WO2015056360A1

WO2015056360A1 - Customer-data analysis/evaluation system

Info

Publication number: WO2015056360A1
Application number: PCT/JP2013/081285
Authority: WO
Inventors: 増田　宗昭
Original assignee: カルチュア・コンビニエンス・クラブ株式会社
Priority date: 2013-10-16
Filing date: 2013-11-20
Publication date: 2015-04-23
Also published as: KR20160071990A; TWI598834B; HK1211719A1; JP2015079331A; CN104813315A; CN104813315B; TW201516927A; JP6059123B2

Abstract

[Problem] The provision of a system that uses a cycle consisting of data accumulation, analysis, action, and effect measurement, as follows: a variety of methods are used to analyze data provided by customers; on the basis of the results of those analyses, actions such as marketing campaigns and recommendations are performed with respect to the customers and actions such as data provision are performed with respect to coalition companies; and the degrees to which said actions were effective are evaluated. [Solution] This customer-data analysis/evaluation system (1) has the following: a means for collecting itemized data on customers of coalition companies; a means for accumulated said itemized data; an analyzing means for analyzing the itemized data; a means for classifying the customers into segments; and an evaluating means for identifying, from the items constituting the itemized data for customers in one segment and the items constituting the itemized data for customers in another segment, important items that affect differences in buying behavior.

Description

Customer data analysis and verification system

The present invention collects and accumulates data related to customers, analyzes the data, and plans and executes various actions such as promotion and recommendation from the analysis result, and can verify the effect of the executed action. It relates to customer data analysis and verification system.

Collect and store data provided by customers who are subscribers of the service point system at a specific company, analyze it, and provide advertisements and samples of the company's products and services (hereinafter referred to as products). It is known to be used for promotions such as the provision of goods and recommendations such as issuance of coupons (cited documents 1 to 5). Until now, however, the analysis and use of such customer data has been confined to a single company. Promotions and recommendations were made to an unspecified number of customers, assuming a customer image. In addition, there is no system to verify in detail whether or not the effects of promotions and recommendations have actually been achieved, and the only way to know the degree of effect is through the transition of monetary events such as how much the sales amount of the target product has increased. there were.

JP 2010-211687 A JP 2009-163533 A JP 2012-247926 A JP 2004-70504 A JP 2012-190061 A

Therefore, the present invention analyzes the data provided by the customer who is a subscriber of the service point system by various methods, and from the analysis result, the customer performs an action such as promotion, recommendation, etc. Plan actions such as provision, execute action provision, verify how effective this action was, and further improve the action, plan action, execute action, verify action, improve action An object is to provide a customer data analysis / verification system having the following cycle.

A customer data analysis / verification system according to the present invention includes means for collecting customer detailed data in an alliance company, means for accumulating the detailed data, analyzing means for analyzing the detailed data, and means for classifying customers into segments. And verification means for determining requirement items that affect the difference in purchasing behavior from items of customer detail data classified into one segment and items of customer detail data classified into another segment It is characterized by that.

According to another aspect of the present invention, the verification means includes items having high non-commonness among items of customer detail data classified into one segment and items of customer detail data classified into another segment. Is a requirement item that affects the difference in purchasing behavior.

According to another aspect of the present invention, the verification means is behavior analysis means for determining a requirement item from an item of history data among items of detailed data.

According to another aspect of the present invention, the verification means is an understanding analysis means for determining a requirement item from an item of research data among items of detailed data.

According to another aspect of the present invention, there is provided an action creation means for listing the requirement items determined by the verification means, and an action execution means for providing the listed requirement items to a company terminal.

According to another aspect of the present invention, there is provided an action creation means for creating a recommendation including the requirement item determined by the verification means, and an action execution means for providing the recommendation to the store terminal and / or the customer terminal. And

According to another aspect of the present invention, there is provided an action creation means for creating a promotion including the requirement item determined by the verification means, and an action execution means for providing the promotion to the store terminal and / or the customer terminal. And

According to another aspect of the present invention, there is provided a means for comparing the detailed data before the execution of the action and the detailed data after the execution of the action, and determining a customer who has caused the purchase behavior after the action.

According to the customer data analysis / verification system of the present invention, it is possible to accurately analyze data provided by a customer and create and execute an action from the analysis result.

It is a block diagram of a customer data analysis / verification system of the present invention. It is a block diagram of a customer data analysis / verification system of the present invention. It is a block diagram of a customer data analysis / verification system of the present invention. It is a block diagram of a customer data analysis / verification system of the present invention. It is a block diagram of a customer data analysis / verification system of the present invention. It is a block diagram of a customer data analysis / verification system of the present invention. It is a flowchart of the customer data analysis and verification system of this invention. It is a flowchart of the customer data analysis and verification system of this invention. It is a flowchart of the customer data analysis and verification system of this invention. It is a flowchart of the customer data analysis and verification system of this invention. It is a schematic diagram of the customer data analysis and verification system of the present invention. It is a schematic diagram of the customer data analysis and verification system of the present invention. It is a schematic diagram of the customer data analysis and verification system of the present invention. It is a schematic diagram of the customer data analysis and verification system of the present invention. It is a schematic diagram of the customer data analysis and verification system of the present invention. It is a schematic diagram of the customer data analysis and verification system of the present invention. It is a schematic diagram of the customer data analysis and verification system of the present invention. It is a schematic diagram of the customer data analysis and verification system of the present invention. It is a schematic diagram of the customer data analysis and verification system of the present invention. It is a schematic diagram of the customer data analysis and verification system of the present invention. It is a schematic diagram of the customer data analysis and verification system of the present invention. It is a flowchart of customer DNA analysis. It is a block diagram of customer DNA analysis. It is a schematic diagram which shows the processing content of the prediction process of customer DNA analysis. It is a schematic diagram which shows the processing content of the prediction process of customer DNA analysis. It is a schematic diagram which shows the processing content of the prediction process of customer DNA analysis. It is a schematic diagram which shows the processing content of the prediction process of customer DNA analysis. It is a schematic diagram which shows the processing content of the prediction process of customer DNA analysis. It is a schematic diagram which shows the processing content of the prediction process of customer DNA analysis. It is a schematic diagram which shows the processing content of the prediction process of customer DNA analysis. It is an example of the data table of the state in which the performance data of the customer entered. It is an example of the data table of the state containing the probability value after probability reasoning. It is an example of the data table of the state in which the map estimated value derived | led-out from the probability value after probability reasoning was contained.

DESCRIPTION OF SYMBOLS 1 Customer data analysis and verification system 2 Management company system 22 Analysis means 23 Questionnaire means 24 Recommendation means 3 Net alliance system 4 Real alliance system 5 Customer terminal 6 Store terminal 7 Corporate terminal

(System configuration)
The customer data analysis / verification system 1 of the present invention will be described below with reference to the drawings. A customer data analysis / verification system 1 (hereinafter, simply referred to as “system 1”) according to the present invention includes an operation company system 2, an operation company system 2 and information that are composed of arithmetic devices, databases, etc. under the jurisdiction of the operation company. Each of each of a plurality of alliance companies connected in a communicable manner with a computing device, a network alliance system 3 composed of a computing device of a plurality of alliance companies and a computing device of a management company, and a management company system 2 It has a real alliance system 4 composed of a store terminal (POS) 6 installed in an actual store, a computing device of an operating company, and a questionnaire means 5.

The outline of the configuration of the system 1 will be described with reference to FIG. 1 and FIG. The system 1 of the present invention gives service points that can be exchanged for products and services to customers and uses the service points, and from the customers via the network alliance system 3, the real alliance system 4, the questionnaire means 5, etc. Collect various data and store it in the database. Then, the accumulated data is analyzed, and segmentation is performed for classifying customers into arbitrary groups as shown in FIG. As shown in FIG. 3 and FIG. 4, recommendations such as distribution of POS coupons and advertisement e-mails suitable for each customer, TV advertisements, newspaper advertisements, social networking, advertisement display on member login screen, promotion of free samples, etc. As shown in FIG. 5 and FIG. 6, various actions for customers or companies such as provision of data for product development and product introduction data for companies are executed, and It is a system that verifies the effect and improves the content of the action.

First, an example of data provided from the customer to the operation system 2 will be described with reference to FIG. The data stored in the operation system 2 from the customer includes attribute data such as basic attributes of the customer such as sex and age provided by the customer when the system 1 is registered for use. Second, every time a customer purchases a product, etc. at an alliance company, it is collected and communicated to the operation system 2, such as the product, product name, visit time zone, historical data such as the store used, and the customer's intention Surveys via e-mail, surveys conducted through two-way communication of viewing content provided to customers, monitoring, interviews, etc., collected and transmitted to the management company system 2, whether there are marriages, children, and type of residence Detailed data including research data such as annual income, driver's license, and customer orientation. It should be noted that network behavior data such as network user access time, usage media, and usage site data may be included as necessary. Third, points related to service points that are granted and used in various ways, such as being granted and used in response to purchases of customer products, etc. by alliance companies, or as rewards for answers to questionnaires. There is system data.

Attribute data is stored in the member master database and updated when there is a change in customer attribute data. Further, the detailed data is transmitted to the operating company system 2 and stored in the analysis database as needed. Point data is stored in the point management database of the point system, and service point addition / subtraction processing is used / stored when customers purchase products at alliance companies and at various other occasions. Each time, that is, each time the service point addition / subtraction data is transmitted, the service point addition / subtraction process is performed by adding / subtracting the service point of the account corresponding to the customer ID of the customer. ing.

Each data collected in this way is managed by an individual customer ID given to the customer when the customer registers in the system 1. An example of a customer ID is a number or character string with an arbitrary number of digits that differs for each customer. If the customer has a point card or a credit card with a point card function, the customer ID Is recorded on the card by various methods such as magnetic and electrical, the customer ID is read by the input means of the store terminal 6 such as the POS terminal of the real alliance system 4, and the customer ID and the customer data are linked. Can be transmitted to the operating company system 2. Further, when the information is transmitted through the network alliance system 3, the customer ID is input by the input means of the customer terminal owned by the customer, and can be transmitted to the operating company system 2 in a state linked to the customer ID and the customer data. ing.

As shown in the flowchart of FIG. 7, the core of the system 1 of the present invention is a collection process of detailed data, a detailed data accumulation process, a detailed data analysis process, a customer segmentation process, a requirement item verification process, Action creation processing, such as recommendations, promotions, product development data, product introduction data, etc., to be provided to the store terminal 6, customer terminal 5 or company terminal 7, the action is the store terminal 6, customer terminal 5 or company terminal 7 and the like, and each of these steps is repeated to form a cycle of action planning, action execution, action verification, and action improvement.

(Detailed data collection and storage processing stage)
In the real alliance system 4, the detailed data is input from the store terminal 6 by the history input means and the customer ID is linked to the detailed data in the real alliance system 4 by the detailed data collecting means. Is transmitted to the operating company system 2 via a communication line such as an Internet line or a dedicated line. Further, in the network alliance system 3, when a customer purchases a product or the like through an EC site or the like, an operating company is connected via a communication line such as an Internet line with the detailed data and the customer ID linked from the customer terminal 5. Is transmitted to the system 2.

Next, the detailed data transmitted from the detailed data storage means to the operating company system 2 through the real alliance system 4 and the net alliance system 3 is stored in the analysis database. In this way, collection and accumulation of detailed data is performed every time detailed data is transmitted. Collection and accumulation of detailed data up to this point is performed each time the customer performs purchase behavior and the data is provided.

(Detailed data analysis stage)
Next, a basic flow from detailed data analysis to action execution of the system 1 will be described with reference to FIG. The detail data accumulated in the analysis database in the database by the detail data collection processing and detail data storage processing is analyzed by the detail data analysis means. The simplest method of the detailed data analysis means is means for analyzing the presence / absence of the purchase history record value or the purchase count of the record record value as shown in FIG. As another example, as shown in FIG. 32, a purchase history is obtained from a probability value obtained as a result of customer DNA analysis (customer profile analysis) described later, and a map estimated value calculated from the probability value as shown in FIG. There is also a method for analyzing the presence and the number of purchases.

For example, the detailed data analysis means analyzes the presence / absence of a history of an arbitrary product etc. G1 from the detailed data items. Or, analyze detailed data by analyzing customers who have purchase history of G1 for any product more than any number in any period, or purchase history but purchase history for any period is less than any number .

(Segmentation stage)
The customer segmentation means classifies customers into purchased users (customers with purchase history) and unpurchased users (customers without purchase history) based on the result of detailed data analysis. Alternatively, the customers are classified into continuous users (customers whose purchase history is an arbitrary number of times or more in an arbitrary period) and withdrawal users (customers who have a purchase history but whose purchase history is an arbitrary number of times or less). Further, when using the detailed data after customer DNA analysis, it is possible to classify customers into customers whose probability value is not less than an arbitrary threshold and customers whose probability value is not more than an arbitrary threshold. In this way, customers are segmented according to any setting.

(Verification stage)
The verification process of requirement items in the verification stage will be described. The requirement item verification means is the above-described customer segmentation means. For example, between the customer group C1 classified as a continuation user and the customer group C2 classified as a withdrawal user, among the items of arbitrary detailed data, Analyzing items that are likely to have an impact on differences in purchasing behavior (for example, items with high non-commonality) through behavioral analysis that analyzes historical data or understanding analysis that analyzes research data, and both customers Determine and extract items that have a significant impact on the group's purchasing behavior differences. As a means for quantitatively determining items that are likely to have an influence on the difference in purchase behavior of customers, for example, an item with a difference between the presence of history and the absence of history is the number of customers equal to or more than an arbitrary threshold. Judgment items that have a significant influence on the difference in purchasing behavior of the group. Items with a difference in the number of purchases exceeding an arbitrary threshold are determined as items that have a significant influence on the difference in purchasing behavior between the two customer groups. There are various methods for doing this.

For example, a customer classified into the customer group C1 of the continuing user has accumulated a lot of history in a certain item of item data, but a customer classified into the customer group C2 of the leaving user is a customer of the customer group C1. It is analyzed that the absence of history is accumulated in many of the items in which the existence of history is accumulated. In this way, items that significantly affect the presence / absence of the purchaser history of the product G1 and the presence / absence of the non-purchase history are extracted as action creation requirement items. The requirement items mentioned here are items that become clear from historical data such as “whether there is a purchase history of product G2 sold together with product G1 (purchased at the same payment opportunity)”, “visiting time”, “store used”, etc. There are various items such as “recognition of specific quality of product G1”, “customer's specific orientation level”, and other items revealed from research data. In addition, since the system 1 of the present invention collects detailed data of a plurality of alliance companies, the requirement item includes “a purchase history of the product G2 of the company B different from the company A that is the provider of the product G1. And the requirement item “There is a store use history of company B different from company A that is the provider of the product G1” can be extracted as the requirement items. There are various analysis methods for verification processing for determining requirement items, such as RFM analysis, side-by-side (basket) analysis, trade area analysis, product development analysis, existing analysis such as purchaser attitude survey, customer DNA analysis unique to the present invention, etc. There are analytical means.

(Action creation stage / Action execution stage)
The action creating means is a means for creating an action (recommendation, promotion, product development data / product introduction data) corresponding to the extracted requirement item. For example, if the requirement item determined to have the customer group C1 of the continuation user but not the customer group C2 of the exit user is the specific “visit time”, the customer group C2 is classified into the customer group C2. When the customer ID is transmitted to the management company system 2 through the store terminal 6, an action is created in which a coupon that can be used for a specific “visit time” is printed on a receipt issued from the store terminal 6. If the requirement item is “recognition of specific quality of product G1”, the one-point information advertisement related to “specific quality of product G1” is displayed on the login screen of all customer websites. Create an action to do. As an example of an action for an alliance company, there is an action of creating a list of requirement items and transmitting it to a company terminal as product development data or product introduction data. Since the system 1 of the present invention is a system that spans a plurality of alliance companies, the creation of actions and the execution of actions can be created and executed to include requirement items that span a plurality of alliance companies. . The above is the basic flow.

(Detailed data analysis stage after next)
Next, the flow after the next time after the action is executed according to the basic flow shown in FIG. 7 will be described with reference to FIG. First, as with the basic flow, collection and storage of detailed data is performed as needed. In the detail data analysis stage, the detail data analysis means analyzes the purchase history after the action. For example, the detailed data analysis means compares the item of the product G1 before the execution of the action with the item of the product G1 after the execution of the action from the customer's detailed data of the action execution target, and adds it to the customer group C2 of “no purchase history” When the actual value of “with purchase history” is assigned to the item of the product G1 corresponding to the customer ID of the classified customer, the customer is analyzed as having an effect of the action.

(Next segmentation stage)
The detailed data analysis described above can be performed for the customers targeted for the action, but since it can be performed for all customers after the action, the customer group is “customer bought with an assumed target”. Customers can be reclassified into new segments, such as “customer purchased with an unexpected target”, “customer withdrawn from an existing user”, and “customer not purchased with an assumed target”.

(Next verification stage)
The verification method is remarkably different in the purchase behavior of “customer bought with an assumed target”, “customer purchased with an unexpected target” and “customer withdrawn from an existing user” and “customer not bought with an assumed target”. The requirement items are judged and extracted by various analysis means.

(Next action creation stage / action execution stage)
Then, the action creating means performs an improvement process for replacing the content of the action based on the newly determined requirement item. The action execution means provides the action after the improvement process to the store terminal 6, the customer terminal 5, the company terminal 7, and the like. In this way, the initial analysis and action creation process, action execution process, re-analysis after action execution process, action improvement process, and improved action execution process are repeated, and the content of the action is A cycle that improves the purchasing effect has been completed.

(Concrete example)
Here, referring to the flowcharts of FIGS. 9 and 10 and the schematic diagrams of FIGS. 11 to 21, the behavior of analyzing history data as a verification means for requirement items, taking an example of an arbitrary product (eg, collagen drink) as an example The specific flow will be described separately for analysis and understanding analysis for analyzing research data.

The “collagen drink” illustrated in FIG. 11 has a relatively high sales among various “collagen drinks”, is pushed by a low-priced product, and the amount of collagen is not much different from other products. Suppose that the content of is rich.

As shown in FIG. 12, with respect to the “collagen drink”, the customer is classified into a customer who has a purchase history of the “collagen drink” and a customer who has no purchase history through detailed data analysis and segmentation processing. From the segmentation processing in the first layer, analyzing the attribute data of customers classified into segments with purchase history, it is analyzed that there are many `` female '', and it is analyzed from the history data that there are many `` night life '' customers It is analyzed that there are many customers with orientations of “low price orientation” and “health orientation”. If necessary, the segmentation means performs segmentation processing on the second hierarchy for customers with purchase histories, using the number of purchases and the purchase period as threshold values, and classifies the customer as a continuing user and a leaving user.

Next, referring to FIG. 13, the behavior analysis means as the verification processing means combines various analysis means such as RFM analysis and side-by-side analysis to change the purchase behavior among the items of the history system data of the continuation user and the withdrawal user. Determine which items are influencing (for example, items with high non-commonality). For example, here, the items “Purchase more than 6 times a month” and the item “Purchase tendency at lunch time” appear to be biased toward continuing users, and the behavior analysis means that these items are not significantly common to both These items are extracted as requirement items that affect the purchasing behavior of this “collagen drink”. On the other hand, among the items of history data, for example, the item “Purchase other company's product at the time of other company's sale” appears on both sides, and the behavior analysis means that this item has no remarkable non-commonality in both It is determined that the items are (that is, common), and these items are not extracted as requirement items that affect the purchasing behavior of the “collagen drink”.

Next, referring to FIG. 14 and FIG. 15, the understanding analysis means as the verification processing means combines various analysis means such as a buyer attitude survey, product development analysis, etc., and the items of research data of the continuing user and the leaving user Among these, an item (for example, an item having a high degree of non-commonality) that affects purchase behavior is determined. For example, the item “Abundance of knowledge about vitamin E (having knowledge that vitamin E helps collagen absorption)” is an item that appears biased to continued users. This item is extracted as a requirement item that affects the purchasing behavior of “collagen drink”. On the other hand, among the items of historical data, the items of “brand recognition”, “brand image”, and “richness of knowledge about collagen” are items that appear on both sides. It is determined that there is no non-commonality (that is, common items), and these items are not extracted as requirement items that affect the purchasing behavior of this “collagen drink”.

Referring to FIG. 16, the preparation means is to understand the requirement items verified from the above-mentioned analysis “low price orientation” and “purchase more than 6 times a month” “purchase at lunch time” “collagen supplementary effect of vitamin E”. “Abundance” and the like are listed and processed into product development data or product introduction data to be provided to the company terminal 7, and an action of creating recommendations and promotions to satisfy the requirement item points is created.

FIG. 17 shows an example of creating an action for a customer. For example, a “collagen drink” is set so as to satisfy the requirement item “purchase at lunchtime”. To create an action to issue a “campaign coupon to be performed” from the store terminal 6. Also, to satisfy the requirement item of “Purchase more than 6 times in a month”, an advertisement mail “Providing a special present application ticket when purchasing more than 6 times” is delivered to the customer terminal 5 Create an action. Or, create an action that creates a media advertisement that “includes the appeal of the effect of vitamin E” so as to satisfy the requirement item of “abundance of understanding of collagen supplementary effects of vitamin E”.

As shown in FIG. 18, a requirement item (for example, an item having a high degree of non-commonality) having a high degree of influence on purchase behavior of a continuation user and a withdrawal user, or a user who has a purchase history and a user who has no purchase history, is analyzed. By making the state of item data items of a user who has a continuation user and a purchase history item data items as close as possible by creating and executing various actions, A user without a purchase history is changed to a continuous user and a user with a purchase history. Here, both behavioral analysis that analyzes historical data and comprehension analysis that analyzes research-based data are used to analyze behavior based on customer consciousness that is analyzed from historical data and research-based data. Requirement items derived from customer conscious behavior will be used to create actions to bridge the differences between the two.

As shown in FIG. 19, the detailed data analysis means analyzes the history data after the action, compares it with the history data before the action, and performs customer segmentation. "" People who bought with an unexpected target "" "People who left with an existing user" "" People who did not buy with an assumed target "

As shown in FIG. 20 and FIG. 21, the above-mentioned detailed data collection / accumulation, detailed data analysis, customer segmentation, verification of requirement items, creation / improvement of actions, and people classified into any segment by repeating actions On the other hand, it is possible to verify why a person classified into a segment did not move to purchasing behavior, and to perform actions such as better recommendation, promotion, and data provision. At this time, in the system 1, since the net action data is also accumulated, the action of the network advertisement can be executed aiming at the access time, the use medium, and the use site that are frequently viewed by the customer as the action target. Furthermore, since the living conditions of individual customers can be analyzed, the action of TV advertisement can be executed aiming at the staying time of the target customer. Furthermore, it is possible to provide a paper advertisement action to a target customer's subscription magazine or newspaper.

Finally, some analysis methods for analyzing detailed data will be described. The first analysis model in the detailed data analysis stage is RFM analysis. RFM analysis is a method of analyzing detailed data from the viewpoint of purchase period, number of purchases, and purchase price. An example of RFM analysis means is the date of purchase by the alliance company recently from the historical data of the detailed data, the number of purchases of products etc. at the alliance company over a certain period, and the purchase amount for a certain period The weights set for each asset management company or alliance company are assigned to each item, the total evaluation score is calculated, and the analysis is performed to evaluate the possibility of purchasing the target product. Since the system 1 of the present invention can analyze the detailed data over a plurality of alliance companies, the detailed data over the plurality of alliance companies of one customer can be analyzed by the cut point of the RFM analysis. Therefore, it is possible to analyze the purchase period, the number of purchases, and the purchase price of purchase behavior across alliance companies of arbitrary customers. As a result, an analysis result or the like for predicting an alliance company or a store where an arbitrary customer may go next, a product that may be purchased, or the like is provided.

The second analysis model is a trade area analysis. The trade area analysis is a technique for analyzing detailed data from a regional perspective. An example of a trade area analysis means is to give a flag to the residence characteristics and / or commuting / commuting area characteristics of the customer's attribute data, and based on the flag, the common residence characteristics and / or commuting / commuting area characteristics can be obtained. Analyzing the number of purchases and purchase price from the historical data of the customer group, assigning weights uniquely set by the asset management company or each alliance company, calculating the total evaluation score, target products, stores, alliance companies Analyzes to evaluate purchase potential at.
Therefore, an analysis result or the like for predicting a customer who may purchase in an arbitrary trade area is provided. Since the system 1 of the present invention can analyze detailed data over a plurality of alliance companies, the detailed data over a plurality of alliance companies of one customer can be analyzed by trade area analysis. Therefore, it is possible to analyze in which trade area an arbitrary customer is making a purchase behavior. In addition, an analysis result is provided for attracting a customer who has historical data in one alliance company in the same trade area to another alliance company in the same trade area that does not have historical data, for example.

The third analysis model is side-by-side analysis (basket analysis). The side-by-side analysis is a technique for analyzing what kind of products a given customer purchases at the same opportunity. An example of side-by-side analysis means is that when historical data is transmitted, historical data transmitted in units of one receipt is collected at a time to analyze products purchased by a customer with a single purchase. Is done. Since the system 1 of the present invention can collect and store detailed data over a plurality of alliance companies, the detailed data over a plurality of alliance companies of one customer can be analyzed by the side-by-side analysis. Therefore, an analysis result for predicting a tendency of a product or the like that an arbitrary customer purchases in one purchasing unit is provided.

The fourth analysis model is product development analysis. Product development analysis is a comprehensive combination of the analysis results of RFM analysis, trade area analysis, or side-by-side analysis based on the above first to third analysis models to determine the purchase period, number of purchases, purchase price, trade area, side-by-side unit, etc. This is a method for analyzing new products that are easier to purchase.

The fifth analysis model is a buyer attitude survey. The operation system 2 is a technique for acquiring research data from a customer in the form of answering an arbitrary questionnaire and analyzing it. In the questionnaire, the customer simply answers “I want” or “I don't want” any product, or the customer ’s orientation (high-class, traditional, health-oriented), etc. There is.

The sixth analysis model is a new detailed data analysis called customer DNA analysis (customer profile analysis). Customer DNA analysis is a data analysis in which probability values are predicted for all items in the customer's detailed data, and the probability values are assigned to each item in the detailed data, including the part where the customer does not provide actual values of the detailed data. It is. In other words, according to customer DNA analysis, customer detail data is analyzed more accurately, and by combining with the above first to fifth analysis methods, analysis of customer detail data that does not have actual value of detail data Can also be done. Hereinafter, this data analysis will be described in detail with reference to FIGS.

Referring to FIGS. 22 and 23, the data summarization process (S3) as the first step of the detailed data analysis stage in the customer DNA analysis after the above-described collection (S1) and accumulation (S2) of detailed data is performed. Will be described. In the data summarization process, the attribute data and / or the detail data are discretized as necessary, and the attribute data and / or the detail data are collected into the category data, the upper hierarchy, or the large / middle classification item group. This is a process for generating a data mart. By generating a data mart, attribute data and / or detailed data can be collected into data items that are easy to analyze.

Next, the sampling process, modeling process (S5), verification process (S6), and scoring process (S7) will be described. Here, the prediction process of the present invention, which is the basis of the modeling process (S5), the verification process (S6), and the scoring process (S7) described later, will be described.

For example, the prediction process uses probability inference represented by a Bayesian network, using multiple items of arbitrary detailed data as objective variables and other multiple items as explanatory variables based on the correlation between the items. The probability value is calculated by probability reasoning.

Hereinafter, the prediction processing of the present invention will be described with reference to the prediction processing models shown in FIGS. The prediction process of Case 1 shown in FIGS. 24 to 26 is a case where the prediction process is a one-layer model. Assuming that the probability of marriage is the objective variable for any customer A's prediction process, if there is no detailed data for customer A, that is, there is no actual value data in the explanatory variable, Presence probability values (married about 0.57, unmarried about 0.43) obtained from the data of customers who provided research-related data are assigned to the probability value of presence / absence. Similarly, in the state where there is no data for other explanatory variables (condiment purchase, side dish purchase, late-night time use ratio), the prior probability value of seasoning purchase obtained from the data of customers who provided the detailed data (many 0.2 less 0.8), pre-probability value of side dish purchase (0.35 with purchase, 0.65 without purchase), prior probability value of usage ratio in midnight hours (more 0.1, normal, 0.3) A small 0.6) is shaken.

According to FIG. 25, when detailed data of “purchased” is obtained for the customer A among the explanatory variables of the customer A, the probability value (1.0) of the side dish purchase is entered, and the target variable is obtained by the prediction process. The probability of being unmarried increases from 0.43 to 0.63.

According to FIG. 26, when detailed data “more” is obtained for use in the midnight time zone among the explanatory variables of customer A, the probability value (1.0) is entered, and the unmarried probability value as the objective variable is 0. .63, and an expectation process of further increasing to 0.90 is performed.

Next, the prediction processing in the case of the two-layer model will be described with reference to the prediction processing model shown in FIGS. Assuming that the probability of handmade orientation is the second objective variable for the forecast processing of any customer A, at this time, if there is no detailed data for customer A, that is, there is no actual value data in the explanatory variable, The handmade-oriented probability value is assigned a prior probability value (about 0.48 high, about 0.52 low) obtained from the data of the customers who provided research-related data. As the probability value, a prior probability value (married about 0.57, unmarried about 0.43) obtained from the data of customers who provided the research data is assigned.

Next, when the detailed data of “purchased” is obtained for the seasoning purchase among the explanatory variables of customer A, the probability value (1.0) of seasoning purchase is entered in the same manner as the one-layer model mechanism described above. As a result of the prediction process, the probability value of handmade orientation increases from 0.48 to 0.71.

Further, when the detailed data of “no purchase” is obtained for the side dish purchase of the explanatory variable of customer A, the probability value (1.0) of the side dish purchase is entered, and the probability of marriage with or without marriage as the first objective variable. The value increases from 0.57 to 0.68. Furthermore, the presence or absence of marriage as the first objective variable is also an explanatory variable for handmade orientation as the second objective variable. The probability value of presence / absence fluctuates, and the fluctuation of the first objective variable is propagated to the fluctuation of the second objective variable, so that the handmade-oriented probability value as the second objective variable is further 0.71 to 0.74. Go up to.

As shown in FIG. 30, the prediction process is a process in which the fluctuation of the probability value of the explanatory variable (parent) changes the probability value of the first objective variable (child), and the first objective variable (child) is updated. Is the process of changing the probability value of the second objective variable (child (grandchild for the first explanatory variable)). That is, one objective variable is also an explanatory variable for the next generation and subsequent objective variables, and it is possible to set all objective variables (nth generation objective variables) linked as necessary. That is, in the prediction process of the present invention, the input of the actual value of an arbitrary explanatory variable affects the probability values of all linked objective variables, and performs the prediction process of calculating the probability value of each item. Also, in the prediction process of two or more layers, the actual value of the explanatory variable of the parent generation is entered, and the influence is propagated to the linked (child, grandchild, nth generation objective variable) fluctuation, and the probability of each item Predictive processing to calculate the value.

Sampling process (S4) is an operation for extracting an arbitrary customer in order to create a provisional model with high accuracy when creating a provisional model for analysis of prediction process for analyzing detailed data for all customers. That is, for the test model for analysis of the prediction process, a process of extracting a data mart (for example, a data mart of 3000 items) after summarizing an arbitrary number of customers (for example, one million people) is performed. This sampling process is a process performed before the modeling process. Customers that are subject to extraction for the analytical test model in the sampling process can be extracted at random, but in order to enable accurate verification in the verification process described later, a customer who has an answer to the probability value, that is, a predetermined value It is appropriate to extract an arbitrary number of customers from customers in which detailed data of actual values of items or more are accumulated.

In the modeling process (S5), before the detailed data for all customers is predicted, the detailed data sampled by the sampling means described above is predicted using the analytical test model and the probability value is calculated. This is a process of generating a provisional model for scoring by verifying the value and the answer obtained from the detailed data actually collected and accumulated, that is, the prior probability value.

In the modeling process, a probability value is calculated for each objective variable subjected to the prediction process for the sampled customer data by a test model for analysis of the prediction process. Next, a verification process for calculating a difference between the probability value obtained by the prediction process using the analysis test model and the prior probability value obtained for each objective variable collected in advance as the actual value is performed. As a result of the verification process, the test model correction process for analysis such as replacement of the item of the objective variable or explanatory variable or change of the discretization method for the item so that the difference from the prior probability value is less than an arbitrary value. Is repeated, a provisional model for scoring in a state where the difference from the prior probability value is equal to or less than an arbitrary value is determined. Since this modeling process becomes older each time the detailed data is collected and accumulated, the modeling process is periodically performed, and a provisional model of a prediction process in accordance with the actual situation is generated each time.

The scoring process is a process for expanding the provisional model generated by the modeling process to the data marts of all customers. Thereby, probability values are calculated for all items of the data mart of all customers.

That is, in a state where scoring is not performed, the data about the customer is actually collected attribute data and detailed data as shown in the result table of FIG. In this regard, when the prediction process is performed using the above-described provisional model, probability values are obtained for all items as shown in a table after probability inference, also called a customer DNA table, shown in FIG. Furthermore, as shown in FIG. 33, it is also possible to obtain a map estimation value including processing such as determining a probability value with a threshold value based on the tendency of the result table as required.

顧客 By analyzing customer data by such processing, it is possible to accurately analyze the data of many customers in a short time. When the number of customers is overwhelmingly small, sampling processing, modeling processing, and verification processing can be omitted, and prediction processing can be performed on detailed data for all customers.

As described above, detailed data analysis, behavior analysis, and understanding analysis can be performed from the actual values in the actual table, but more accurate requirement items are extracted for customers with a small amount of actual value data and actions are created. Analyze detailed data of probability values or map estimates after customer DNA analysis, segment customers, conduct behavior analysis and understanding analysis on probability values or map estimates after customer DNA analysis By determining, a more accurate action can be created.

As described above, according to the system 1 of the present invention, data provided by a customer who is a subscriber of a service point system is analyzed by various methods, and actions such as promotion and recommendation are performed on the customer based on the analysis result. Plans actions such as data provision, executes actions, verifies how effective this action is, and further improves the action plan, action execution, action effect measurement, action It is possible to provide an analysis / verification system for customer data having an improvement cycle.

Moreover, since the system 1 of the present invention can analyze customer data across a plurality of alliance companies, the customer can use the data across a plurality of companies, and receives recommendations and promotions according to his / her living conditions and preferences. The alliance company can make recommendations and promotions to the right customers as targets without owning customer data, and can obtain data useful for product development and product introduction. it can.

Claims

The alliance company has a means for collecting customer detailed data, a means for accumulating the detailed data, an analyzing means for analyzing the detailed data, and a means for classifying customers into segments, and is classified into one segment. A customer data analysis / verification system comprising verification means for determining a requirement item that affects a difference in purchase behavior from an item of customer detail data and an item of customer detail data classified into another segment .
The verification means a requirement that affects the difference in purchasing behavior between items of customer detail data classified into one segment and items of customer detail data classified into other segments. The customer data analysis / verification system according to claim 1, wherein the system is determined as an item.
3. The customer data analysis / verification system according to claim 1 or 2, wherein the verification means is behavior analysis means for determining a requirement item from an item of history data among items of detailed data.
4. The customer data analysis / verification system according to claim 1, wherein the verification means is an understanding analysis means for determining a requirement item from an item of research data among items of detailed data.
5. The method according to claim 1, further comprising: an action creation unit that lists the requirement items determined by the verification unit; and an action execution unit that provides the list of requirement items to a company terminal. Customer data analysis and verification system.
6. The method according to claim 1, further comprising: an action creating unit that creates a recommendation including the requirement item determined by the verification unit; and an action executing unit that provides the recommendation to the store terminal and / or the customer terminal. Customer data analysis and verification system described in 1.
7. The method according to claim 1, further comprising action creating means for creating a promotion including the requirement item determined by the verifying means, and action executing means for providing the promotion to a store terminal and / or a customer terminal. Customer data analysis and verification system described in 1.
8. The customer data according to claim 1, further comprising means for comparing the detailed data before the execution of the action and the detailed data after the execution of the action, and determining a customer who has caused the purchase behavior after the action. Analysis and verification system.