the similarity between the lost customer profile and the target customer profile comprises the following steps:

[0028] A similarity calculation model is constructed using the first churned customer sub-portrait, the second churned customer sub-portrait, the mth churned customer sub-portrait, the first target customer sub-portrait, the second target customer sub-portrait, and the mth target customer sub-portrait, and the similarity calculation model is used to obtain the similarity between the churned customer portrait and the target customer portrait.

[0029] Further optionally, the similarity calculation model is:

[0030] , ,in, Indicates the similarity between the lost customer profile and the target customer profile. i and s are both numbers. The value of i does not exceed m, where m is the number of performance dimensions. The value of s does not exceed S, where S is the number of lost customers. represents the i-th target customer sub-portrait, represents the metric vector corresponding to the sth lost customer in the ith lost customer sub-portrait on the ith performance dimension, represents the similarity between the metric vector corresponding to the sth lost customer in the ith target customer sub-portrait and the sth lost customer in the ith lost customer sub-portrait on the ith performance dimension, It represents the highest similarity value of the metric vector of lost customers in the ith target customer sub-portrait and the ith performance dimension. j is a number, and the value of j does not exceed n. n represents the number of metric indicators corresponding to the ith performance dimension. represents the jth metric representation value of the target customer on the i-th performance dimension, It represents the representation value of the jth metric of the sth lost customer on the ith performance dimension.

[0031] Optionally, generating a target customer churn warning according to a comparison result between the similarity and a similarity threshold comprises the following generation rules:

[0032] If the similarity exceeds the similarity threshold, a target customer churn warning is generated; if the similarity does not exceed the similarity threshold, the target customer is continuously monitored.

[0033] The analysis method based on the customer data cabin provided by the present invention has the following beneficial effects:

[0034] The present invention sets lost customer screening conditions to screen out lost customers, thereby laying the foundation for subsequent analysis; at the same time, by extracting performance characteristics on multiple performance dimensions from historical data of lost customers, it helps enterprises to deeply understand the reasons for loss; in addition, the present invention uses a similarity model to quickly and accurately determine the similarity between lost customer portraits and target customer portraits, and then generate a loss warning in time by comparing the similarity with a preset threshold.

[0035] The present invention enables enterprises to take proactive intervention measures to win back potential lost customers, improve customer retention rate, improve the accuracy of customer management, and enhance the enterprise's ability to respond in the market, ultimately achieving higher customer satisfaction and economic benefits.

[0036] The present invention can also enable enterprises to continuously optimize marketing strategies and service quality and establish a stable customer relationship network by continuously monitoring and analyzing customer data. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 A flow chart of an analysis method based on a customer data cabin provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0038] In one embodiment, see Figure 1 , Figure 1 A flow chart of an analysis method based on a customer data cabin provided in an embodiment of the present invention.

[0039] It should be noted that the customer data cabin described in the present invention is an integrated database system that collects, stores and manages customer-related data from different channels and systems.

[0040] like Figure 1 As shown, the analysis method based on the customer data cabin provided in the embodiment of the present invention includes the following steps:

[0041] S01. Setting lost customer screening conditions, wherein the lost customer screening conditions are used to screen out lost customers from a customer data cabin.

[0042] In this embodiment, the lost customers refer to customers who have not conducted any transactions or interactions within a certain period of time; further, the lost customer screening conditions are standards or indicators for identifying and judging which customers have been lost, which can be set by the enterprise according to actual conditions.

[0043] In this embodiment, the setting of lost customer screening conditions in step S01 includes the following steps: setting a first lost customer screening condition based on a time threshold.

[0044] Furthermore, the first lost customer screening condition satisfies the following characterization model: ,in, Indicates the last transaction time of customer c, Indicates the time threshold, Indicates the current time. That is, when the time interval between the last transaction time and the current time exceeds a certain time threshold, the customer is identified as a lost customer.

[0045] In one or some other embodiments, the setting of lost customer screening conditions in step S01 includes the following steps: setting a second lost customer screening condition based on a transaction frequency threshold.

[0046] Furthermore, the second lost customer screening condition satisfies the following characterization model: ,in, Indicates that customer c is active during The number of transactions within Indicates the length of time that customer c is active. Indicates the transaction frequency threshold.

[0047] S02: based on the lost customer screening condition, screen out lost customers from the customer data cabin, and extract historical data of lost customers from the customer data cabin.

[0048] In this embodiment, based on the first lost customer screening condition ( ), the set of lost customers selected from the customer data cabin satisfies the following characterization model: ,in, ' indicates the set of lost customers, Represents a customer, Represents the set of all customers.

[0049] In one or some other embodiments, based on the second lost customer screening condition ( ), the set of lost customers selected from the customer data cabin satisfies the following characterization model: ,in, Represents the set of lost customers. Represents a customer, Represents the set of all customers.

[0050] S03. Obtain lost customer characteristics in different performance dimensions through historical data of lost customers, and create lost customer portraits based on the lost customer characteristics.

[0051] It can be understood that the performance dimensions are categories used to measure and evaluate the characteristics or indicators of customers, products or services in specific aspects, which include but are not limited to transaction dimensions, interaction dimensions, satisfaction dimensions, value dimensions, etc., and any performance dimension includes at least one measurement indicator.

[0052] Specifically, the transaction dimension is used to measure the characteristics of customer transaction activities and consumption behaviors within a certain period of time, which can be measured through indicators such as total transaction amount, number of transactions, and average transaction amount; the interaction dimension is used to measure the interaction between customers and enterprises, including the degree and frequency of customer participation in various channels, which can be measured through indicators such as login frequency and page visits; the satisfaction dimension is used to evaluate customer satisfaction with the company's products or services, which can be measured through indicators such as satisfaction score, number of complaints, and willingness to recommend; the value dimension measures the economic value and potential value that customers create for the company, which can be measured through indicators such as customer lifetime value (CLV) and profit contribution.

[0053] Furthermore, the step S03 of obtaining lost customer characteristics in different performance dimensions through historical data of lost customers includes the following steps:

[0054] S0311. Set multiple performance dimensions, any one of which is used to represent an attribute characteristic of a customer, and any two performance dimensions represent different attribute characteristics.

[0055] In this embodiment, step S0311 is configured with m representation dimensions, which can be characterized as the following representation dimension set: ,in, represents the first performance dimension, represents the second performance dimension, Represents the mth performance dimension.

[0056] S0312. Set measurement indicators for each performance dimension respectively, and the number of measurement indicators for any performance dimension is one or more.

[0057] In this embodiment, a performance dimension in step S032 There are n metrics, which can be represented as the following metric set: , where i is a number whose value does not exceed m. represents the first metric of the i-th performance dimension, represents the second metric of the i-th performance dimension, Represents the nth metric indicator of the ith performance dimension. It should be noted that the number of metric indicators corresponding to any two performance dimensions may be the same or different.

[0058] Furthermore, any metric (where i is a positive integer less than n) corresponds to a specific metric acquisition model, such as the first metric of the first performance dimension is the transaction amount indicator of the transaction dimension, then , It means to sum up all transaction amounts of customers. Similarly, other metrics and their corresponding metric acquisition models need to be set according to the actual meaning of the set metrics.

[0059] S0313. Utilize the historical data of lost customers to obtain the representation value of each metric indicator, and obtain the lost customer characteristics on different performance dimensions through one or more metric indicator representation values ​​corresponding to each performance dimension.

[0060] Furthermore, the step S03 of creating a lost customer portrait based on the lost customer characteristics includes the following steps: respectively aggregating one or more metric characterization values ​​corresponding to all lost customers in different performance dimensions to obtain a lost customer portrait, including a first lost customer sub-portrait, a second lost customer sub-portrait, and an mth lost customer sub-portrait.

[0061] Specifically, the steps are as follows:

[0062] S0321. Summarize the measurement indicator vectors of all lost customers on the first performance dimension to obtain a first lost customer sub-portrait, where the first lost customer sub-portrait is a set of measurement indicator vectors corresponding to all lost customers on the first performance dimension.

[0063] In detail, the second churn customer sub-profile can be characterized as the following model: ,in, Represents the metric vector of the first lost customer on the first performance dimension, Represents the metric vector of the second lost customer on the first performance dimension, Represents the metric vector of the Sth lost customer on the first performance dimension.

[0064] More specifically, the metric vector of any lost customer on the first performance dimension = , is the number of metrics in the first performance dimension.

[0065] S0322. Summarize the measurement indicator vectors of all lost customers on the second performance dimension to obtain a second lost customer sub-portrait, where the second lost customer sub-portrait is a set of measurement indicator vectors corresponding to all lost customers on the second performance dimension.

[0066] Similarly, the second lost customer sub-profile can be represented by the following model: ,in, Represents the metric vector of the first lost customer on the second performance dimension, Represents the measurement indicator vector of the second lost customer on the second performance dimension, A vector of metrics representing the third lost customer on the second performance dimension.

[0067] Furthermore, the metric vector of any lost customer on the second performance dimension = , is the number of measurement indicators in the second performance dimension.

[0068] …

[0069] S032m. Summarize the measurement indicator vectors of all lost customers on the mth performance dimension to obtain the mth lost customer sub-portrait, where the mth lost customer sub-portrait is a set of measurement indicator vectors corresponding to all lost customers on the mth performance dimension.

[0070] Similarly, the mth lost customer sub-profile can be represented by the following model: ,in, represents the metric vector of the first lost customer on the mth performance dimension, represents the metric vector of the second lost customer on the mth performance dimension, Represents the metric vector of the third lost customer on the mth performance dimension.

[0071] Furthermore, the metric vector of any lost customer on the mth performance dimension is = , is the number of metrics in the mth performance dimension.

[0072] S04. Extract the historical data of the target customer from the customer data cabin, obtain the customer characteristics of the target customer in the different performance dimensions through the historical data of the target customer, and create a target customer portrait based on the customer characteristics.

[0073] In this embodiment, referring to the method for obtaining the lost customer portrait in step S03, step S04 can obtain the corresponding target customer portrait.

[0074] Furthermore, the step S04 of obtaining the customer characteristics of the target customer in the different performance dimensions through the historical data of the target customer includes the following steps: obtaining each metric characterization value by using the historical data of the target customer, and obtaining the customer characteristics in different performance dimensions through one or more metric characterization values ​​corresponding to each performance dimension.

[0075] Furthermore, creating a target customer portrait based on the customer characteristics as described in step S04 includes the following steps: respectively summarizing one or more metric representation values ​​corresponding to the target customers in different performance dimensions to obtain a lost customer portrait, including a first target customer sub-portrait, a second target customer sub-portrait, and an mth target customer sub-portrait.

[0076] The first target customer sub-portrait is a metric vector corresponding to the target customer in the first performance dimension, which can be represented as: ,in, Represents the target customer's metric vector on the first performance dimension, Indicates the first metric value of the target customer in the first performance dimension. Indicates the second metric value of the target customer on the first performance dimension. Indicates the target customer's first performance dimension. Metric value.

[0077] Similarly, the second target customer sub-portrait is the measurement indicator vector corresponding to the target customer in the second performance dimension, and its corresponding representation model can be inferred by analogy and will not be elaborated here; the mth target customer sub-portrait is the measurement indicator vector corresponding to the target customer in the mth performance dimension, and its corresponding representation model can be inferred by analogy and will not be elaborated here.

[0078] S05: Obtain the similarity between the lost customer portrait and the target customer portrait, and generate a target customer lost warning according to a comparison result of the similarity and a similarity threshold.

[0079] In this embodiment, the step S05 of obtaining the similarity between the lost customer profile and the target customer profile includes the following steps:

[0080] A similarity calculation model is constructed using the first churned customer sub-portrait, the second churned customer sub-portrait, the mth churned customer sub-portrait, the first target customer sub-portrait, the second target customer sub-portrait, and the mth target customer sub-portrait, and the similarity calculation model is used to obtain the similarity between the churned customer portrait and the target customer portrait.

[0081] Furthermore, the similarity calculation model is: , ,in, Indicates the similarity between the lost customer profile and the target customer profile. i and s are both numbers. The value of i does not exceed m, where m is the number of performance dimensions. The value of s does not exceed S, where S is the number of lost customers. represents the i-th target customer sub-portrait, represents the metric vector corresponding to the sth lost customer in the ith lost customer sub-portrait on the ith performance dimension, represents the similarity between the metric vector corresponding to the sth lost customer in the ith target customer sub-portrait and the sth lost customer in the ith lost customer sub-portrait on the ith performance dimension, It represents the highest similarity value of the metric vector of lost customers in the ith target customer sub-portrait and the ith performance dimension. j is a number, and the value of j does not exceed n. n represents the number of metric indicators corresponding to the ith performance dimension. represents the jth metric representation value of the target customer on the i-th performance dimension, It represents the representation value of the jth metric of the sth lost customer on the ith performance dimension.

[0082] It is understandable that the similarity threshold is a specific value used as a judgment standard in similarity comparison to identify potential customer churn risks, that is, by comparing the target customer with the churned customer's similarity to determine whether the target customer is likely to churn.

[0083] It is also understandable that the reasonable setting of its value directly affects the effect of customer relationship management. In actual situations, the similarity threshold is usually determined based on a combination of factors such as historical data analysis, statistical methods, business experience and model training.

[0084] In this embodiment, the step S05 generates a target customer churn warning according to the comparison result between the similarity and the similarity threshold, including the following generation rules:

[0085] If the similarity exceeds the similarity threshold, a target customer churn warning is generated; if the similarity does not exceed the similarity threshold, the target customer is continuously monitored.

[0086] Specifically, continuously monitor the behavior of target customers, such as transaction records, interaction frequency, satisfaction feedback and other data to determine their engagement and activity levels; further, dynamically update the performance dimensions and metrics in the customer portrait to ensure that the monitoring data reflects the customer's latest status; further, recalculate the similarity between the target customer and the churned customer portraits within a set time interval to determine whether the target customer is at potential risk of churn.

[0087] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described or recorded in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0088] It should be noted that the above embodiments can be freely combined as needed. The above are only preferred implementations of the present invention; it should be noted that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. An analysis method based on a customer data cabin, characterized in that: The steps include: Setting a lost customer screening condition, wherein the lost customer screening condition is used to screen out lost customers from the customer data cabin; Based on the lost customer screening condition, screen out lost customers from the customer data cabin, and extract historical data of lost customers from the customer data cabin; Obtain the lost customer characteristics in different performance dimensions through the historical data of lost customers, and create lost customer portraits based on the lost customer characteristics; Extracting historical data of target customers from the customer data cabin, obtaining customer characteristics of target customers in different performance dimensions through the historical data of target customers, and creating a target customer profile based on the customer characteristics; The similarity between the lost customer portrait and the target customer portrait is obtained, and a target customer lost warning is generated according to a comparison result between the similarity and a similarity threshold.

2. The analysis method based on the customer data cabin according to claim 1, characterized in that: The step of setting the lost customer screening condition comprises the following steps: Based on the time threshold, a first lost customer screening condition is set, and the first lost customer screening condition satisfies the following characterization model: ,in, Indicates the customer The last transaction time of represents the time threshold, Indicates the current time.

3. The analysis method based on customer data cabin according to claim 1, characterized in that: The step of setting the lost customer screening condition comprises the following steps: Based on the transaction frequency threshold, a second lost customer screening condition is set, and the second lost customer screening condition satisfies the following characterization model: ,in, Indicates that customer c is active during The number of transactions within Indicates the length of time that customer c is active. Indicates the transaction frequency threshold.

4. The analysis method based on customer data cabin according to claim 1, characterized in that: The method of obtaining lost customer characteristics in different performance dimensions through historical data of lost customers includes the following steps: Set multiple performance dimensions, any one of which is used to represent an attribute characteristic of a customer, and any two performance dimensions represent different attribute characteristics; Set the measurement indicators for each performance dimension separately. The number of measurement indicators for any performance dimension can be one or more. The historical data of lost customers are used to obtain the representation value of each metric indicator, and the characteristics of lost customers in different performance dimensions are obtained through one or more metric indicator representation values ​​corresponding to each performance dimension.

5. The analysis method based on customer data cabin according to claim 4, characterized in that: The step of creating a lost customer profile according to the lost customer characteristics comprises the following steps: Summarize the one or more metric representation values ​​corresponding to all lost customers in different performance dimensions to obtain lost customer portraits, including the first lost customer sub-portrait, the second lost customer sub-portrait, and the mth lost customer sub-portrait; Among them, the first churned customer sub-portrait is a set of measurement indicator vectors corresponding to all churned customers on the first performance dimension, the second churned customer sub-portrait is a set of measurement indicator vectors corresponding to all churned customers on the second performance dimension, and the mth churned customer sub-portrait is a set of measurement indicator vectors corresponding to all churned customers on the mth performance dimension, where m is the number of performance dimensions.

6. The analysis method based on customer data cabin according to claim 5, characterized in that: The step of obtaining the target customer's customer characteristics in the different performance dimensions through the target customer's historical data includes the following steps: Utilize the historical data of target customers to obtain the representation value of each metric indicator, and obtain customer characteristics on different performance dimensions through one or more metric indicator representation values ​​corresponding to each performance dimension.

7. The analysis method based on customer data cabin according to claim 6, characterized in that: The step of creating a target customer profile based on the customer characteristics comprises the following steps: Summarize the representation values ​​of one or more metric indicators corresponding to the target customers in different performance dimensions respectively to obtain the lost customer portrait, including the first target customer sub-portrait, the second target customer sub-portrait, and the mth target customer sub-portrait; Among them, the first target customer sub-portrait is a measurement indicator vector corresponding to the target customer in the first performance dimension, the second target customer sub-portrait is a measurement indicator vector corresponding to the target customer in the second performance dimension, and the mth target customer sub-portrait is a measurement indicator vector corresponding to the target customer in the mth performance dimension.

8. The analysis method based on customer data cabin according to claim 7, characterized in that: The obtaining of the similarity between the lost customer profile and the target customer profile comprises the following steps: A similarity calculation model is constructed using the first churned customer sub-portrait, the second churned customer sub-portrait, the mth churned customer sub-portrait, the first target customer sub-portrait, the second target customer sub-portrait, and the mth target customer sub-portrait, and the similarity calculation model is used to obtain the similarity between the churned customer portrait and the target customer portrait.

9. The analysis method based on customer data cabin according to claim 8, characterized in that: The similarity calculation model is: , ,in, Indicates the similarity between the lost customer profile and the target customer profile. i and s are both numbers. The value of i does not exceed m, where m is the number of performance dimensions. The value of s does not exceed S, where S is the number of lost customers. represents the i-th target customer sub-portrait, represents the metric vector corresponding to the sth lost customer in the ith lost customer sub-portrait on the ith performance dimension, represents the similarity between the metric vector corresponding to the sth lost customer in the ith target customer sub-portrait and the sth lost customer in the ith lost customer sub-portrait on the ith performance dimension, It represents the highest similarity value of the metric vector of lost customers in the ith target customer sub-portrait and the ith performance dimension. j is a number, and the value of j does not exceed n. n represents the number of metric indicators corresponding to the ith performance dimension. represents the jth metric representation value of the target customer on the i-th performance dimension, It represents the representation value of the jth metric of the sth lost customer on the ith performance dimension.

10. The analysis method based on customer data cabin according to claim 1, characterized in that: The generating of target customer churn warning according to the comparison result between the similarity and the similarity threshold comprises the following generating rules: If the similarity exceeds the similarity threshold, a target customer churn warning is generated; If the similarity does not exceed the similarity threshold, the target customer will continue to be monitored.

Images