JP2018077636A - Program and combination extraction system - Google Patents

Abstract

Compared to a case where a product is recommended based on a purchase tendency of a product by a purchaser, the effect of recommending the purchaser to purchase the product is enhanced. A sales support device 100 proposes a purchase probability calculating unit 116 that calculates a purchase probability of each product by each customer from a purchase history in which the product is associated with a customer who has purchased the product, and proposes the product and the product. A proposal probability calculation unit 117 that calculates a proposal probability of each product by each sales from a proposal history that is associated with the sales, and a purchase probability that is higher than the reference value of the purchase probability and the proposal probability is a reference value of the proposal probability A recommendation list generation unit 119 that generates a recommendation list including a combination of a product, a sales, and a purchaser. [Selection] Figure 4

Description

  The present invention relates to a program and a combination extraction system.

  For example, Patent Literature 1 receives a content recommendation request from a communication terminal device of a target user, determines a group in which the target user is participating, and other users who are participating in the group, The usage history information of the user's content is acquired from the content usage history DB, the usage parameters relating to the usage of the content included in the usage history information of the target user satisfy the given condition, and are included in the usage history information of other users It is disclosed that content to be specified is specified as recommended content, and information related to the recommended content is transmitted to a target user's communication terminal device so as to be viewable.

  Further, for example, Patent Document 2 includes means for receiving product group information from a sales representative, customer selection means for selecting customers as sales target candidates based on purchase behavior for each product in the product group, and candidate selection means. Sales support device comprising analysis means for analyzing commonality by referring to selected customer stratification attributes and search means for searching customer information from a customer database based on a search condition based on the result of the analysis Is disclosed.

JP 2013-206194 A JP 2002-157272 A

In a company or the like, a so-called sales person may make a proposal (recommendation) to a purchaser. In addition, based on the purchaser's preference distribution, purchase history, and the like, a product may be recommended to the purchaser. In such a case, the sales representative as a recommender recommends a product that the purchaser purchases without recommending it in the first place, or a product for which the purchaser has a willingness to purchase. There is a possibility of not recommending.
An object of the present invention is to increase the effect of recommending a purchaser to purchase a product as compared with the case of recommending a product based on the purchase tendency of the product by the purchaser.

According to the first aspect of the present invention, information on a recommended product purchase history obtained by associating a plurality of products, a plurality of recommenders who recommend products to customers, and a plurality of purchasers who have purchased products is acquired in a computer. A function for outputting a recommended index based on each product recommended by each recommender from the acquired product recommended purchase history information, and a product for which the recommended index satisfies a predetermined criterion And a function for extracting a combination of a recommender and a purchaser.
The invention according to claim 2 is the program according to claim 1, wherein the recommendation index is a probability that one recommender is expected to propose one product to one customer.
According to a third aspect of the present invention, the recommendation index includes, when one recommender proposes one product to one customer, the probability that the one customer is expected to purchase the one product, The program according to claim 1, wherein a probability that the one customer is predicted to purchase the one product when the recommender does not propose the one product to the one customer.
The invention according to claim 4 is characterized in that, when one recommender proposes one product to one customer, the recommendation index is a probability that the one customer actually purchased the one product. The program according to claim 1.
The invention according to claim 5 is characterized in that the output function further outputs another recommended index based on each product purchased by each purchaser, and the function to be extracted is a function in which the other recommended index is predetermined. 5. The program according to claim 1, wherein a combination of one product, one recommender, and one purchaser that further satisfies the criteria is extracted.
According to a sixth aspect of the present invention, the other recommended index is a probability that one customer purchases the one product when one recommender proposes one product to one customer. 6. The program according to claim 5, wherein the program is characterized in that:
The invention according to claim 7 is an acquisition unit that acquires information on a recommended product purchase history in which a plurality of products, a plurality of recommenders who recommend products to customers, and a plurality of purchasers who have purchased products are associated with each other. And an output unit that outputs a recommendation index based on each product recommended by each recommender from each of the acquired product recommended purchase history information, and a product that satisfies the predetermined criterion. It is a combination extraction system including an extraction unit that extracts combinations of recommenders and purchasers.

According to the first aspect of the present invention, it is possible to enhance the effect of the recommendation to cause the purchaser to purchase the product as compared with the case of recommending the product based on the purchase tendency of the product by the purchaser.
According to the invention described in claim 2, it is possible to recommend a product that the recommender is unlikely to recommend to the purchaser.
According to the third aspect of the present invention, it is possible to recommend a product that can be urged to be purchased by the purchaser through the presence of the recommender.
According to the invention described in claim 4, it is possible to recommend a product that can promote purchase by the purchaser by exhibiting the ability of the recommender.
According to the invention described in claim 5, it is possible to recommend a product that the purchaser is likely to purchase.
According to the sixth aspect of the present invention, it is possible to more reliably recommend a product that the purchaser is likely to purchase.
According to the seventh aspect of the present invention, it is possible to enhance the effect of the recommendation to cause the purchaser to purchase the product as compared with the case of recommending the product based on the purchase tendency of the product by the purchaser.

It is a figure for demonstrating the structure used as the premise of each embodiment. It is a figure for demonstrating the combination of a customer, goods, and sales. It is a figure which shows the hardware structural example of the sales support apparatus with which this Embodiment is applied. 3 is a block diagram illustrating a functional configuration example of a sales support apparatus according to Embodiment 1. FIG. (A) is a figure which shows an example of the purchase history stored in a purchase history storage part, (b) is a figure which shows an example of the proposal history stored in a proposal history storage part. 5 is a flowchart illustrating an example of a recommendation list output procedure according to the first embodiment. It is a figure for demonstrating the purchase probability of each goods by a target customer. It is a figure for demonstrating the similarity between customers based on purchase goods. It is a figure for demonstrating the similarity between customers based on a customer attribute. It is a figure for demonstrating the similarity between customers based on machine learning. It is a figure for demonstrating the proposal probability (the 1) of each goods with respect to an object customer by object sales. It is a figure for demonstrating the proposal probability (the 2) of each goods with respect to an object customer by object sales. It is a figure for demonstrating the similarity between sales based on a proposal goods. It is a figure for demonstrating the similarity between sales based on a sales attribute. It is a figure for demonstrating the similarity between sales based on machine learning. It is a figure which shows an example of a recommendation list. 6 is a block diagram illustrating an example of a functional configuration of a sales support apparatus according to Embodiment 2. FIG. 12 is a flowchart illustrating an example of a recommendation list output procedure according to the second embodiment. It is a figure for demonstrating the purchase probability of each goods via sales of an object customer. It is a figure for demonstrating the purchase probability of each goods which do not go through sales of an object customer. FIG. 10 is a block diagram illustrating a functional configuration example of a sales support apparatus according to Embodiment 3. 15 is a flowchart illustrating an example of a recommendation list output procedure according to the third embodiment. It is a figure for demonstrating the average order rate of each goods of object sales. FIG. 10 is a block diagram illustrating a functional configuration example of a sales support apparatus according to a fourth embodiment. 15 is a flowchart illustrating an example of a recommendation list output procedure according to the fourth embodiment.

Embodiments of the present invention will be described below in detail with reference to the accompanying drawings.
<Prerequisite configuration>
FIG. 1 is a diagram for explaining a configuration that is a premise of each embodiment described below. FIG. 1 shows the relationship between a product to be bought and sold, a customer who purchases (purchases) the product, and a business that proposes (recommends) the product to the customer. In the example illustrated in FIG. 1, it is assumed that there are 6 sales (sales S1 to S6), 5 types of products (products P1 to P5), and 6 customers (customers C1 to C6). . Here, the customer is an example of a purchaser, and the business is an example of a recommender.

  In this example, sales S1 to S6 make proposals to urge customers C1 to C6 to purchase products P1 to P5. The customers C1 to C6 who have received the proposal purchase the products P1 to P5 as necessary. Further, the customers C1 to C6 may purchase products P1 to P5 that have not received proposals from the sales S1 to S6 based on, for example, advertisements and reviews.

  In each embodiment described below, the sales support device analyzes the relationship between the customer, the product, and the sales, and uses a plurality of recommended indicators (details will be described later) to use the customer, the product, and the sales. Extraction about the combination of.

<Combination of customers, products and sales>
FIG. 2 is a diagram for explaining combinations of customers, products, and sales. FIG. 2 shows seven combinations of customers, products and sales depending on which factors are fixed and which are variable.

  The first combination is when the customer is “variable”, the product is “variable”, and the sales is “variable” (all are variable). The first combination is used when it is desired to decide which product should be proposed to which customer.

  The second combination is a case where the customer is “variable”, the product is “variable”, and the business is “fixed”. The second combination is used when a target sales (target sales), such as a new business development activity, wants to determine which product should be proposed to which customer.

  The third combination is a case where the customer is “variable”, the product is “fixed”, and the business is “variable”. The third combination is used when it is desired to determine which sales should propose to which customer the target product (target product), such as when the priority product is determined in the sales department.

  The fourth combination is a case where the customer is “fixed”, the product is “variable”, and the sales is “variable”. The fourth combination is used when it is desired to determine which sales should propose which product to a target customer (target customer), for example, when a priority customer is determined in the sales department.

  The fifth combination is a case where the customer is “variable”, the product is “fixed”, and the business is “fixed”. The fifth combination is used, for example, when it is desired to determine to which customer a product (target product) targeted by the target business (target sales), such as a new business development activity, should be proposed.

  The sixth combination is a case where the customer is “fixed”, the product is “variable”, and the business is “fixed”. The sixth combination is used when it is desired to determine which product should be proposed to the target customer (target customer) for the target business (target sales).

  The seventh combination is a case where the customer is “fixed”, the product is “fixed”, and the business is “variable”. The seventh combination is used when it is desired to determine which business should be proposed when a target customer (target customer) and a target product (target product) are determined.

  Here, combinations (1, 3, 4, 7) in which the sales are “variable” are assumed to be used by the management side (including the marketing department and the sales planning department) that manages the sales (person in charge). On the other hand, combinations (2, 5, 6) in which sales are “fixed” are assumed to be used by individual sales (persons in charge).

  Embodiments 1 to 3 to be described later are based on the sixth combination (customer: fixed, product: variable, business: fixed) shown in FIG. Further, Embodiment 4 described later is based on the seventh combination (customer: fixed, product: fixed, business: variable) shown in FIG. However, the first to seventh combinations are all included in the present invention.

<Embodiment 1>
[Example of hardware configuration of sales support device]
FIG. 3 is a diagram illustrating a hardware configuration example of the sales support apparatus 100 to which this exemplary embodiment is applied.
A sales support device 100 as an example of a combination extraction system is a computer device that is used by a salesperson to propose (recommend) a product to a customer in, for example, a retail business. The sales support apparatus 100 according to the present embodiment includes a CPU (Central Processing Unit) 101 that is a calculation means, a main memory 102 that is a storage means, and a magnetic disk device 103.

  Here, the CPU 101 executes various programs such as an OS (Operating System) and applications to realize various functions of the sales support apparatus 100. The main memory 102 is a storage area for storing various programs and data used for execution thereof. The magnetic disk device 103 is a storage area for storing input data for various programs, output data from various programs, and the like.

  Furthermore, the sales support apparatus 100 includes a communication interface (communication I / F) 104 for performing communication with the outside, a display mechanism 105 including a video memory and a display, and an input device 106 including a keyboard and a mouse. Prepare.

[Functional configuration example of sales support device]
FIG. 4 is a block diagram illustrating a functional configuration example of the sales support apparatus 100 according to the first embodiment.
The sales support apparatus 100 according to the present embodiment includes a purchase history storage unit 111 that stores a purchase history of products by a customer, and a proposal history storage unit 112 that stores a proposal history of products by sales. Further, the sales support device 100 includes a product attribute information storage unit 113 that stores attribute information (product attribute information) of each product, a customer attribute information storage unit 114 that stores attribute information (customer attribute information) of each customer, A sales attribute information storage unit 115 that stores attribute information (sales attribute information) of each business. Furthermore, the sales support apparatus 100 includes a purchase probability calculation unit 116 that calculates a probability that a customer purchases a product (purchase probability), and a proposal probability calculation unit 117 that calculates a probability that the sales will propose a product (proposition probability). Prepare. Further, the sales support apparatus 100 uses the purchase probability calculated by the purchase probability calculation unit 116 and the proposal probability calculated by the proposal probability calculation unit 117 to determine the usefulness of the relationship between the sales, the product, and the customer. Based on the determination result by the determination unit 118, the usefulness determination unit 118, a recommendation list generation unit 119 that generates a list (recommendation list) of combinations of sales, products, and customers to be recommended, and a recommendation list generation unit 119. A recommendation list output unit 120 for outputting the created recommendation list to the outside.

  The purchase history storage unit 111 stores a purchase history of products by customers. The purchase history of the present embodiment indicates a record relating to each product that has been purchased by a customer.

  Here, FIG. 5A is a diagram illustrating an example of a purchase history stored in the purchase history storage unit 111. The purchase history shown in FIG. 5A associates the purchased product, the customer who purchased the product, the business that proposed the product, and the purchase date of the product. For example, as shown in the uppermost row in the figure, the customer “C3” purchases the product “P2” from the sales “S1” on the purchase date “2016/09/01”. For example, as shown in the third row in the figure, the customer “C3” purchases the product “P4” on the purchase date “2016/09/02” without going through sales. In addition to these, the number of products sold and the number of sales (such as the number of sales) may be included in the purchase history.

  The proposal history storage unit 112 stores a product proposal history by sales. This proposal history shows a record relating to each of the products that the business has proposed to the customer.

  Here, FIG. 5B is a diagram illustrating an example of the proposal history stored in the proposal history storage unit 112. The proposal history shown in FIG. 5B is a table in which the proposed product, the customer to whom the product is proposed, the business that has proposed the product, and the proposal date of the product are associated with each other. For example, as shown in the uppermost row in the figure, the customer “C3” proposes the product “P2” from the sales “S1” on the proposal date “2016/08/03”. Further, as shown in the third row in the figure, the customer “C2” proposes the product “P3” on the proposal date “2016/08/04” from the sales “S4”.

  Note that these purchase history and proposal history are examples of the recommended product purchase history.

  The product attribute information storage unit 113 stores attribute information of each product (here, five types of products P1 to P5). Here, the attribute information of the product includes, for example, “manufacturer”, “use”, “color”, “price range”, and the like.

  The customer attribute information storage unit 114 stores attribute information of each customer (here, six companies of customers C1 to C6). Here, as customer attribute information, for example, “industry”, “number of employees”, “region”, “business relationship”, and the like can be cited.

  The sales attribute information storage unit 115 stores attribute information of each sales (here, 6 people of sales S1 to S6). Here, the sales attribute information includes, for example, “affiliation department”, “sales experience years”, “region”, “work style”, and the like.

  The purchase probability calculation unit 116 calculates the purchase probability (an example of another recommended index) of each product by each customer based on the purchase history acquired from the purchase history storage unit 111. Details of the purchase probability will be described later.

  The proposal probability calculation unit 117 calculates a proposal probability (an example of a recommended index) of each product by each business based on the proposal history acquired from the proposal history storage unit 112. Details of the proposal probability will be described later.

  The usefulness determination unit 118 determines the usefulness of the relationship between sales, products, and customers using the purchase probability calculated by the purchase probability calculation unit 116 and the proposal probability calculated by the proposal probability calculation unit 117.

  The recommendation list generation unit 119 generates a list (recommendation list) of combinations of sales, products, and customers to be recommended based on the determination result by the usability determination unit 118. The recommendation list generation unit 119 of the present embodiment registers a combination of sales, a product, and a customer determined to be highly useful by the usefulness determination unit 118 in the recommendation list. Details of the recommendation list will be described later.

  The recommendation list output unit 120 outputs the recommendation list created by the recommendation list generation unit 119 to the outside. Here, the recommendation list output unit 120 outputs the recommendation list in a state of being printed on a sheet, for example, or in the state of electronic data.

  Here, in the present embodiment, the purchase probability calculation unit 116 and the proposal probability calculation unit 117 function as an acquisition unit and an output unit, and the usefulness determination unit 118 and the recommendation list generation unit 119 function as an extraction unit. doing.

  Note that each functional unit constituting the sales support apparatus 100 shown in FIG. 4 and the like is realized by cooperation of software and hardware resources. The program executed by the CPU 101 can be provided in a state of being stored in another computer-readable recording medium in addition to a form stored in the magnetic disk device 103 in advance. Further, the program may be provided from the Internet 4 to the CPU 101 via the wireless base station device 2 or the like.

[Recommendation list output procedure]
FIG. 6 is a flowchart showing an example of a recommendation list output procedure according to the first embodiment. Here, as described above, the present embodiment is based on the sixth combination (customer: fixed, product: variable, sales: fixed) shown in FIG. 2, and the target sales (target sales) is the target. This is used when it is desired to decide which product should be proposed to the customer (target customer).

  First, the purchase probability calculation unit 116 acquires a purchase history from the purchase history storage unit 111 and acquires customer attribute information from the customer attribute information storage unit 114. Then, the purchase probability calculation unit 116 calculates the degree of similarity between customers based on the acquired purchase history and customer attribute information (step 110).

  Next, the purchase probability calculation unit 116 calculates the purchase probability of each product of the customer based on the purchase tendency of similar customers obtained using the similarity between customers calculated in step 110 (step 120). ).

  Subsequently, the proposal probability calculation unit 117 acquires a proposal history from the proposal history storage unit 112. And the proposal probability calculation part 117 calculates the proposal probability of each product of sales based on the acquired proposal history (step 130).

  Then, the usefulness determination unit 118 proposes a product whose purchase probability calculated in step 120 is a certain value or more and whose proposal probability calculated in step 130 is a certain value or less to the target customer. A product to be (recommended) is selected (step 140). Here, in step 140, one or more products may be selected. Thereafter, the recommendation list generation unit 119 generates a recommendation list in which the target customer, the target sales, and the product to be recommended are associated with each other, and the recommendation list output unit 120 generates the recommendation list generated by the recommendation list generation unit 119 as an external Output to.

[Specific Example of Embodiment 1]
Now, the recommendation list output procedure in the present embodiment will be described with a specific example.
Here, as an example, the business S6 that is the target business wants to determine which of the products P1 to P5 is proposed to the customer C6 that is the target customer.

[Purchase probability]
FIG. 7 is a diagram for explaining the purchase probability of the products P1 to P5 by the target customer (customer C6). FIG. 7 corresponds to step 120 shown in FIG.
FIG. 7 shows the purchase history (whether or not purchased) of the products P1 to P5 by the customers C1 to C5 other than the target customer (customer C6), and the similarity (between customers) of the customer C6 and the other customers C1 to C5 ), The purchase probability based on the simple total, and the purchase probability when weighting is performed based on the similarity between customers. In FIG. 7, when there is a purchase history, it is expressed as “1”, and when there is no purchase history, it is expressed as “0” (the same applies below).

  In this example, the similarities of the customers C1 to C5 (similarity between customers) with respect to the customer C6 are “0.8”, “0.8”, “0.6”, “0.4”, “ 0.4 ". A method for determining the degree of similarity between customers will be described later.

  The product P1 is purchased by the customers C1, C2, C4, and C5, the purchase probability based on the simple total is 4/5 (80%), and the purchase probability weighted by the similarity is 2.4 / 3 (80 %). Further, the product P2 is purchased by the customers C3 and C4, the purchase probability based on the simple total is 2/5 (40%), and the purchase probability weighted by the similarity is 1/3 (33%). . Further, the product P3 is purchased by the customers C1 to C5, the purchase probability based on the simple total is 5/5 (100%), and the purchase probability weighted by the similarity is 3/3 (100%). . Furthermore, the product P4 is purchased by the customer C2, the purchase probability based on the simple total is 1/5 (20%), and the purchase probability weighted by the similarity is 0.8 / 3 (27%). Become. The product P5 is purchased by the customers C1, C3, and C4. The purchase probability based on the simple total is 3/5 (60%), and the purchase probability weighted by the similarity is 1.8 / 3 (60 %).

[Similarity between customers]
Next, the degree of similarity between customers will be described. Here, three examples of the method for determining the similarity between customers are given. These correspond to step 110 shown in FIG.

[Similarity between customers 1]
FIG. 8 is a diagram for explaining the similarity between the target customer (customer C6) and the other customers C1 to C5 based on the purchased product. Here, the similarity between customers is calculated | required from the similarity of the purchased goods of customer C6 who is an object customer, and the other customers C1-C5.
FIG. 8 shows the purchase history (whether or not purchased) of the products P1 to P5 by all the customers C1 to C6 including the target customer C6 and the dice coefficient and cosine similarity as an example of the similarity between customers. Showing the relationship. Here, the dice coefficient is represented by “the number of matches of the proposed products / the average of the number of proposed products”. Further, the cosine similarity is represented by “the product of the vector inner product / vector absolute value of the proposed product”.

  In this example, it is assumed that the customer C6 has purchased the products P2 to P5 but has not purchased the product P1. Note that the purchase histories of the customers C1 to C5 are the same as those shown in FIG.

  In this case, the dice coefficients of the customer C6 with respect to the customers C1 to C5 are “0.57”, “0.57”, “0.86”, “0.75”, and “0.33”, respectively. Further, the cosine similarity of the customer C6 with respect to the customers C1 to C5 is “0.58”, “0.58”, “0.87”, “0.75”, and “0.35”, respectively.

[Similarity between customers 2]
FIG. 9 is a diagram for explaining the similarity between the target customer (customer C6) and the other customers C1 to C5 based on the customer attributes. Here, the similarity between customers is calculated | required from the similarity of the customer attribute of customer C6 who is an object customer, and the other customers C1-C5.
FIG. 9 shows customer attributes (industry, number of employees, region, business relationship) of all customers C1 to C6 including the target customer C6, weights (0 to 1) set for each element of the customer attributes, The relationship between the dice coefficient (simple total) and the dice coefficient (after weighting) as an example of the similarity between customers is shown.

In this example, for example, the customer C6 has the type of industry “manufacturing”, the number of employees “11 to 100 (persons)”, the region “Kanto”, and the business relationship “new customer”. And Detailed descriptions of other customers C1 to C5 are omitted.
In this example, it is assumed that the weight of the business type and the number of employees is “1”, and the weight of the region and the business relationship is “0.5”. However, the weighting value may be changed.

  In this case, the dice coefficients (simple total) of the customer C6 for the customers C1 to C5 are “0.50”, “0.50”, “0.50”, “0.75”, and “0.25”, respectively. It becomes. Further, the dice coefficients (after weighting) of the customer C6 with respect to the customers C1 to C5 are “0.66”, “0.50”, “0.50”, “0.50”, and “0.33”, respectively. Become.

[Similarity between customers 3]
FIG. 10 is a diagram for explaining the similarity between the target customer (customer C6) and other customers C1 to C5 based on machine learning. Here, the degree of similarity between the customer C6 and other customers C1 to C5, determined in advance by some method, the purchase rate prediction of the customer C6 based on the purchase history of the customers C1 to C5, and the actual purchase by the customer C6. The correction is made sequentially based on the presence or absence. More specifically, the update is repeated until the similarity between customers converges, for example, by logarithmic likelihood differential distribution so that the purchase rate prediction is close to the actual purchase / non-purchase.
FIG. 10 shows purchase histories (whether or not purchased) of the products P1 to P5 by the customers C1 to C5 other than the target customer (customer C6), prediction of the purchase rate of the products P1 to P5 by the customer C6, and actual purchases. It shows the relationship between presence / absence and similarity between customers.

  In this example, it is assumed that the customer C6 has purchased the products P2 to P5 after predicting the purchase rate, but has not purchased the product P1. The purchase history of the customers C1 to C5 is the same as that shown in FIG.

  In this case, the purchase rate prediction of the product P1 by the customer C6 is 4/5 (80%), the purchase rate prediction of the product P2 is 2/5 (40%), and the purchase rate prediction of the product P3 is 5/5 (100%). ), The purchase rate prediction of the product P4 is 1/5 (20%), and the purchase rate prediction of the product P5 is 3/5 (60%). Further, the similarities of the customer C6 to the customers C1 to C5 are “0.8”, “0.8”, “0.6”, “0.4”, and “0.4”, respectively.

[Proposed probability]
Next, the proposal probability will be described. Here, two examples of the method for determining the proposal probability are given. These correspond to step 130 shown in FIG.

[Proposal probability 1]
In the first determination method, the target sales (sales S6) obtains a proposal probability based on products previously proposed to customers C1 to C5 other than the target customer (customer C6).

FIG. 11 is a diagram for explaining the proposal probability (part 1) of the products P1 to P5 for the target customer (customer C6) by the target sales (sales S6).
FIG. 11 is based on the proposal history of the products P1 to P5 for the customers C1 to C5 by the sales S6 (whether there is a proposal), the similarity between the customer C6 and the customers C1 to C5 (similarity between customers), and a simple total. The relationship between the proposal probability and the proposal probability when weighting is performed according to the similarity between customers is shown. In FIG. 11, when there is a proposal history, it is expressed as “1”, and when there is no suggestion history, it is expressed as “0” (the same applies below).

  In this example, the similarities of the customers C1 to C5 (similarity between customers) with respect to the customer C6 are “0.8”, “0.8”, “0.6”, “0.4”, “ 0.4 ". The degree of similarity between customers is the same as that shown in FIG.

  The business S6 proposes the product P1 to the customers C1, C2, C4, and C5. The proposal probability based on the simple total is 4/5 (80%), and the proposal probability weighted based on the similarity is 2.4. / 3 (80%). Sales S6 proposes the product P2 to the customers C3 and C4, the proposal probability based on the simple sum is 2/5 (40%), and the proposal probability weighted by the similarity is 1/3 (33 %). Furthermore, the sales S6 proposes the product P3 to the customers C1 to C5, the proposal probability based on the simple sum is 5/5 (100%), and the proposal probability weighted by the similarity is 3/3 (100 %). Furthermore, the sales S6 proposes the product P4 to the customer C2, the proposal probability based on the simple sum is 1/5 (20%), and the proposal probability weighted by the similarity is 0.8 / 3 ( 27%). Sales S6 proposes the product P5 to the customers C1, C3, and C4, the proposal probability based on the simple sum is 3/5 (60%), and the proposal probability weighted based on the similarity is 1.8. / 3 (60%).

[Proposal probability 2]
In the second determination method, the sales probabilities S1 to S5 other than the target sales (sales S6) obtain the proposal probability based on the products proposed to the target customer (customer C6).

FIG. 12 is a diagram for explaining the proposal probability (part 2) of the products P1 to P5 for the target customer (customer C6) by the target sales (sales S6).
FIG. 12 shows the proposal probability of the products P1 to P5 for the customer C6 by the sales S1 to S5, the similarity between the sales S6 and the sales S1 to S5 (similarity between sales), the proposal probability by a simple total, and the sales It shows the relationship with the proposal probability when weighting is performed according to the similarity.

  In this example, the similarities of sales S1 to S5 (similarity between sales) with respect to sales S6 are “0.8”, “0.8”, “0.6”, “0.4”, “ 0.4 ". A method for determining the similarity between sales will be described later.

  Proposal probabilities of the product P1 to the customer C6 by the sales S1 to S5 are “0.8”, “0.4”, “0.2”, “0.6”, and “0”, respectively. The proposal probability by 0.4 is 0.4 (40%) and the proposal probability weighted by similarity is 0.44 (44%). In addition, the proposal probabilities of the product P2 for the customer C6 by the sales S1 to S5 are “0.2”, “0”, “0.2”, “0.4”, and “0.6”, respectively. The proposal probability based on the simple sum is 0.28 (28%), and the proposal probability weighted based on the similarity is 0.23 (23%). Furthermore, the proposal probabilities of the product P3 to the customer C6 by the sales S1 to S5 are “0.6”, “1”, “1”, “0.4”, and “0.8”, respectively. The proposal probability is 0.76 (76%), and the proposal probability weighted by similarity is 0.79 (79%). Furthermore, the proposal probabilities of the product P4 to the customer C6 by the sales S1 to S5 are “0.6”, “1”, “1”, “1”, “0.2”, respectively, and are based on simple totals. The proposal probability is 0.76 (76%), and the proposal probability weighted by the similarity is 0.79 (79%). And the proposal probability of the goods P5 with respect to the customer C6 by sales S1-S5 is "0.2", "0.2", "0.8", "0", "0.4", respectively. The proposal probability based on the simple sum is 0.32 (32%), and the proposal probability weighted based on the similarity is 0.32 (32%).

[Similarity between sales]
Next, the degree of similarity between sales will be described. Here, three examples of the method for determining the similarity between customers are given.

[Similarity between sales 1]
FIG. 13 is a diagram for explaining the similarity between the target business (business S6) and the other business S1 to S5 based on the proposed product. Here, the similarity between sales is calculated | required from the similarity of the proposal goods of sales S6 which is object sales, and other sales S1-S5.
FIG. 13 shows the proposal history (the presence / absence of proposals) of the products P1 to P5 and the dice coefficient and cosine similarity as an example of the similarity between sales by all the sales S1 to S6 including the sales S6 that is the target sales. Showing the relationship.

  In this example, the business S6 proposes products P2 to P5, but does not propose the product P1. In addition, since the proposal history of sales S1-S5 is the same as what was shown in FIG. 12, description is abbreviate | omitted.

  In this case, the dice coefficients of the sales S6 for the sales S1 to S5 are “0.57”, “0.57”, “0.86”, “0.75”, and “0.33”, respectively. Further, the cosine similarity of the business S6 with respect to the business S1 to S5 is “0.58”, “0.58”, “0.87”, “0.75”, and “0.35”, respectively.

[Similarity between sales 2]
FIG. 14 is a diagram for explaining the similarity between the target business (business S6) and the other business S1 to S5 based on the business attributes. Here, the similarity between sales is calculated | required from the similarity of the sales attribute of sales S6 which is object sales, and other sales S1-S5.
FIG. 14 shows the sales attributes (affiliation department, years of sales experience, region, work style) of all sales S1 to S6 including sales S6, which is the target sales, and the weights (0 to 1) set for each element of the sales attributes. And the relationship between the dice coefficient (simple total) and the dice coefficient (after weighting) as an example of the similarity between sales.

In this example, for example, in sales S6, the department to which the department belongs is “Department A”, the sales experience years are “3 to 10 (years)”, the region is “Kanto”, and the working style is “full time”. is there. In addition, detailed description is abbreviate | omitted about other sales S1-S5.
In this example, it is assumed that the weight of the department and the sales experience years is “1”, and the weight of the region and the work style is “0.5”. However, the weighting value may be changed.

  In this case, the dice coefficients (simple total) of sales S6 for sales S1 to S5 are “0.50”, “0.50”, “0.50”, “0.75”, and “0.25”, respectively. It becomes. The dice coefficients (after weighting) of sales S6 for sales S1 to S5 are “0.66”, “0.50”, “0.50”, “0.50”, and “0.33”, respectively. Become.

[Similarity between sales 3]
FIG. 15 is a diagram for explaining the similarity between the target business (business S6) and the other business S1 to S5 based on machine learning. Here, the degree of similarity between the sales S6 and the other sales S1 to S5, determined in advance by some method, is used to estimate the proposal rate of the sales S6 based on the sales history of the sales S1 to S5 and the actual proposal by the sales S6. The correction is made sequentially based on the presence or absence. More specifically, the update is repeated until the similarity between sales is converged by, for example, logarithmic likelihood differential distribution so that the proposal rate prediction is close to the presence or absence of the actual proposal.
FIG. 15 shows the proposal history of the products P1 to P5 (presence / absence of proposals) by the sales S1 to S5 other than the target sales (sales S6), the proposal rate prediction of the products P1 to P5 by the sales S6, and the actual proposal. It shows the relationship between presence / absence and similarity between sales.

  In this example, it is assumed that the business S6 has proposed the products P2 to P5 after predicting the proposal rate, but has not proposed the product P1. In addition, since the purchase history of business S1-S5 is the same as what was shown in FIG. 11, description is abbreviate | omitted.

  In this case, the proposal rate prediction of the product P1 by the sales S6 is 4/5 (80%), the proposal rate prediction of the product P2 is 2/5 (40%), and the proposal rate prediction of the product P3 is 5/5 (100%) ), The proposal rate prediction of the product P4 is 1/5 (20%), and the proposal prediction of the product P5 is 3/5 (60%). The similarities of the sales S6 with respect to the sales S1 to S5 are “0.8”, “0.8”, “0.6”, “0.4”, and “0.4”, respectively.

[Recommendation List]
FIG. 16 is a diagram illustrating an example of a recommendation list.
FIG. 16 ranks the products to be proposed for the target customer (customer C6) and the target sales (sales S6). Here, the product P1 is selected as the first place, the product P3 is selected as the second place, and the product P4 is selected as the third place.

[Summary of Embodiment 1]
In the present embodiment, the purchase probability of the customer is expected to be greater than or equal to a certain value, i.e., purchase by the customer, and the sales proposal probability is equal to or less than the certain value, i.e., sales are not expected. It was selected as a product that should be recommended. As a result, it is possible to increase the effect of the recommendation that the sales person wants the customer to purchase the product.

[Proposed probability]
The purchase history shown in FIG. 5 (a) is generally organized and recorded in many cases, but the proposal history shown in FIG. 5 (b) is generally not present or organized. It can be hidden by other information due to lack of it.
In this case, for example, a proposal history may be extracted from a daily business report described by each business by natural language processing. In addition, even when a product proposed to a customer is not explicitly described, it is possible to infer from the context. Furthermore, when the sales person submits the estimate of the product to the customer, the submission history of this estimate may be regarded as the proposal history.

  Further, when there is no means for estimating the proposal history, the purchase history may be regarded as the proposal history, assuming that at least the purchased product has been proposed by sales. In this case, weighting may be performed in consideration of a possibility that a product that has not been purchased is not proposed, but has not been purchased but has been proposed.

<Embodiment 2>
In the first embodiment, the target sales selects a product to be proposed to the target customer by using the purchase probability of the target customer's product and the proposal probability of the target sales product. In contrast, in the present embodiment, the target sales should propose to the target customer using the purchase probability of each product through the sales of the target customer and the purchase probability of each product not through the sales of the target customer. A product is selected. In the present embodiment, the same components as those in the first embodiment are denoted by the same reference numerals, and detailed description thereof is omitted.

[Functional configuration example of sales support device]
FIG. 17 is a block diagram illustrating a functional configuration example of the sales support apparatus 100 according to the second embodiment.
The sales support apparatus 100 of this embodiment includes a purchase history storage unit 111, a product attribute information storage unit 113, a customer attribute information storage unit 114, a sales attribute information storage unit 115, a purchase probability calculation unit 116, and a useful A sex determination unit 118, a recommendation list generation unit 119, and a recommendation list output unit 120 are provided. That is, the sales support apparatus 100 according to the present embodiment is different from the first embodiment (see FIG. 4) in that the proposal history storage unit 112 and the proposal probability calculation unit 117 are not provided.

Further, the purchase probability calculation unit 116, based on the purchase history acquired from the purchase history storage unit 111, the purchase probability (an example of a recommended index) of each product via the customer's business and each product not via the customer's business Is different from the first embodiment in that the purchase probability (an example of a recommendation index) is calculated.
Further, the usefulness determination unit 118 uses the purchase probability of each product via the customer's sales and the purchase probability of each product not through the customer's sales, calculated by the purchase probability calculation unit 116. The first embodiment is different from the first embodiment in that the usefulness of the relationship between the customer and the customer is determined.

  Here, in the present embodiment, the purchase probability calculation unit 116 functions as an acquisition unit and an output unit, and the usefulness determination unit 118 and the recommendation list generation unit 119 function as an extraction unit.

[Recommendation list output procedure]
FIG. 18 is a flowchart illustrating an example of a recommendation list output procedure according to the second embodiment. Here, as described above, the present embodiment is based on the sixth combination (customer: fixed, product: variable, sales: fixed) shown in FIG. 2, and the target sales (target sales) is the target. This is used when it is desired to decide which product should be proposed to the customer (target customer).

  First, the purchase probability calculation unit 116 acquires a purchase history from the purchase history storage unit 111 and acquires customer attribute information from the customer attribute information storage unit 114. Then, the purchase probability calculation unit 116 calculates the similarity between customers based on the acquired purchase history and customer attribute information (step 210). In step 210, the same processing as in step 110 shown in FIG. 6 is performed.

  Next, the purchase probability calculation unit 116 uses the similarities between customers calculated in step 210 to purchase each product of the customer at the time of the sales proposal based on the purchase tendency through sales of similar customers. A purchase probability is calculated (step 220).

  Subsequently, the purchase probability calculation unit 116 uses the similarities between customers calculated in step 210 to obtain each product of the customer at the time of non-suggestion based on the purchase tendency not through the sales of similar customers. The purchase probability is calculated (step 230).

  And the usefulness determination unit 118 has a purchase probability at the time of sales proposal calculated in step 220 equal to or higher than a certain value, and a purchase probability at the time of non-sales proposal calculated in step 230 is equal to or less than a certain value. The target sales is selected as a product to be proposed (recommended) to the target customer (step 240). Here, in step 240, one or more products may be selected. Thereafter, the recommendation list generation unit 119 generates a recommendation list in which the target customer, the target sales, and the product to be recommended are associated with each other, and the recommendation list output unit 120 generates the recommendation list generated by the recommendation list generation unit 119 as an external Output to.

[Specific Example of Embodiment 2]
Now, the recommendation list output procedure in the present embodiment will be described with a specific example.
Here, as in the first embodiment, a case where sales S6, which is the target business, wants to determine which of the products P1 to P5 is proposed to the customer C6, which is the target customer, is taken as an example.

[Purchase probability at the time of sales proposal]
FIG. 19 is a diagram for explaining the purchase probabilities of the products P1 to P5 through the sales S1 to S6 of the target customer (customer C6). FIG. 19 corresponds to step 220 shown in FIG.
FIG. 19 shows purchase histories (purchase / non-purchase) of products P1 to P5 via sales S1 to S6 by customers C1 to C5 other than the target customer (customer C6), and customer C6 and other customers C1 to C5. And the purchase probability in the case where weighting based on the similarity between customers is performed.

  In this example, the similarities of the customers C1 to C5 (similarity between customers) with respect to the customer C6 are “0.8”, “0.8”, “0.6”, “0.4”, “ 0.4 ".

  The product P1 is purchased by the customers C1, C2, C4, and C5 through the sales operations S1 to S6. The purchase probability based on the simple total is 4/5 (80%), and the purchase probability weighted based on the similarity is 2.4 / 3 (80%). Further, the product P2 is purchased by the customers C3 and C4 through the sales operations S1 to S6. The purchase probability based on the simple total is 2/5 (40%), and the purchase probability weighted by the similarity is 1 / 3 (33%). Further, the product P3 is purchased by the customers C1 to C5 through the sales operations S1 to S6, the purchase probability based on the simple total is 5/5 (100%), and the purchase probability weighted by the similarity is 3 / 3 (100%). Furthermore, the product P4 is purchased by the customer C2 through the sales operations S1 to S6, the purchase probability based on the simple total is 1/5 (20%), and the purchase probability weighted by the similarity is 0.8. / 3 (27%). The product P5 is purchased by the customers C1, C3, and C4 through the sales operations S1 to S6. The purchase probability based on the simple total is 3/5 (60%), and the purchase probability weighted by the similarity is 1.8 / 3 (60%).

[Purchase probability when sales are not proposed]
FIG. 20 is a diagram for explaining the purchase probability of the products P1 to P5 that do not go through the sales S1 to S6 of the target customer (customer C6). FIG. 20 corresponds to step 230 shown in FIG.
FIG. 20 shows purchase histories (whether or not purchased) of products P1 to P5 not through sales S1 to S6 by customers C1 to C5 other than the target customer (customer C6), and customer C6 and other customers C1 to C5. And the purchase probability in the case where weighting based on the similarity between customers is performed.

  In this example, the similarities of the customers C1 to C5 (similarity between customers) with respect to the customer C6 are “0.8”, “0.8”, “0.6”, “0.4”, “ 0.4 ". Note that the degree of similarity between customers is the same as that shown in FIG.

  The product P1 is purchased by the customer C5 without going through the sales S1 to S6, the purchase probability based on the simple total is 1/5 (20%), and the purchase probability weighted by the similarity is 0.4 / 3. (13%). Further, the product P2 is purchased by the customers C1 and C5 without going through the sales operations S1 to S6, the purchase probability based on the simple total is 2/5 (40%), and the purchase probability weighted by the similarity is 1 2/3 (40%). Further, the product P3 is purchased by the customer C4 without going through the sales S1 to S6, the purchase probability based on the simple total is 1/5 (20%), and the purchase probability weighted by the similarity is 0.4. / 3 (13%). Furthermore, the product P4 is purchased by the customer C3 without going through the sales operations S1 to S6, the purchase probability based on the simple total is 1/5 (20%), and the purchase probability weighted by the similarity is 0. 6/3 (20%). The product C5 is purchased by the customer C1 without going through the sales operations S1 to S6, the purchase probability based on the simple total is 1/5 (20%), and the purchase probability weighted by the similarity is 0.8. / 3 (27%).

  Also in the present embodiment, a recommendation list as shown in FIG. 16 is output as in the first embodiment.

[Summary of Embodiment 2]
In the present embodiment, a product having a customer purchase probability at the time of sales proposal of a certain value or more and a customer purchase probability at the time of non-sales proposal of a certain value or less is selected as a product to be recommended. . This is considered to be synonymous with selecting a product that a customer purchases due to the presence of sales. As a result, it is possible to increase the effect of the recommendation that the sales person wants the customer to purchase the product.

<Embodiment 3>
In the second embodiment, the target sales selects the product to be proposed to the target customer by using the purchase probability of the product at the time of the target customer's sales proposal and the purchase probability of the product at the time of the target customer's non-suggestion of sales. It was. In contrast, this embodiment uses the product purchase probability at the time of the target customer's business proposal and the average order rate of the product at the time of the target business proposal to determine the product that the target business should propose to the target customer. It is what you choose. In the present embodiment, the same components as those in the second embodiment are denoted by the same reference numerals, and detailed description thereof is omitted.

[Functional configuration example of sales support device]
FIG. 21 is a block diagram illustrating a functional configuration example of the sales support apparatus 100 according to the third embodiment.
The sales support apparatus 100 of the present embodiment includes a purchase history storage unit 111, a proposal history storage unit 112, a product attribute information storage unit 113, a customer attribute information storage unit 114, a sales attribute information storage unit 115, a purchase A probability calculation unit 116, a usefulness determination unit 118, a recommendation list generation unit 119, a recommendation list output unit 120, and an average order rate calculation unit 121 are provided. That is, the sales support apparatus 100 of the present embodiment is different from the first embodiment (see FIG. 4) in that an average order rate calculation unit 121 is provided instead of the proposal probability calculation unit 117.

Further, the purchase probability calculation unit 116 is different from the first embodiment in that the purchase probability of each product is calculated through the sales of the customer based on the purchase history acquired from the purchase history storage unit 111.
Further, the usefulness determination unit 118 uses the purchase probability of each product through the sales of the customer calculated by the purchase probability calculation unit 116 and the average order rate calculated by the average order rate calculation unit 121 to It differs from the first embodiment in that the usefulness of the relationship between the product and the customer is determined.

  Based on the purchase history acquired from the purchase history storage unit 111 and the proposal history acquired from the proposal history storage unit 112, the average order rate calculation unit 121 calculates the average order rate (an example of a recommended index) of each product at the time of business proposal. calculate.

  Here, in the present embodiment, the purchase probability calculation unit 116 and the average order rate calculation unit 121 function as an acquisition unit and an output unit, and the usefulness determination unit 118 and the recommendation list generation unit 119 serve as an extraction unit. It is functioning.

[Recommendation list output procedure]
FIG. 22 is a flowchart illustrating an example of a recommendation list output procedure according to the third embodiment. Here, as described above, the present embodiment is based on the sixth combination (customer: fixed, product: variable, sales: fixed) shown in FIG. 2, and the target sales (target sales) is the target. This is used when it is desired to decide which product should be proposed to the customer (target customer).

  First, the purchase probability calculation unit 116 acquires a purchase history from the purchase history storage unit 111 and acquires customer attribute information from the customer attribute information storage unit 114. Then, the purchase probability calculation unit 116 calculates the degree of similarity between customers based on the acquired purchase history and customer attribute information (step 310). In step 310, the same processing as in step 110 shown in FIG. 6 is performed.

  Next, the purchase probability calculation unit 116 calculates the purchase probability of each product of the customer at the time of business proposal based on the purchase tendency of similar customers obtained using the similarity between customers calculated in step 310. (Step 320). In step 320, the same processing as in step 220 shown in FIG. 18 is performed.

  Subsequently, the average order rate calculation unit 121 acquires a purchase history from the purchase history storage unit 111 and also acquires a proposal history from the proposal history storage unit 112. Then, the average order rate calculation unit 121 calculates the average order rate of each product at the time of business proposal based on the acquired proposal history and purchase history (step 330).

  The usefulness determination unit 118 then selects a product having a purchase probability at the time of business proposal calculated at step 320 that is equal to or higher than a certain value, and an average order rate at the time of business proposal calculated at step 330 is equal to or higher than a predetermined value. The target sales is selected as a product to be proposed (recommended) to the target customer (step 340). Here, in step 340, one or more products may be selected. Thereafter, the recommendation list generation unit 119 generates a recommendation list in which the target customer, the target sales, and the product to be recommended are associated with each other, and the recommendation list output unit 120 generates the recommendation list generated by the recommendation list generation unit 119 as an external Output to.

[Specific Example of Embodiment 3]
Now, the recommendation list output procedure in the present embodiment will be described with a specific example.
In addition, here, as in the first and second embodiments, a case where the sales S6, which is the target sales, wants to determine which of the products P1 to P5 is proposed to the target customer C6 is taken as an example. To do.

[Average order rate]
FIG. 23 is a diagram for describing an average order rate of products P1 to P5 of the target business (business S6). FIG. 23 corresponds to step 330 shown in FIG.
FIG. 23 shows the relationship between the presence / absence of proposals for products P1 to P5 by the target sales S6 (proposal / no proposal) and the presence / absence of orders (orders / missed) by customers C1 to C5 other than the target customer (customer C6) The relationship between the degree of similarity between the customer C6 and the other customers C1 to C5 (similarity between customers), the average order rate based on the simple total, and the average order rate when weighting is performed based on the degree of similarity between customers. Show.

  In this example, the similarities of the customers C1 to C5 (similarity between customers) with respect to the customer C6 are “0.8”, “0.8”, “0.6”, “0.4”, “ 0.4 ".

  For example, for the product P3, the sales S6 proposes to the customers C1 and C5 and receives an order (denoted as “proposal-order”). The product P3 is not proposed to the customer C2 by the sales S6 (indicated as “no proposal”). Further, the product P3 has been proposed by the sales S6 to the customers C3 and C4, but has been lost (denoted as “suggestion-rejection”). In this case, the average order rate of the product P3 at the time of business proposal by simple total is 2/4 (50%), and the average order rate of the product P3 at the time of business proposal by weighting by similarity is 1.2 / 2. .2 (55%). In addition, detailed description is abbreviate | omitted about the other goods P1, P2, P4, P5.

  Also in the present embodiment, a recommendation list as shown in FIG. 16 is output as in the first embodiment.

[Summary of Embodiment 3]
In this embodiment, a product having a purchase probability at the time of business proposal of a certain value or more and an average order rate at the time of business proposal of a certain value or more is selected as a product to be recommended. Here, a high average order rate at the time of business proposal is considered to be synonymous with a high proposal ability (skill) of the business. As a result, it is possible to increase the effect of the recommendation that the sales person wants the customer to purchase the product.

<Embodiment 4>
In the first to third embodiments, the recommendation list is generated by the target sales selecting a product to be proposed to the target customer. On the other hand, as described above, the present embodiment is based on the seventh combination (customer: fixed, product: fixed, sales: variable) shown in FIG. This is used when it is desired to determine which business should be proposed for the target product (target product). In addition, in this Embodiment, about the thing similar to Embodiment 1-3, the same code | symbol is attached | subjected and the detailed description is abbreviate | omitted.

[Functional configuration example of sales support device]
FIG. 24 is a block diagram illustrating a functional configuration example of the sales support apparatus 100 according to the fourth embodiment.
The sales support apparatus 100 of the present embodiment includes a purchase history storage unit 111, a proposal history storage unit 112, a product attribute information storage unit 113, a customer attribute information storage unit 114, a sales attribute information storage unit 115, a purchase A probability calculation unit 116, an average order rate calculation unit 121, a cooperation necessity determination unit 122, a request candidate determination unit 123, and a request candidate output unit 124 are provided. That is, the sales support apparatus 100 according to the present embodiment replaces the usefulness determination unit 118, the recommendation list generation unit 119, and the recommendation list output unit 120 with a cooperation necessity determination unit 122, a request candidate determination unit 123, and a request candidate output. The third embodiment is different from the third embodiment (FIG. 21) in that the portion 124 is provided.

  The cooperation necessity determination unit 122 uses the purchase probability at the time of business proposal calculated by the purchase probability calculation unit 116 and the average order rate at the time of business proposal calculated by the average order rate calculation unit 121 to cooperate with other sales. Whether or not is necessary is determined.

  The request candidate determination unit 123 determines which business is requested to cooperate when it is determined by the cooperation necessity determination unit 122 that cooperation by another business is necessary.

  The request candidate output unit 124 outputs a business request candidate based on the determination result by the request candidate determination unit 123.

  Here, in the present embodiment, the purchase probability calculation unit 116 and the proposal probability calculation unit 117 function as an acquisition unit and an output unit, and the cooperation necessity determination unit 122 and the request candidate determination unit 123 serve as an extraction unit. It is functioning.

[Recommendation list output procedure]
FIG. 25 is a flowchart illustrating an example of a recommendation list output procedure according to the fourth embodiment.
First, a purchase history is acquired from the purchase history storage unit 111, and customer attribute information is acquired from the customer attribute information storage unit 114. Then, the purchase probability calculation unit 116 calculates the degree of similarity between customers based on the acquired purchase history and customer attribute information (step 410). In step 410, the same processing as in step 110 shown in FIG. 6 is performed.

  Next, the purchase probability calculation unit 116 predicts the purchase probability of each product of the customer at the time of business proposal based on the purchase tendency of similar customers obtained using the similarity between customers calculated in step 410. (Step 420). In step 420, the same processing as in step 220 shown in FIG. 18 is performed.

  Then, the cooperation necessity determination unit 122 selects a product having the highest purchase probability at the time of the business proposal calculated in Step 420 (Step 430).

  Subsequently, the average order rate calculation unit 121 acquires a purchase history from the purchase history storage unit 111 and also acquires a proposal history from the proposal history storage unit 112. Then, the average order rate calculation unit 121 calculates the average order rate of each product at the time of business proposal based on the acquired proposal history and purchase history (step 440). In step 440, the same processing as step 330 shown in FIG. 22 is performed.

  And the cooperation necessity determination part 122 determines whether the average order-receiving rate of the goods (selected goods) selected at step 430 is more than a fixed value (step 450). When an affirmative (YES) determination is made in step 450, the cooperation necessity determination unit 122 recommends the selected product to the target business (step 460), and the process is completed.

  On the other hand, if a negative (NO) determination is made in step 450, the request candidate determining unit 123 determines another business with a high average order rate for the selected product as a cooperation request candidate (step 470). Thereafter, the request candidate output unit 124 outputs the request candidates (recommendation list) determined in step 470 to the outside.

[Summary of Embodiment 4]
In the present embodiment, sales with a high average order rate are selected as supporters for products with a high purchase probability by customers. As a result, it is possible to increase the effect of the recommendation that the sales person wants the customer to purchase the product.

<Others>
In each of the above-described embodiments, the description has been made based on the relationship between the product, the customer (purchaser), and the sales (recommendant), but the scope of application of the present invention is not limited to this. For example, the method described above can also be applied to a case where friends (one is a recommender and the other is a purchaser) and the first person recommends a product to the second person.

  Moreover, in each embodiment mentioned above, although the sales support apparatus 100 was implement | achieved by one apparatus, it is good also as dividing and implement | achieving in a several apparatus. For example, various function units for performing computation may be realized by a client device, and for example, various storage units for storing information may be realized by a server device connected to the client device via a network.

DESCRIPTION OF SYMBOLS 100 ... Sales support apparatus, 111 ... Purchase history storage part, 112 ... Proposal history storage part, 113 ... Merchandise attribute information storage part, 114 ... Customer attribute information storage part, 115 ... Sales attribute information storage part, 116 ... Purchase probability calculation part DESCRIPTION OF SYMBOLS 117 ... Proposal probability calculation part, 118 ... Usefulness determination part, 119 ... Recommendation list production | generation part, 120 ... Recommendation list output part, 121 ... Average order rate calculation part, 122 ... Cooperation necessity determination part, 123 ... Request candidate determination Part, 124 ... request candidate output part

Claims (7)

On the computer,
A function of acquiring information on a product recommendation purchase history in which a plurality of products, a plurality of recommenders who recommend products to customers, and a plurality of purchasers who have purchased products are associated with each other;
A function of outputting a recommendation index based on each product recommended by each recommender from each acquired product recommended purchase history information;
A program for realizing a function of extracting a combination of a product, a recommender, and a purchaser, wherein the recommendation index satisfies a predetermined criterion.   The program according to claim 1, wherein the recommendation index is a probability that one recommender is expected to propose one product to one customer.   The recommendation index includes a probability that, when one recommender proposes one product to one customer, the one customer is expected to purchase the one product, and the one recommender is the one customer. The program according to claim 1, wherein the probability is that the one customer is predicted to purchase the one product when the one product is not proposed.   The program according to claim 1, wherein the recommendation index is a probability that the one customer actually purchases the one product when one recommender proposes one product to one customer. . The function to output further outputs other recommended indicators based on each product purchased by each purchaser,
2. The extracting function extracts a combination of one product, one recommender, and one purchaser, wherein the other recommendation index further satisfies another predetermined criterion. The program of any one of thru | or 4.   6. The other recommendation index is a probability that, when one recommender proposes one product to one customer, the one customer purchases the one product. program. An acquisition unit that acquires information on a product recommendation purchase history that associates a plurality of products, a plurality of recommenders that recommend products to customers, and a plurality of purchasers that have purchased the products;
From the acquired product recommended purchase history information, an output unit that outputs a recommended index based on each product recommended by each recommender to each purchaser;
A combination extraction system including an extraction unit that extracts a combination of a product, a recommender, and a purchaser, wherein the recommendation index satisfies a predetermined criterion.

Images