KR20010109402A - System and method for measuring customer's activity value on internet - Google Patents

Abstract

The present invention relates to a system for calculating a value for customer behavior by assigning an appropriate value to each site that generates revenue directly or indirectly among various customer behaviors displayed on the Internet.

According to the present invention, a system for calculating a behavior value of a customer using a predetermined Internet site, comprising: a customer DB (database) management unit for collecting and managing information about the customer from the internet site, and the Internet from the customer DB management unit. Collecting information about the site and information about the customer so that the new customer value representing the value at the time when the customer joins the Internet site as a new customer and the customer if the customer is already subscribed to the Internet site; Define a value-occurring customer behavior that is applied to calculate a current customer value that represents a value at a current time of, and each value for the new customer value and the current customer value based on the defined value-occurring customer behavior. Value calculations, including means for creating output models A model generator-the value generating customer behavior is an action that can generate revenue in the internet site; and the data of the customer from the customer DB management unit is reflected in the value calculation model to reflect each customer value. A customer behavior value calculation system on the Internet is provided that includes a customer value calculator that calculates.

Description

SYSTEM AND METHOD FOR MEASURING CUSTOMER'S ACTIVITY VALUE ON INTERNET}

The present invention relates to a system for calculating the value of customer behavior on the Internet. In particular, the present invention relates to a system for calculating customer value by directly or indirectly generating profits for each site among various customer behaviors on the Internet. It's about a system that produces value.

Recently, with the rapid increase in Internet users, the importance of customer management and marketing techniques unique to the Internet service industry has emerged. Since the early days of the Internet, most Internet companies have tried to attract customers with a variety of services, such as free e-mail, community, free content delivery, and free software downloads.

Many of these customers visit sites on the Internet, leaving behind a variety of behaviors, each of which has a different value in contributing to the site, depending on who they are and what actions / reactions they have taken on which pages. For example, in a site with an average of thousands of new members registered each day, not all of those new members have the same value. Compared to the case where a teenage child registers as a new member on a e-commerce site that sells home appliances as a main product, and when the 40s head of household registers, from the site's point of view, the 40s head of household who usually has a higher purchasing power than the teenager You will be this worthy customer. For example, even if a teenager uses the site more often than the 40's most often, the site's primary marketed customer will be the 40's most customer than the teenager in that it is not the target customer of the site. Therefore, it is desirable that the target of the new customer attraction campaign of the site is focused on the most customers in their 40s rather than their teens.

As such, each customer has different characteristics, ranging from basics such as age, place of residence, gender, occupation, and marital status, to psychological matters such as hobbies, specialties, types of residence, and favorite food, information, and friends. Because of these different characteristics, the behavior on the site is also distinguished, so the customer value seen by Internet companies is different. The information needed to calculate customer value is derived from the information that the customer provides at the time of registration as a new member, and then from the various behavior histories the customer has visited the site, and is calculated based on the different behaviors that customers with different characteristics show in different situations. Customer value can be the basis for gaining satisfaction for both customers and the site by allocating marketing costs accordingly.

However, at present, most Internet sites do not have a concrete way of estimating customer value. In some sites, the customer value is calculated based on a scorecard determined by the behavior or response indicated by the customer, but only the customer value is roughly calculated. Therefore, there is a need for a method that can present criteria and techniques for more accurately estimating the value of each customer.

Accordingly, an object of the present invention is to systematically manage and utilize customer data for each customer, to arrange which of the various behaviors of customers appearing in various sites on the Internet generate revenue, and to quantify the revenue To provide a customer behavior value calculation system and method for calculating customer behavior value.

According to an aspect of the present invention, in the system for calculating the behavior value of a customer using a predetermined Internet site, the customer DB (database) management unit for collecting and managing information about the customer from the Internet site, the customer DB management unit Collecting information about the Internet site and information about the customer from the client, wherein the new customer value representing the value at the time when the customer joins the Internet site as a new customer, and the customer is already subscribed to the Internet site. Define a value-occurring customer behavior that is applied to calculate a current customer value indicative of the value at the current time of the customer, and based on the defined value-occurring customer behavior for the new customer value and the current customer value. Means for creating each value calculation model Value calculation model creation unit-The value generation customer behavior is an action that can generate revenue in the Internet site-and each of the customer by reflecting the data for the customer from the customer DB management unit in the value calculation model A customer behavior value calculation system on the Internet is provided that includes a customer value calculator that calculates value.

The value calculation model creation unit defines a value generating customer behavior that is applied to calculate a future customer value representing a value for the behavior that the customer is expected to show in the future, and based on the defined value generating customer behavior Preferably, the method further includes means for creating a value calculation model for the future customer value.

The value calculation model creation unit may further include a value behavior definition unit defining a value behavior for each customer value to be calculated by the customer value calculation unit, and a value calculation item that is a specific measure for quantifying the defined value behavior. A value calculation item defining unit defining a value, and a point value representing a value given by the Internet site for each of the defined value calculation items, and a degree to which each of the value calculation items is reflected in calculating the value of the customer. It is preferable to include a value calculation model generator which defines a weight to represent and creates a value calculation model.

In addition, when the new customer value is calculated, membership is defined as the value-generating customer behavior, and the value calculation item for the membership is an amount that is substantially paid by the customer to become a member on the Internet site. The demographic characteristics of the customer compared to the demographic distribution of the customer of the Internet site, the generality indicating the similarity, the accessibility indicating the availability and convenience of communication with the customer, the basic characteristics of the customer, the propensity, the It is advantageous to include at least one of an identity indicating whether or not a behavior pattern and the like within the Internet site are known, and prediction information predicting the behavior of the customer that cannot be directly calculated at the time of the new customer registration. .

The prediction information may be information calculated from a value calculation model for the future customer value.

In addition, the value-generating customer behavior is individually valued at each occurrence, and the individually calculated value may be integrated at a predetermined time to constitute the current customer value.

In the calculation of the future customer value, a predictive model for predicting the purchase possibility of the customer for a specific product / service or advertisement in the internet site, a product / service or advertisement in the internet site by the customer for a predetermined period, etc. The value is at least one of a prediction model for predicting the amount of purchase in the case of single or comprehensive purchase, and a prediction model for predicting the possibility of how long the customer will remain as a customer of the Internet site. It is preferable to define as a calculation item.

Each of the prediction models may be predicted based on existing data on a customer group having characteristics similar to those of the customers of the Internet site.

According to another aspect of the invention, the method for calculating the behavioral value of a customer using a predetermined Internet site, the method comprising the steps of: receiving data from the Internet site about a predetermined action that the customer visited the Internet site, Determining whether the customer is a new subscribed customer of the Internet site; when the customer is a new subscribed customer, calculating a new customer value indicating a value at the time when the customer subscribes to the Internet site as a new customer; If the customer is an existing customer, determining whether the predetermined action performed by the customer is set as a value-generating customer action that generates revenue on the Internet site, and associated with the predetermined action. Based on the preset value calculation model, the customer's present The method of calculating the value of customer behavior on the Internet, comprising the step of calculating the current value of the customer is provided showing the value of a point.

In this case, the method may further include calculating a future customer value indicating a value for an action that the customer is expected to show in the future, based on a value calculation model that is preset in relation to the predetermined action.

In addition, when it is determined that the predetermined behavior is a value generating customer behavior, it is preferable to further include updating the content of the value calculation model.

In addition, the step of calculating the new customer value, the step of detecting whether there is a value-generating customer behavior that is selected as the customer behavior that generates a new customer value in the behavior of the customer, is predetermined for the detected value-generating customer behavior Collecting customer data for each value calculation item that is present, and generating supplemental data that is additionally required to calculate the new customer value in addition to the collected customer data, and collecting the collected customer data and the supplemental data. And calculating the new customer value.

The value-generating customer behavior is a subscription, and the value calculation item is an amount of money actually paid by the customer to become a member of the Internet site, and demographic characteristics of the customer are demographics of the Internet site customer. Whether or not the generality of the similarity, the accessibility and the convenience of communication with the customer, the basic characteristics of the customer, the propensity, and the behavioral pattern in the Internet site can be known as compared with the geographical distribution. And at least one of an identity indicating the degree and prediction information predicting the behavior of the customer, which cannot be directly calculated at the time of joining the new customer.

The present customer value calculating step may include detecting a type of a value generating customer behavior that is preset as a customer behavior that generates the current customer value, and calculating a value that is preset for the detected value generating customer behavior. Collecting and generating customer data of each item and additionally required supplementary data, storing and managing a history of each value-producing customer behavior for the customer, and collecting the collected customer data, the supplementary data And calculating the current customer value using the customer history.

The calculating of the future customer value may include selecting a preset prediction model as a value calculation item for the future customer value, collecting data necessary for each of the selected prediction models, and additionally required supplementary data. Generating a value, reflecting the collected data and the supplementary data to the prediction model, calculating a predetermined prediction value from the prediction model, and calculating a future customer value using the calculated prediction values. It is desirable to.

In addition, the step of generating the predictive model, the step of predicting the purchase possibility of the customer for a particular product / service or advertisement, etc. in the Internet site, the customer is a product / service or advertisement in the Internet site for a predetermined period Predicting a purchase amount in the case of single or comprehensive purchase of the back and the like, and predicting the likelihood that the customer will remain as a customer of the Internet site for a certain period of time. It is advantageous.

According to another aspect of the present invention, a method for creating a predictive model for calculating future behavioral values for a customer using a predetermined Internet site, the method comprising: defining a subject to be predicted in the predictive model, the defined Defining a data item necessary for analyzing a subject, collecting data on the data item, refining and converting the data into a form suitable for analysis, and generating an analysis table by generating derivatives required for analysis And a method for creating a predictive model for calculating future behavioral values for a customer on an internet site comprising observing the analysis table and deriving an appropriate forecast item to generate a predictive model.

According to another aspect of the present invention, in a method for creating a value calculation model for calculating a value for a customer's behavior using a given Internet site, the value of the customer's behavior is determined in consideration of the characteristics of the Internet site. Defining a behavior to generate, defining a value calculation item that is a specific measure for quantifying the defined value action, and a point value representing the value assigned by the Internet site to each of the defined value calculation items And creating a value calculation model by defining a weight indicating a degree in which each of the value calculation items is reflected in calculating the value of the customer.

1 is a schematic diagram of a customer behavior value calculation system 200 in accordance with one embodiment of the present invention.

FIG. 2 is a flowchart of a process performed by the value calculation model preparation unit 213 of FIG. 1.

3 is a view schematically showing the configuration and operation of the customer DB management unit 220 of FIG.

4 is a flowchart showing the overall processing flow of the customer behavior value calculation system 200 according to the present invention.

5 is a flowchart illustrating a processing flow of a new customer value calculator according to an embodiment of the present invention.

FIG. 6 is a diagram illustrating value calculation items that can be defined for each of the paid member 600 and the free member 610 of FIG. 5 and unit items for each value calculated item.

FIG. 7 is a diagram showing the contents of customer data collected for each value calculation item in calculating a new customer value. FIG.

FIG. 8 is a diagram illustrating a point value designated for each item of FIG. 7. FIG.

9 is a flowchart illustrating a processing flow of a current customer value calculator according to an embodiment of the present invention.

10 is a diagram showing the contents of customer data collected for each value calculation item at the time of calculating the current customer value.

FIG. 11 is a diagram illustrating a point value designated for each item of FIG. 10. FIG.

12 is a flowchart specifically showing a processing flow in the predictive modeling step according to the present invention.

13 is a flow chart showing the processing flow of the future customer value calculation step in accordance with one embodiment of the present invention.

14 is a diagram illustrating a specific example of calculating a future value of a customer by applying a purchaseability prediction model and a residual probability prediction model.

15 is a diagram illustrating an example of calculating a customer future value using a purchase amount prediction model and a probability of survival prediction model.

<Explanation of symbols for the main parts of the drawings>

100a to c: Site 1 to 3

200: customer behavior value calculation system

211: customer value calculation unit

211a: New Customer Value Calculator

211b: Current customer value calculator

211c: Future Customer Value Calculator

213: Value calculation model preparation unit

213a: Value Behavior Justice

213b: Value calculation item definition unit

213c: value calculation model generator

220: customer DB management unit

220a: data collector

220b: Supplementary data generation and addition

220c: data analysis unit

221: integrated DB

Next, an embodiment of the present invention will be described in detail with reference to the accompanying drawings.

1 is a schematic diagram 200 of a customer behavior value calculation system in accordance with one embodiment of the present invention. As shown in the drawing, the customer behavior value calculation system 200 according to the present embodiment selects a customer DB manager 220 that manages data of each customer X using the Internet sites 100a, 100b, and 100c. Customer value calculation unit 211 that calculates customer value based on the established criteria, and a value calculation model generator that defines value behaviors for each type of customer value and creates a value calculation model for calculating customer value based on the defined value behavior (213). Here, the type of customer value means that the value of the customer can be calculated in various aspects according to the characteristics of each Internet site and the characteristics of the customer, and means a type that defines the value of each calculated customer. In this embodiment, the customer value types are classified into three types according to the customer life cycle: 1) New customer value, which calculates the value at the time customer X joins as a new customer, Existing Customer Value Current customer value, which yields behavior-specific values, and future customer value, which calculates value for behaviors that existing customers are expected to show in the future.

The customer value calculator 211 includes a new customer value calculator 211a, a current customer value calculator 211b, and a future customer value calculator 211c. The value calculated by these value calculators 211a, 211b, and 211c is combined to calculate the lifetime value of the customer, which means the total value generated while the customer (X) registers as a new customer and then exists as a site customer. It can also be calculated.

The value calculation model creation unit 213 is a value behavior definition unit 213a which defines the value behavior for each customer value to be calculated by the customer value calculation unit 211, and becomes a concrete measure for quantifying the defined value behavior. And a value calculation model generator 213c for defining a value calculation item, and a value calculation model generator 213c for creating a value calculation model for defining point values and weights of the defined value calculation items.

2 is a flowchart of a process performed by the value calculation model preparation unit 213 of FIG. First, select customer behaviors that generate customer value to be calculated (step 10). Here, value generating customer activity is an action that can generate revenue on the site in any manner by such an action and is selected according to the characteristics of each site. Value-generating customer behavior includes, for example, "sign up" for new customer value, "buy goods", "click on ads", "paid content clicks" for current customer value, or " Probable predictive model "," purchase predictive model ", and" preservable predictive model "(described below).

Subsequently, a specific value calculation item is selected to quantify each value generating customer behavior (step 20). For example, a specific value calculation item for the "membership" behavior may include a membership fee in the case of a paid membership and universal / identity / accessibility (described later). In addition, examples of the specific value calculation item for the "buy goods", "ad clicks", "paid content clicks" action may include the quantity / click counts, the unit price and the margin of the purchase of the goods, advertisements or contents. In the case of the "purchasing probability prediction model", the "purchasing predictive model", and the "survivability predictive model", each of the predicted values can be adopted as a specific value calculation item. These value calculation items can also be selected according to the characteristics of each site. For example, in a quiz site, the value calculation item may be a question of participating in a quiz, winning a quiz competition, or sharing a product experience opinion on a site that divides product information among consumers.

Next, a value calculation model is created by defining point values and weights for each value calculation item in consideration of characteristics of each site (step 30).

3 is a view schematically showing the configuration and operation of the customer DB management unit 220 of FIG. As shown, the customer DB management unit 220 includes a data collection unit 220a, supplemental data generation and addition unit 220b, and data analysis unit 220c. The data collection unit 220a collects data representing various types of information about the sites 1 to n, for example, customer data, history data, product / service / advertisement data, log data, and the like to build an integrated DB 221.

The supplementary data generation and addition unit 220b generates supplementary data additionally required to calculate the value for the customer X, in addition to the information held in the integrated DB 221, and also relates to the customer X. By adding prediction data predicted by the prediction model 1000 to be described later. Here, the supplementary data may include supplementary data due to data conversion and derivative data generation, and supplementary data using data obtained from an external research institution. Supplementary data resulting from data conversion and derivative data generation may be derived from, for example, information such as age, gender, etc., which may be derived from the customer's social security number, information such as residence, which may be derived from the postal code, and annual purchases. It may include data representing information such as monthly average purchase amount or average purchase amount over the last three months. In addition, supplementary data from data obtained from research institutes, etc. may be used for information such as Internet user behavior conducted by research institutes (ie, the general purpose of Internet use, behavior, areas of interest, personal computer possession status, Data indicating the number of times of e-commerce use, Internet satisfaction, etc.).

The data analysis unit 220c creates the data for each value calculation item of the value calculation model based on the integrated DB 221, the supplementary data, and the prediction data. The calculated item-specific data 222 may be used by the new customer value calculator 211a, the current customer value calculator 211b, and the future customer value calculator 211c of the customer value calculator 211.

4 is a flowchart showing the overall processing flow of the customer behavior value calculation system 200 according to the present invention. First, when customer X visits site 1 and performs action A, data about action A performed by customer X is received from each site (step 400). Then, it is determined whether the behavior A of the customer X is the new subscription behavior (step 410). If behavior A is determined to be a new subscription behavior, processing proceeds to calculating a new customer value at step 420. If it is determined in step 410 that it is not a new subscription behavior, it is determined whether customer X is an existing registered customer of site 1 (step 430). If it is determined in step 430 that the customer is not an existing customer, the process ends. If it is determined to be an existing customer, it is determined whether the action A performed by the customer X on the site 1 is set as a profit generating action on the site 1, that is, a value generating customer action (step 440). If it is determined in step 440 that action A is a value generating customer behavior, the process is terminated. If it is determined that value A is customer behavior, the content previously set in relation to action A is updated with the new data received in step 400 (step 450). Then, the content of the value calculation model associated with the updated value calculation item is updated (step 460). Processing then proceeds to a current customer value calculation step 470 or a future customer value calculation step 480. Optionally, lifetime customer value may be calculated using the new customer value, current customer value, and future customer value calculated in steps 420, 470, and 480, respectively (step 490).

Next, the new customer value calculation step, the current customer value calculation step and the future customer value calculation step of steps 420, 470 and 480 will be described in more detail.

<New customer value calculation stage>

5 is a flowchart illustrating the processing flow of the new customer value calculation step 420 according to the present invention.

First, it is detected whether any of the actions of the customer X is selected as a customer action that generates new customer value (step 421). In this embodiment, "membership" is selected as the value-generating customer behavior associated with the new customer value, which means that customer X joins as a new customer of site 1.

Next, with respect to the detected value generating customer behavior, customer data for each value calculation item is collected from the customer DB manager 220 (step 422). In the case of defining a subscription as a value-generating customer behavior as in the present embodiment, a value calculation item may be defined differently for a paid subscription and a free subscription. In this case, the contents of the value calculation item for the member registration behavior of the customer and the specific unit items under each item may be different according to the nature of the site and the purpose of the site. Therefore, hereinafter, only portions that can be commonly used regardless of the nature or purpose of the site will be described, and descriptions of portions that cannot be commonly used will be omitted.

FIG. 6 is a diagram illustrating value calculation items that can be defined for each of the paid member 600 and the free member 610 of FIG. 5 and unit items for each value calculated item. .

In the case of the paid membership 600, the amount 601 (for example, the amount of membership fee, fee, subscription fee, etc.) actually paid by the customer X to the site 1 may be defined as a basic value calculation item. In addition, it is also possible to add the universality, accessibility, identity (620, 630, 640) (to be described later) to be applied to the free member as a value calculation item.

In addition, regardless of the paid membership 600 / free membership 610, when calculating the new customer value, it is possible to apply the future value of the customer in addition to the basic new customer value. In other words, by applying the built-in prediction model based on the information provided by the customer at the time of signing up, each customer's likelihood of future value behaviors is calculated, combined with the basic new customer value. Can be calculated. However, in order to score (scoring) the possibility of expressing the future value behavior of new customers, a prediction model based on the existing customer data must be prepared first. Therefore, whether to apply the future value can be determined according to the site-specific situation.

In the case of free membership 610, universality 620, accessibility 630, and identity 640 are defined as value calculation items for quantifying the value of subscription behavior. These three items can vary in their applicability according to site-specific features and purposes.

Universality (620) is basically a measure of how similar the demographic characteristics of customer (X) are to the demographic distribution of the site's customers. Age 621, gender 622, region 623 and the like are defined as unit items. In other words, the age, gender, and location of customer (X) are compared with the distribution of customers across the site, so that the demographic characteristics of customer (X) are closer to the main customer base of the site, the higher the value of universality. To give. This is because if the customer (X) belongs to the main customer base of the site, the future activity on the site is expected to be more active than the non-main customer base. In particular, in the case of community or content-oriented sites where there are few or no direct revenue generations, there is a higher frequency of contributing to the site than when a new member is a main customer. It can be expected. In order to calculate the generality more precisely, each group item is subdivided again to form a large number of customer groups 624, and statistical analysis of the average behavior of each group for each customer group 624 statistically analyzed The universality of the new customer may be calculated based on the information 625. For example, after setting the average number of group memberships, the average number of visits to the site, and the number of content clicks for each customer group as specific unit items, the generality of new subscribers can be determined using the unit item value of the customer group to which the new customer belongs. Can be calculated. In addition, in the case of an e-commerce site, the act of simply registering itself may be considered to be of low value or, in some cases, of no value. Therefore, in this case, it is better to reduce the universality weight and increase the identity weight to be described later.

The accessibility 630 indicates whether communication with the customer X is possible and the degree of convenience. For example, the accessibility 630 may adopt communication channels such as a customer's e-mail, a phone number, and an address as a unit item. The weight given to each of these unit items may be different for each site. For example, a site in an Internet environment can typically be given the highest weight on e-mail holding (631), which is the most convenient and inexpensive channel for communicating with customers. It is much more expensive and can be given a lower weight than e-mail. However, for e-commerce sites, an address or phone may be more important than e-mail for order delivery.

The identity 640 indicates whether or not the basic characteristics of the customer X, the propensity, the behavioral pattern in the site, and the like can be known, and the basic information 641 and the profile information 642 about the customer X are known. It is adopted as a unit item. Here, the basic information 641 may indicate information about the age, gender, address, occupation, etc. of the customer, and the profile information 642 may indicate information about the hobby, lifestyle, interests, etc. of the customer.

The basic information 641 and the profile information 642 may adopt the number of the corresponding information items 641a and 642a and the fidelity of the information items (write or not) 641b and 642b as specific unit items, respectively.

In addition to the universality 620, accessibility 630, and identity 640, prediction information 650 predicting customer behaviors that cannot be directly calculated at the time of membership registration may be selectively applied. The prediction information 650 may include prediction values 651, which are obtained by predicting, through a built-in prediction model, value behaviors on the site that the customer will show after signing up, such as a response to a campaign, a product / service purchase, and an advertisement / content click. 652 may be included as a specific unit item, and may include information calculated in a future customer value calculation step 480 to be described later. To enable this, first, a prediction model is generated based on the behavior patterns of existing members, and a prediction value for future value behaviors of new members can be calculated based on the generated prediction model (described later). In addition, in addition to the prediction information 650, information indicating the possibility of departure of the customer may be added to refine the customer value, which depends on the environment of each site. The prediction information 650 may be defined as a value calculation item separate from the identity, or reflected in calculating the identity 640 by combining the basic information 641 and the profile information 642 at the time of membership. May be

Referring back to FIG. 5, in addition to the customer data collected in step 422, supplemental data is additionally required to calculate new customer value (step 423). Such supplementary data may be generated based on the collected customer data, or may be generated based on the data constructed in the integrated DB 221. The collected customer data and supplemental data are then reflected in the new customer value calculation model to calculate new customer value (step 424). The process then ends or proceeds to steps 470, 480, and / or 490.

FIG. 7 is a diagram showing the content of customer data collected for each value calculation item when calculating new customer value, and FIG. 8 illustrates a point value designated according to the content of item data for each item of FIG. 7. It is a figure. As shown in FIG. 7, information on the customer provided for each item in FIG. 7 is converted into a point value specified in the new customer value point value designation table of FIG. 8, and each value is calculated for the converted point value. The new customer value 700 is calculated by adding up the weighted values that are preset for each item as a whole. At this time, the summed final values can be adjusted within the standard range (eg 0-100). Here, the point value represents the value given to each site for each item, and the weight represents the degree to which each item is reflected in calculating the customer value.

<Current customer value calculation stage>

9 is a flowchart illustrating a processing flow of a current customer value calculator according to an embodiment of the present invention.

First, the type of value generating customer behavior that is currently selected as a customer behavior generating customer value is detected (step 471). Value-generating customer behavior related to current customer value is a value that is directly or indirectly valued from the site's point of view, among the various and complex behaviors or responses that an existing customer may display within the site until they visit the site on the Internet. Means. In this case, it is obvious that value-generating customer behavior may vary from site to site, because customer behavior that is not valuable on other sites may be valuable on certain sites, depending on the nature or purpose of the site. However, for the sake of convenience of explanation, the following will describe a part which can be commonly evaluated by general Internet sites.

In this application, value-generating customer behavior currently applied in calculating customer value is selected by way of example, product purchases, ad clicks, and paid content clicks (see FIG. 10). Individual value can be estimated. In each value action, the individual value calculated is integrated to a certain point in time to form the customer's present value.

On the other hand, the value-generating customer behavior that is currently applied in the calculation of customer value, which is not a direct product purchase or click, but a lead generation, such as a request for additional information about the product and an inquiry, etc. Actions may also be adopted. In this case, the predictive model can be used to calculate the current customer value by calculating the likelihood that the customer's current response will lead to future quantifiable behavior.

Next, looking at the value calculation items for each of the value calculation customer behavior, in the case of product purchase, the actual amount of revenue in consideration of the purchase price and cost, and in the case of an ad click, the operating cost at the price per click set by the site at the time of advertisement attraction In addition, the unit price per click may be adopted, and in the case of paid content click, the unit price per click considering the operation cost may be adopted in addition to the unit price per click of paid content.

Referring back to FIG. 9, customer data of each value calculation item for the value-occurring customer behavior detected in step 471 is collected (step 472), additionally generated supplementary data required (step 473), and each value The history of occurrence customer behavior is stored and managed by customer (step 474). Here, the history of the customer behavior is a variety of information that can occur during each customer behavior, for example, in the case of a product purchase, the product purchase time, product type, product amount, product receiving means and point value given to each of them It means information such as.

When such a history is accumulated, the customer present value at a certain reference point can be calculated, which is calculated by integrating the three occurrences of the value behavior among all transactions occurring after registering as a new customer. For example, site-specific analytical purposes and applications, such as average value behavior scores (values) over the past three months, total value behavior scores from the time of the new enrollment up to now, or total value behavior scores as monthly averages. By calculating the present value of the customer.

The collected customer data, supplemental data, and customer history are then reflected in the current customer value calculation model to calculate the current customer value (step 475). The process then ends or proceeds to step 490.

FIG. 10 is a diagram illustrating the contents of customer data collected for each value calculation item at the time of calculating the current customer value, and FIG. 11 is a diagram illustrating a point value designated for each item of FIG. 10. In FIG. 10, the information on the customer provided for each item is converted into a point value defined in the new customer value calculation model of FIG. 11, and a weight preset for each value calculation item in each converted point value. The sum of these values is added together to yield the current customer value. At this time, the summed final values can be adjusted within the standard range (eg 0-100). In this case, the point value may be defined as the sales amount itself, but based on the actual revenue allocated to the site, a more accurate present value may be calculated.

<Future customer value calculation stage>

Future customer value is the present value of the expected future expenditure or revenue of the customer (X), where present value is the value applied at an appropriate discount rate that reflects the value of time on the money the site sees. it means.

As a value-generating customer behavior that is applied when calculating future customer value, one embodiment of the present invention relates to the possibility of future occurrences for three value actions (product purchase, ad click, paid content click) that are applied when calculating current customer value. The degree is predicted and used as future spending or revenue.

The items used to predict the future likelihood and extent of these customer value behaviors are: whether customer (X) will purchase additional goods / services in the future, if so, what products will be purchased, and what is the total purchase estimate? The average monthly purchase estimate will vary and how long will the customer be a customer of the site (customer retention). These value calculations are based on existing customer data, and predictive future values can be obtained through predictive models.

Predictive models related to future customer value calculation can be divided into two types according to the characteristics of the predicted values to be predicted. One is to predict whether the customer behaves, such as whether a particular product has been purchased or the site has left, etc., which is commonly called classification prediction. The other is a value prediction that predicts the total amount of the product purchase of the customer in the future (Value Prediction, Regression, Estimation). Creating a model for classification and value prediction is called predictive modeling, which is done using data mining techniques. More specifically, the model is generated using decision tree, neural network, logistic regression, etc. for classification prediction, and the model is generated using regression, neural network, etc. for value prediction. do. The decision tree technique is a tree representation of the decision making process. It is a classification prediction / fragmentation modeling technique that divides the entire target group into several subgroups. The neural network technique mimics the neural networks of the human brain to A modeling technique that finds patterns hidden in data through an iterative learning process from possessed data, and logistic regression analysis obtains an event of interest as a function of one or more input variables, that is, a binary or ordinary goal. Modeling technique for estimating / predicting the probability of a variable.

The method of predicting future customer behavior may include a predictive modeling step and a model application step. The predictive modeling step is a step of generating a model based on the data including the content to be predicted, and the model application process is a model generated in the predictive modeling process on the data that has no content to obtain, that is, the target data to be predicted. Apply to predict future behavior

Hereinafter, the predictive modeling step will be described in more detail with reference to FIG. 12.

12 is a flowchart specifically showing a processing flow in the predictive modeling step according to the present invention. First, an analysis topic such as whether to predict whether to purchase or not to click on an advertisement is defined (step 120). Once the subject of analysis is defined, the data needed to analyze the subject is defined and data is obtained from the data source in which the data exists (step 121). At this time, the data source is a data warehouse or integrated DB 221 including data 121a such as basic information, profile information, history information (campaign response history, purchase history, advertisement / content click history) of existing customers, and the like. Can be.

Data obtained from various data sources is purified and converted into a form suitable for analysis (step 122). More specifically, in order to create a sophisticated predictive model, not only the integration of relevant raw data but also an outlier (a value outside the range or category of a data item, for example a gender item other than female / male) or Processing such as missing values (values that are not present in the data item, for example, when a job item field is not filled in) may be required. Therefore, the analysis table is created by transforming the data according to the business logic or generating derived variables that are considered necessary for analysis. Here, the business logic means a business, process or data rule used for each business. For example, in the case of a site where only a female customer is a member, when the seventh digit of the social security number is 1, an error occurs in the site business rule. Is processed. Derived variable means a new item created by modifying / combining the standard data items, and may include items such as site visit experience, average monthly visits, average monthly purchase amount, and age group. Including these derived variables, you can create an analysis table like, for example, Table 1.

The analysis table thus prepared is then observed and the appropriate predicted item is derived to generate a model (step 123). In this step, the prepared analysis table is used to determine which items are related to the item to be predicted and the degree of the related items, and generate a model by applying one or more algorithms. The algorithm used is appropriately selected in terms of analysis subject, data and other relevant hardware or software (tools). If you can use multiple algorithms, you can create multiple models with different algorithms. In addition, even when using one algorithm, several models are created as the algorithm adjusts various options and items.

The models thus generated are compared and evaluated according to various measures, and finally the model to be used for prediction is selected (step 124). Commonly used measures for model comparison and evaluation include classification accuracy and lift and gain charts. The classification accuracy indicates how closely the classification predicted by the model matches the actual classification, and the lift chart is a chart indicating the fit provided by the model. Model comparison and evaluation compares not only the predictive power between models, but also the effects within and without the model itself. Accordingly, the effect can be predicted when using the model, and the final model is selected in consideration of the predicted effect and convenience of use. At this time, if the model is inappropriate or more sophisticated to use as the final model of the generated model to return to the appropriate or necessary of the previous four steps (120, 121, 122, 123) and repeat the operation. The customer, product and service and transaction data of the Internet site are constantly changing. This new model of predictive modeling requires repetitive refinement because new customers come in, some of the existing customers opt out, update information, and visit the site every day and behave differently from yesterday.

Next, the step of calculating the future customer value by applying the model created in the predictive modeling step as described above will be described.

13 is a flowchart showing the processing flow of the future customer value calculation step (step 480 of FIG. 4) according to an embodiment of the present invention. First, a predictive model selected as a value calculation item related to future customer value is selected (step 481). In one embodiment of the present invention, three types of predictive models are adopted as value calculation items related to future expenditures or revenues of customers: 1) predictability models for purchase of specific goods / services or advertisements; A) a single or comprehensive model for predicting purchases, such as goods / services or advertisements, over a period of time;

The likelihood of purchasing model predicts, for example, an additional purchasing model that indicates which customers with additional characteristics make additional purchases and those with other characteristics based on existing customer data. After generating, scores of future purchases by a specific customer are scored to determine that customers with high predicted scores are more likely to purchase. In this case, the additional purchase product or service is usually set as a specific item and then the purchase possibility is predicted.

The purchase amount prediction model is a model that calculates the future value of the customer based on the total amount of the customer's forecast purchases within a certain period. Even in this case, based on the existing customer data, after generating a model indicating how much (amount) when a customer who has a certain characteristic additionally purchases a specific product, the model predicts the future of the specific customer. Score the purchase amount.

The likelihood of retention (maintenance, drop-off) predictive model quantifies whether a customer will be a customer of a site for a period of time. As with other forecasting models, you can predict the future site survival of a particular customer based on existing customer data.

Referring back to FIG. 13, the data needed to calculate the predictive value in each prediction model selected in step 481 is collected (step 482), and supplemental data other than the collected data is generated (step 483), and the collected data And supplemental data are applied to each prediction model to calculate each prediction value (step 484). Subsequently, the calculated future predictions are reflected in the future customer value calculation model to calculate future customer values (step 485). The process then ends or proceeds to step 420 or step 490.

FIG. 14 is a diagram illustrating a specific example of calculating a future value of a customer by applying a purchaseability prediction model and a residual probability prediction model, and FIG. 14A illustrates a product purchase possibility and an advertisement click possibility for each customer. FIG. 14 is a diagram illustrating an example of calculating a future value of a customer by using predicted values for clickable content and remaining payability, and FIG. 14B is used to calculate a total point value in FIG. It is a figure which calculates the score about the individual goods which become. As shown in the drawing, each point value is calculated by multiplying a price and a profit margin for each item, and the total point value is obtained by summing values obtained by multiplying each point value by a purchase possibility for the item. At this time, in addition to the point value and the purchase possibility for the item, a weight set in advance for the item may be applied to the total point value. The future value for the customer is calculated by multiplying the total point value thus obtained by the customer's possibility of remaining.

15 illustrates an example of calculating a customer future value by using a purchase amount prediction model and a residual probability prediction model. In this case, the future value of the customer is calculated by multiplying the total purchase amount predicted for the customer X by the remaining probability.

As such, the future value of the customer may be calculated by appropriately reflecting the prediction values through various prediction models.

By using the customer behavior value calculation system according to the present invention as described above, the present invention can be utilized as a basis for deriving an appropriate customer management method for each customer value from a customer perspective. In other words, the same 20:80 rule that 20% of top customers account for 80% of total profits can be applied to Internet companies as well as offline companies. Therefore, profits can be maximized by appropriately allocating marketing activities and costs. . Therefore, for high value customers, it is possible to induce customer satisfaction by providing incentives through the development of a continuous loyalty management program, and to induce continuous value generation by maintaining an active customer relationship by providing it as an active departure prevention program. In addition, low-value customers can focus on aggressive marketing strategies that can increase customer value by increasing value, such as cross-selling or up-selling.

Claims (19)

In a system for calculating the behavioral value of a customer using a predetermined Internet site, Customer DB (database) management unit for collecting and managing information about the customer from the Internet site, Collecting information about the Internet site and information about the customer from the customer DB management unit, the new customer value indicating the value at the time when the customer joins the Internet site as a new customer and the customer has already Define a valued customer behavior that is applied to calculate a current customer value that is indicative of the value at the current time of the customer when subscribed, and based on the defined valued customer behavior, the new customer value and the current A value calculation model generator comprising means for creating a respective value calculation model for customer value, wherein the value generating customer behavior is an action that can generate revenue on the internet site; and A customer value calculator that calculates each customer value by reflecting data about the customer from the customer DB manager in the value calculation model Customer behavior value calculation system on the Internet comprising a. The method of claim 1, wherein the value calculation model creation unit defines a value-generating customer behavior that is applied to calculate a future customer value representing a value for the behavior that the customer is expected to show in the future, and the defined value And a means for creating a value calculation model for the future customer value based on the occurrence customer behavior. The said value calculation model creation part of Claim 1 or 2, A value behavior definition unit that defines a value behavior for each customer value to be calculated by the customer value calculation unit, A value calculation item definition unit that defines a value calculation item that is a specific measure for quantifying the value behavior defined above, and For each of the defined value calculation items, a value calculation model is created by defining a point value indicating a value given by the Internet site and a weight indicating a degree to which each of the value calculation items is reflected in calculating the value of the customer. Value calculation model generator Customer behavior value calculation system on the Internet comprising a. The method of claim 1 or 2, wherein, in calculating the new customer value, membership is defined as the value inducing customer behavior. The value calculation item for the membership, The amount that the customer substantially pays the internet site, The generality of the demographic characteristics of the customer compared to the demographic distribution of the Internet site customer, Accessibility indicating whether communication with the customer is possible and the degree of convenience, An identity indicating whether and how much of the customer's basic characteristics, dispositions, and behavioral patterns within the Internet site are known, and Prediction information predicting the behavior of the customer that cannot be calculated directly at the time of the new customer registration A customer behavior value calculation system on the internet comprising at least one of the following. The system of claim 4, wherein the predictive information is information derived from a value calculation model for the future customer value. The method of claim 1 or 2, wherein the value generating customer behavior is individually valued at each occurrence, The individually calculated value is integrated at a given point in time customer behavior value calculation system on the Internet that constitutes the current customer value The method of claim 2, wherein in calculating the future customer value, A predictive model for predicting the purchase possibility of the customer for a specific product / service or advertisement in the Internet site, A predictive model for predicting a purchase amount when the customer purchases a product / service or advertisement in the Internet site for a predetermined period in a single or comprehensive manner, and Prediction model for predicting the likelihood of how long the customer will remain as a customer of the Internet site. Customer behavior value calculation system on the Internet, wherein at least one of the models is defined as the value calculation item. The method of claim 7, wherein each prediction model, A system for calculating a value of customer behavior on the Internet, which is predicted based on existing data on a customer group among the customers of the Internet site having characteristics similar to the customer. In the method of calculating the behavioral value of a customer using a predetermined Internet site, Receiving data from the internet site about a predetermined action that the customer visits the internet site, Determining whether the customer is a new subscriber of the internet site, Calculating a new customer value representing the value at the time when the customer joins the Internet site as a new customer if the customer is a new subscriber; If the customer is an existing customer, determining whether the predetermined action performed by the customer is set as a value-generating customer action that generates revenue on the Internet site; and Calculating a current customer value representing a value at the present time of the customer based on a value calculation model preset in relation to the predetermined behavior; Method of calculating the value of customer behavior on the Internet comprising a. 10. The method of claim 9, further comprising calculating future customer values indicative of the value for the behavior that the customer is expected to show in the future, based on a value calculation model that is predetermined in relation to the predetermined behavior. How to calculate the value of customer behavior on the Internet. The method according to claim 9 or 10, further comprising updating the contents of the value calculation model when it is determined that the predetermined behavior is a value generating customer behavior. The method of claim 9 or 10, wherein the new customer value calculation step, Detecting whether there is a value generating customer behavior selected as the customer behavior generating new customer value among the customer's actions; Collecting customer data for each of the predetermined value calculation items for the detected value-occurring customer behavior, and generating supplemental data additionally required to calculate the new customer value in addition to the collected customer data. , And Calculating the new customer value using the collected customer data and the supplemental data Method of calculating the value of customer behavior on the Internet comprising a. 13. The method of claim 12, wherein the value generating customer behavior is membership The value calculation item, The amount that the customer substantially pays the internet site, The generality of the demographic characteristics of the customer compared to the demographic distribution of the Internet site customer, Accessibility indicating whether communication with the customer is possible and the degree of convenience, An identity indicating whether and how much of the customer's basic characteristics, dispositions, and behavioral patterns within the Internet site are known, and Prediction information predicting the behavior of the customer that cannot be calculated directly at the time of the new customer registration A method of calculating customer behavior value on the Internet, comprising at least one of the following. The method of claim 13, wherein the predictive information is information calculated from a value calculation model for the future customer value. The method of claim 9 or 10, wherein the current customer value calculation step, Detecting a type of value generating customer behavior that is preset as the customer behavior that generates the current customer value, Collecting and generating customer data and additionally required supplementary data of each value calculation item preset for the detected value-generating customer behavior; Storing and managing a history of each value-bearing customer behavior for the customer, and Calculating current customer value using the collected customer data, the supplemental data and the customer history Method of calculating the value of customer behavior on the Internet comprising a. The method of claim 10, wherein the step of calculating the future customer value, Selecting a preset prediction model as a value calculation item for the future customer value, Collecting data necessary for each of the selected prediction models, and further generating supplementary data required; Reflecting the collected data and the supplementary data to the prediction model, calculating a predetermined prediction value from the prediction model, and Calculating future customer value using the calculated predictions Method of calculating the value of customer behavior on the Internet comprising a. The method of claim 16, wherein the step of generating the prediction model comprises: Predicting the likelihood of purchase of the customer for a specific product / service or advertisement in the Internet site, Estimating a purchase amount when the customer purchases a single or comprehensive product / service or advertisement in the Internet site for a predetermined period of time, and Predicting the likelihood of how long the customer will remain as a customer of the Internet site And at least one step of the method. In a method of creating a predictive model for calculating future behavioral values for a customer using a given internet site, Defining a topic to be predicted in the prediction model, Defining a data item necessary for analyzing the defined subject, collecting data for the data item, Purifying the data into a form suitable for analysis, generating derivative variables required for analysis, and preparing an analysis table; and Observing the analysis table and deriving an appropriate prediction item to generate a prediction model How to create a predictive model for calculating the future behavioral value for the customer on the Internet site comprising a. In the method of creating a value calculation model for calculating the value of the behavior of a customer using a given Internet site, Taking into account the characteristics of the Internet site, defining behaviors that generate value among the actions of the customer, Defining a value calculation item that is a specific measure for quantifying the value behavior defined above, and For each of the defined value calculation items, a value calculation model is created by defining a point value indicating a value given by the Internet site and a weight indicating a degree to which each of the value calculation items is reflected in calculating the value of the customer. Steps to How to create a value calculation model that includes.