Measure formulation assistance device, method, and program

Abstract

This measure formulation assistance device: acquires data pertaining to a purchase in a store, and instruction information indicating the trend of a sales promotion measure that includes the type of a measure target and an index to be improved by implementing the measure; aggregates store data for each segment which is a combination of classification attribute values corresponding to a combination of classification attributes of the store data according to the instruction information; extracts, from the aggregated segments, a set of a measure target segment that serves as an implementation target of the measure and a measure goal segment that serves as a goal of the measure; calculates a measure expectation effect to be expected when the measure is performed on the measure target segment for each set; and outputs the calculated measure expectation effect in association with the set of the measure target segment and the measure goal segment.

Description

Policy formulation support device, method, and program

[0001] The disclosed technology relates to a policy formulation support device, a policy formulation support method, and a policy formulation support program.

[0002] In the marketing of physical stores, when planners of sales promotion measures, such as store operators or analysts of store data related to purchases at the store, formulate measures that are consistent with the direction of the measures, they formulate the measures through observing the store and analyzing the store data. The direction of the measures is a requirement for implementing the measures, and is mainly the type of measure target when implementing the measures, the measure KPI (Key Performance Indicator) that is to be improved by the measures, etc. The types of measure targets are, for example, products, purchases, customers, etc., and the measure KPIs are, for example, purchase amounts, number of purchases, number of store visits, etc. Therefore, the direction of the measures might be, for example, "We want to increase the number of purchases (measure KPI) of products (type of measure target)."

[0003] For example, in the method of Non-Patent Document 1, segments that are combinations of classification attribute values ​​are extracted for the type of campaign target based on the classification attributes and classification attribute values ​​in store data. Then, in this method, a segment with a low value of the campaign KPI that is to be improved is selected from the extracted segments, and sales promotion campaigns are implemented with that segment as the campaign target.

[0004] In addition, the method of Non-Patent Document 2 utilizes the characteristics of each customer to estimate products and sales times, and implements sales promotion measures.

[0005] Michiko Watanabe, "Classification Techniques for Large-Scale Data - Behavioral Metrics Utilization of Latent Class Models," Proceedings of the 1st Conference of the Japan Society of Interdisciplinary Studies, 2005, Vol. 2005, Session ID K2-54, p. 223, Published June 27, 2006; Tomoyuki Yamane, Kotaro Sugawara, Naoki Nishimura, "Analysis of Product Promotion Effects Using Time Series Models," Communications of the Operations Research Society of Japan, 61(2), February 2016, Operations Research Society of Japan, pp. 65-70

[0006] With conventional technology, it is possible to target a campaign on segments with low campaign KPI values ​​or products that customers are likely to purchase. However, when there are many types of classification attributes and classification attribute values ​​in store data, and when there are many combinations of classification attribute values, it is difficult to predict in advance the campaign effects, such as the improvement in campaign KPI values, that will be achieved by implementing a campaign for all segments, and to find segments that will be highly effective. Similarly, it is also difficult to find high-quality segments that can be used as campaign targets, such as those with high campaign KPI values. Therefore, it is difficult to select campaign targets that will be highly effective depending on the direction of the campaign.

[0007] For example, let's say that by focusing on the classification attributes of store data, such as purchase time period (morning, noon, afternoon, evening, etc.) and purchased product category, store observations and store data reveal that many people purchase bread in the morning, but few people purchase beverages in the morning, and even fewer people purchase sweets in the morning. It is possible to select the segment with fewer purchases, "morning sweets purchases," as the target of a measure. However, it is difficult to determine from the store data whether a measure for "morning sweets purchases" or "morning beverage purchases" will be more effective.

[0008] The disclosed technology has been made in consideration of the above points, and aims to support the selection of policy targets that will have high policy effects depending on the policy direction.

[0009] A first aspect of the present disclosure is a policy development support device that includes: an acquisition unit that acquires store data, which is data related to purchases at a store, and instruction information that indicates the direction of a sales promotion policy, including the type of policy target and indicators that are to be improved by implementing the policy; an aggregation unit that aggregates the store data for each segment, which is a combination of classification attribute values ​​corresponding to a combination of classification attributes of the store data, in accordance with the instruction information; an extraction unit that extracts from the aggregated segments pairs of policy target segments that are the targets of the policy implementation and policy target segments that are the goals of the policy; a calculation unit that calculates, for each pair, an expected policy effect that is expected when a policy is implemented for the policy target segment; and an output unit that outputs the calculated expected policy effect in association with the pair of the policy target segment and the policy target segment.

[0010] A second aspect of the present disclosure is a policy formulation support method executed by a policy formulation support device including an acquisition unit, an aggregation unit, an extraction unit, a calculation unit, and an output unit, in which the acquisition unit acquires store data, which is data related to purchases at a store, and instruction information indicating the direction of a sales promotion policy, including the type of policy target and an indicator to be improved by implementing the policy, the aggregation unit aggregates the store data by segment, which is a combination of classification attribute values ​​corresponding to a combination of classification attributes of the store data, in accordance with the instruction information, the extraction unit extracts from the aggregated segments pairs of policy target segments, which are the targets of the policy implementation, and policy target segments, which are the goals of the policy, the calculation unit calculates, for each pair, an expected policy effect that is expected when a policy is implemented for the policy target segment, and the output unit outputs the calculated expected policy effect in association with the pair of the policy target segment and the policy target segment.

[0011] A third aspect of the present disclosure is a policy development support program for causing a computer to function as each unit of the policy development support device.

[0012] The disclosed technology can assist in the selection of a policy target that will have a high policy effect depending on the policy direction.

[0013] 1 is a block diagram showing a hardware configuration of a policy formulation support device. FIG. 2 is a block diagram showing an example of a functional configuration of a policy formulation support device. FIG. 3 is a diagram showing an example of store data in this embodiment. FIG. 4 is a diagram showing an example of a counting result in this embodiment. FIG. 5 is a diagram showing an example of an extraction result in this embodiment. FIG. 6 is a diagram showing an example of support information in this embodiment. FIG. 7 is a diagram showing an example of a list of support information in this embodiment. FIG. 8 is a diagram showing an example of a generation definition of policy contents in this embodiment. FIG. 9 is a diagram showing an example of a list display of rankings of combinations of classification attributes in this embodiment. FIG. 10 is a diagram showing an example of a generation definition of policy examples in this embodiment. FIG. 11 is a diagram showing an example of a list display of rankings of support information in this embodiment. FIG. 12 is a diagram showing an example of an evaluation result in this embodiment. FIG. 13 is a flowchart showing an example of policy formulation support processing. FIG. 14 is a flowchart showing an example of an output processing. FIG. 15 is a flowchart showing an example of an evaluation processing. FIG. 16 is a diagram showing an example of store data in another application example. FIG. 17 is a diagram showing an example of a counting result in another application example. FIG. 18 is a diagram showing an example of an extraction result in another application example. FIG. 19 is a diagram showing an example of a list of support information in another application example. FIG. 19 is a diagram showing an example of a generation definition of policy contents in another application example. FIG. 19 is a diagram showing an example of a list display of rankings of combinations of classification attributes in another application example. FIG. 19 is a diagram showing an example of a generation definition of policy examples in another application example. FIG. 19 is a diagram showing an example of a list display of rankings of support information in another application example.

[0014] An example of an embodiment of the disclosed technology will be described below with reference to the drawings. Note that the same or equivalent components and parts in each drawing are given the same reference numerals. Also, the dimensional proportions in the drawings are exaggerated for the convenience of explanation and may differ from the actual proportions.

[0015] 1 is a block diagram showing the hardware configuration of a policy formulation support device 10 according to this embodiment. As shown in FIG. 1, the policy formulation support device 10 includes a CPU (Central Processing Unit) 11, a ROM (Read Only Memory) 12, a RAM (Random Access Memory) 13, a storage 14, an input unit 15, a display unit 16, and a communication I / F (Interface) 17. Each component is connected to each other via a bus 19 so as to be able to communicate with each other.

[0016] The CPU 11 is a central processing unit that executes various programs and controls each component. That is, the CPU 11 reads programs from the ROM 12 or the storage 14 and executes the programs using the RAM 13 as a work area. The CPU 11 controls the above components and performs various arithmetic processing in accordance with the programs stored in the ROM 12 or the storage 14. In this embodiment, the ROM 12 or the storage 14 stores a policy formulation support program, which will be described later.

[0017] The ROM 12 stores various programs and various data. The RAM 13 temporarily stores programs or data as a working area. The storage 14 is configured by a storage device such as an HDD (Hard Disk Drive) or an SSD (Solid State Drive), and stores various programs including an operating system and various data.

[0018] The input unit 15 includes a pointing device such as a mouse and a keyboard, and is used to input various types of information. The display unit 16 is, for example, a liquid crystal display, and displays various types of information. The display unit 16 may also function as the input unit 15 by employing a touch panel system.

[0019] The communication I / F 17 is an interface for communicating with other devices, and the communication may use a wired communication standard such as Ethernet (registered trademark) or FDDI, or a wireless communication standard such as 4G, 5G, or Wi-Fi (registered trademark).

[0020] Next, the functional configuration of the policy formulation support device 10 will be described. FIG. 2 is a block diagram showing an example of the functional configuration of the policy formulation support device 10. As shown in FIG. 2, the policy formulation support device 10 includes, as its functional configuration, an acquisition unit 21, a counting unit 22, an extraction unit 23, a calculation unit 24, and an output unit 25. Furthermore, a predetermined storage area of ​​the policy formulation support device 10 stores a store data DB (database) 31, a counting result DB 32, an extraction result DB 33, and a support information DB 34. Each functional configuration is realized when the CPU 11 reads out a policy formulation support program stored in the ROM 12 or the storage 14, expands it into the RAM 13, and executes it.

[0021] As shown in FIG. 2, the policy formulation support device 10 receives input of policy policy information indicating the direction of a sales promotion policy. The policy formulation support device 10 aggregates store data for each segment in accordance with the input policy policy information. The policy formulation support device 10 extracts, from the aggregated segments, pairs of policy target segments, which are segments targeted by the policy implementation, and policy target segments, which are segments that are the targets of the policy. Furthermore, for each pair of segments, the policy formulation support device 10 calculates the expected effect of the policy that is expected when a policy is implemented for the policy target segment, with the policy KPI of the policy target segment as the target. The policy formulation support device 10 then outputs support information that associates the calculated expected effect of the policy with the extracted pair of policy target segment and policy target segment.

[0022] Sales promotion measures are measures implemented in stores to promote sales. Measure direction refers to the requirements for implementing a measure, such as the type of measure target when implementing the measure and the measure KPIs that are to be improved by the measure, and measure policy information specifies these requirements. Measure target type refers to the type of target for which the measure is implemented, such as customers, purchases, or products. If the measure target is customers, measures are formulated to be implemented for one customer; if the measure target is purchases, measures are formulated to be implemented for one purchase; and if the measure target is products, measures are formulated to be implemented for one product. Measure KPIs are indicators whose values ​​are to be improved by implementing the measure, such as purchase amount, number of purchases, or number of store visits.

[0023] Furthermore, store data is data related to purchases made at a store. Classification attributes are attributes in the store data related to customers, purchases, products, etc. The classification attributes are attributes used to classify data during data analysis. In this embodiment, the target of a measure is narrowed down by combining classification attributes. For example, classification attributes of store data related to "customers" include gender, age, etc. For example, classification attributes of store data related to "purchases" include purchase dates, time periods, etc. For example, classification attributes of store data related to "products" include concurrent purchase patterns, number of products purchased, etc. Classification attribute values ​​are specific values ​​of the classification attributes. For example, classification attribute values ​​for the classification attribute "time period" are morning, noon, evening, night, etc. For example, classification attribute values ​​for the classification attribute "concurrent purchase pattern" are coffee only, coffee and bread, beverages and meals, etc.

[0024] A segment is a collection of data obtained by aggregating store data for each combination of classification attribute values ​​corresponding to a combination of classification attributes. For example, if store data is aggregated by a combination of classification attribute values ​​for a combination of classification attributes (e.g., "time of day" and "dual purchase pattern") on a purchase basis, a collection of data for purchases of only coffee in the morning will form one segment. By aggregating store data for each segment, the frequency of each segment can be determined. The frequency of a segment is the number of data items included in that segment (the number of purchases of only coffee in the morning).

[0025] Furthermore, a measure target segment is a segment for which a measure is implemented, and a measure target segment is a segment that is the target when implementing a measure targeting the measure target segment. For example, if the segment "Coffee only in the morning" is the measure target segment, and a measure to target this measure target segment is to sell a set of coffee and bread, then the segment "Coffee and bread in the morning" could be the measure target segment.

[0026] Furthermore, the expected effect of a measure is the effect of the measure that is expected before the measure is implemented. The effect of the measure is the improvement in the measure KPI due to the measure. Specifically, it is the difference between the measure KPI of the target segment before the measure and the measure KPI after the measure is implemented. For example, if the measure KPI is purchase amount, the effect of the measure is the total purchase amount gained by implementing the measure. More specifically, if the total purchase amount of the target segment before the measure is implemented is 100,000 yen and the total purchase amount of the target segment after the measure is implemented is 200,000 yen, the effect of the measure is 100,000 yen (= 200,000 yen - 100,000 yen).

[0027] Sales promotion measures are formulated based on information about what measures are likely to be effective when implemented in what segments, obtained from the policy formulation support device 10. Each functional unit of the policy formulation support device 10 will be described in detail below.

[0028] The store data DB 31 stores store data for each predetermined classification attribute related to customers, purchases, products, etc.

[0029] The acquisition unit 21 acquires the policy information input by the user. As described above, the policy information includes information such as the policy target type when implementing the policy, the policy KPI to be improved by the policy, etc. In this embodiment, an example will be described in which the user is considering "increasing purchase sales (purchase amount per purchase)" as the direction of the sales promotion policy. In this example, policy information including "purchase" as the policy target type and "purchase amount" as the policy KPI is acquired.

[0030] The acquiring unit 21 also acquires store data (store data recorded in units of "purchases" in this example) for a predetermined period (e.g., one month's worth) from the store data DB 31 in accordance with the acquired policy information. FIG. 3 illustrates an example of store data recorded in units of "purchases." In the example of FIG. 3 , each row (each record) corresponds to purchase data for one purchase, and each purchase data includes information on a "purchase ID," "gender," "generation," "date," "time period," "concurrent purchase pattern," "number of purchased items," and "purchase amount." The "purchase ID" is an identifier for the purchase data. The "gender," "generation," "date," "time period," "concurrent purchase pattern," and "number of purchased items" are examples of classification attributes. These are information indicating the gender and generation of the customer who made the purchase, the date and time period when the purchase was made, the category of items purchased simultaneously, and the number of items purchased simultaneously, respectively. The "purchase amount" is an example of an indicator that serves as a policy KPI.

[0031] The acquisition unit 21 transfers the acquired policy information and store data in units of “purchase” to the aggregation unit 22 .

[0032] The aggregation unit 22 aggregates the store data for each "purchase" unit passed from the acquisition unit 21 based on the policy direction information passed from the acquisition unit 21. Specifically, for each combination of two classification attributes included in the store data for each "purchase", the aggregation unit 22 aggregates data based on the policy KPI for each segment, which is a combination of classification attribute values ​​of the combination of classification attributes. For example, based on the policy KPI being "purchase amount", the aggregation unit 22 aggregates values ​​related to the "purchase amount" of the same segment for each combination of classification attributes of the store data for each "purchase". The aggregation unit 22 may calculate, for example, the median purchase amount or the frequency of purchases (number of purchases) as the aggregated value of the "purchase amount".

[0033] 4 shows an example of the aggregation results for a combination of the classification attributes "time of day" and "multiple purchase pattern" in store data broken down by "purchase." In FIG. 4, for example, the segment "morning, coffee" indicates that the classification attribute value of the classification attribute "time of day" is "morning" and the classification attribute value of the classification attribute "multiple purchase pattern" is "coffee." Furthermore, the example in FIG. 4 shows that, for purchases in the segment "morning, coffee," for example, when store data broken down by "purchase" for one month is aggregated, the median purchase amount is "250 yen" and the frequency of purchases is "1,000."

[0034] Note that, when the policy direction information received from the acquisition unit 21 does not include any requirements other than the policy target type and the policy KPI, the aggregation unit 22 aggregates all combinations of the two classification attributes included in the store data for each "purchase" in the same manner as described above. Furthermore, when the policy direction information includes a specification of classification attributes to be combined, the aggregation unit 22 aggregates the specified combinations of classification attributes in the same manner as described above. The aggregation unit 22 stores the aggregation results in the aggregation result DB 32 and passes the policy direction information to the extraction unit 23. In the example, there are two classification attributes, but the number of classification attributes may be three or more (e.g., "time period," "multiple purchase patterns," "age group," etc.). Furthermore, the number of classification attributes to be combined can be specified in the policy direction information. When the number of classification attributes is specified, the aggregation unit 22 aggregates all combinations of the specified number of classification attributes included in the store data for each "purchase."

[0035] The extraction unit 23 acquires the corresponding aggregation results from the aggregation result DB 32 based on the policy direction information passed from the aggregation unit 22. The extraction unit 23 extracts, from the segments included in the acquired aggregation results, pairs of policy target segments, which are segments targeted for implementing policies, and policy target segments, which are segments that serve as targets when implementing policies targeting the policy target segments, based on the policy direction information.

[0036] Specifically, the extraction unit 23 extracts, from pairs of two segments included in the aggregation results, pairs of segments that satisfy predetermined extraction conditions. The predetermined extraction conditions include statistical conditions based on statistical differences in the policy KPIs between the segments and characteristic conditions related to the similarity between the segments. Furthermore, the extraction unit 23 sets the segment with the lower policy KPI as the policy target segment and the segment with the higher policy KPI as the policy target segment.

[0037] More specifically, the extraction unit 23 may use as a statistical condition that there is a statistical difference in the distribution of the campaign KPIs between the segments, for example, that there is a significant difference and the effect size is greater than a threshold. The presence or absence of a significant difference may be determined, for example, by performing a Steele-Duwas multiple comparison test on the medians of the campaign KPIs between the campaign target segment and the campaign target segment to determine whether the difference in the medians of the campaign KPIs is statistically significant. The effect size may be calculated using, for example, the Cliff delta method, which compares the magnitude of all segment pairs included in the aggregated results. The extraction unit 23 may use as a characteristic condition that the campaign can bring the campaign target segment closer to the campaign target segment, i.e., that it can change purchasing behavior. For example, the characteristic condition may be that the classification attribute values ​​of the classification attribute "time zone" are the same and that the classification attribute values ​​of the classification attribute "multiple purchase pattern" are in an inclusive relationship.

[0038] The extraction unit 23 stores the extraction results of pairs of policy target segments and policy target segments extracted from the tabulation results in the extraction result DB 33. Figure 5 shows an example of the extraction results in which pairs of policy target segments and policy target segments are extracted from an example of the tabulation results for the combination of the classification attributes "time period" and "concurrent purchase pattern" shown in Figure 4. In the example of Figure 5, the extraction result DB 33 stores, for each "combination of classification attributes," the "segment," "median purchase amount," and "frequency" of each of the "policy target segment" and the "policy target segment" as extraction results. In addition, in the example of Figure 5, the "significance" and "effect size" obtained to determine the above statistical conditions are also stored as "segment pair statistics."

[0039] 5, an example of extraction processing will be described under extraction conditions in which the statistical conditions are a significant difference and an effect size > 0.15, and the characteristic conditions are that the classification attribute values ​​for "time of day" are the same and the classification attribute values ​​for "combined purchasing pattern" are in an inclusive relationship (classification attribute value 1 is a proper subset of classification attribute value 2). Note that classification attribute value 1 is the classification attribute value for the "combined purchasing pattern" of the segment set as the target segment, and classification attribute value 2 is the classification attribute value for the "combined purchasing pattern" of the segment set as the target segment. Also, the inclusive relationship here does not refer to an inclusive relationship regarding the content of product categories, but rather means that when each product category included in a combination purchasing pattern is considered as an element of a set, the product categories of one segment are a subset of the product categories of the other segment.

[0040] In this example, the extraction unit 23 first extracts a pair of two segments from the segments included in the tabulation result shown in Fig. 4. For example, if the extraction unit 23 extracts a pair of segments, "morning, coffee" and "morning, beverage," the segment "morning, beverage" with the lowest median of the campaign KPI (here, purchase amount) within this pair of segments is set as the campaign target segment, and the segment "morning, coffee" with the highest median is set as the campaign target segment. Since the classification attribute values ​​"coffee" and "beverage" of the "dual purchase pattern" are not in an inclusive relationship between "morning, coffee" and "morning, beverage" and the characteristic condition is not satisfied, the extraction unit 23 does not extract this pair of segments.

[0041] Furthermore, when the extraction unit 23 extracts the segment pair "morning, coffee" and "morning, coffee and bread" from the aggregation results, it sets the measure target segment to "morning, coffee" and the measure target segment to "morning, coffee and bread" based on the measure KPI. In this case, there is a significant difference, and the effect size of 0.4 is greater than or equal to the threshold, so the statistical condition is met. Furthermore, the classification attribute value of "time period" is the same as "morning," and the classification attribute value of "combined purchase pattern" is in an inclusive relationship in which "coffee" is a proper subset of "coffee and bread," so the characteristic condition is also met. Therefore, the extraction unit 23 extracts this segment pair and stores it in the extraction result DB 33.

[0042] In the above, we have explained a case where the median of purchase amount is used as the campaign KPI, the Steele-Dwas multiple comparison test is used to test for significance, and the Cliff delta method is used to measure the effect size. This is because it is assumed that normality is not necessarily guaranteed for the statistical distribution of purchase amount. If normality is guaranteed for the purchasing trends of the target stores, the average purchase amount can be used as the campaign KPI, and a parametric multiple comparison test such as the Tukey method can be used, and Cohen's d can be used as the effect size. Any method can be used as long as it can measure the statistical difference between the two groups, the campaign target segment and the campaign target segment.

[0043] When the extraction unit 23 has finished extracting pairs of policy target segments and policy target segments for all the tabulation results of the combinations of classification attributes and storing them in the extraction result DB 33 , it passes the policy direction information to the calculation unit 24 .

[0044] The calculation unit 24 acquires the corresponding extraction results from the extraction result DB 33 based on the policy direction information passed from the extraction unit 23. For each pair of a policy target segment and a policy target segment included in the acquired extraction results, the calculation unit 24 calculates an expected policy effect that is expected when a policy is implemented for the policy target segment, with the policy target segment as the policy KPI. Specifically, the calculation unit 24 calculates, as the expected policy effect, the amount of change in the policy KPI of the policy target segment that is expected when a policy is implemented so as to bring the policy KPI for the policy target segment closer to the policy KPI for the policy target segment.

[0045] More specifically, for a pair of a policy target segment and a policy target segment, the calculation unit 24 calculates the expected effect of the policy for the policy target segment from the difference in the policy KPIs of the two segments and the similarity between the two segments, with the aim of bringing the policy target segment closer to the policy target segment.

[0046] For example, suppose the calculation method for the expected effect of a campaign is defined as shown below. Expected effect of a campaign = a x b x c - d a: Frequency value of the segment with the lowest median purchase amount b: Median purchase amount of the segment with the highest median purchase amount - Median purchase amount of the segment with the lowest median purchase amount c: Response rate (or similarity) to the campaign d: Cost amount related to the campaign

[0047] The above definition is used to calculate the expected effect of a measure from the perspective of the median purchase amount of purchases with a frequency (number of purchases) belonging to the measure target segment approaching the median purchase amount of the measure target segment as a result of the measure.Assuming that the measure will not necessarily bring purchases belonging to all measure target segments closer to the median purchase amount of the measure target segment, the expected effect of the measure is calculated by dividing by the response rate c to the measure, which is set to a value of 0<c<1, and subtracting the cost amount related to the measure.

[0048] In this case, the calculation unit 24 extracts, for example, a combination of "morning coffee" as the measure target segment and "morning coffee and bread" as the measure target segment from the extraction result shown in FIG. 5 . The calculation unit 24 calculates the expected effect of the measure for this combination in accordance with the definition of the above-mentioned calculation method of the expected effect of the measure. The calculation unit 24 uses values ​​stored in the extraction result DB 33 as the median purchase amount and frequency of each of the measure target segment and the measure target segment included in the above-mentioned definition. Furthermore, the calculation unit 24 uses data defined in advance for each combination of classification attributes for the response rate and cost amount included in the above-mentioned definition. Furthermore, the response rate and cost amount may also be defined for each segment combination.

[0049] When implementing a campaign targeting the "morning coffee" segment with the "morning coffee and bread" segment as the campaign target segment, assuming a response rate of 0.5 and a cost of 10,000, the expected campaign effect is calculated as 65,000 yen, as shown below. Expected campaign effect = a x b x c - d = 1000 x (400 - 250) x 0.5 - 10000 = 65000

[0050] 6, the calculation unit 24 associates the calculated expected effect of the measure with a pair of the measure target segment and the measure target segment, and stores the result as support information in the support information DB 34. When the calculation unit 24 has finished calculating the expected effect of the measure for all pairs of the measure target segment and the measure target segment included in the extraction result and storing the results in the support information DB 34, it passes the measure policy information to the output unit 25.

[0051] The output unit 25 acquires support information corresponding to the policy direction information passed from the calculation unit 24 from the support information DB 34, and outputs the acquired support information in a list to the display unit 16. As a result, the support information is displayed in a list on the display unit 16. Fig. 7 shows an example of a list of support information for a combination of the classification attributes "time period" and "multiple purchase pattern."

[0052] The output unit 25 may also be configured to tally up the expected effects of measures for each combination of classification attributes and output a list in which the expected effects of measures for each combination of classification attributes are ranked in descending order. The output unit 25 may also add the content of measures corresponding to the combination of classification attributes to the ranked combination of classification attributes. This case will now be described in detail.

[0053] The output unit 25 retrieves all support information for a certain combination of classification attributes from the support information DB 34, and calculates the maximum value, average value, total value, etc. of the expected effect of the policy included in the retrieved support information as a representative value for the combination of classification attributes. In the case of support information for the combination of classification attributes "time period" and "multiple purchase pattern" shown in Figure 7, if the representative value is taken as the maximum value, the expected effect of the policy for the combination of classification attributes "time period" and "multiple purchase pattern" is 95,000 yen.

[0054] The output unit 25 generates and assigns a policy content for each combination of classification attributes in accordance with a predetermined policy content generation definition. As shown in Fig. 8, the generation of the policy content is defined based on the combination of the policy target type, the policy KPI, and the classification attributes. In the example of Fig. 8, when the policy target type is "purchase," the policy KPI is "purchase amount," and the combination of classification attributes is "time period, multiple purchase pattern," the policy content is "product set sale." Fig. 9 shows an example in which policy content is assigned to each combination of classification attributes and displayed in a list sorted in descending order of expected policy effect.

[0055] The output unit 25 may also output a list in which the policy target segments, policy target segments, and expected policy effects corresponding to the combination of classification attributes selected by the user from the list shown in Fig. 9 are ranked in descending order of expected policy effects, and examples of policy implementations are also provided. This case will now be described in detail.

[0056] The output unit 25 retrieves support information corresponding to a combination of classification attributes specified by the user from the support information DB 34. For example, in FIG. 9 , when the combination of the first-ranked classification attributes "time zone" and "multiple purchase pattern" is specified, the output unit 25 retrieves support information corresponding to this combination from the support information DB 34. In addition, the output unit 25 generates a policy example for each of the retrieved support information in accordance with a predetermined policy example generation definition. As shown in FIG. 10 , the policy example is generated using a template defined based on the combination of the policy target type, policy KPI, and classification attribute, and the support information.

[0057] The output unit 25 generates a campaign example by inserting the classification attribute values ​​included in each of the campaign target segment and campaign target segment included in the support information into a template in accordance with the generation definition. For example, FIG. 10 shows an example of a generation definition of a campaign example under the conditions that the campaign target type is "purchase," the campaign KPI is "purchase amount," and the combination of classification attributes is "time period, concurrent purchase pattern." In this example, the template is defined as follows: "When purchasing <a set of classification attribute values ​​of the campaign target segment>, is sold as a set." The output unit 25 generates a campaign example by inserting data obtained from the support information within the < > of the template. In the example of FIG. 10, the <a set of classification attribute values ​​of the campaign target segment> is "coffee in the morning" (underlined in FIG. 10), and the is "coffee and bread" (wavy line in FIG. 10). In this case, the campaign example generated is "When purchasing coffee in the morning, coffee and bread are sold as a set."

[0058] The output unit 25 outputs a list of the support information shown in Fig. 7, for example, by adding the generated example of the measures to each of the support information, and ranking the measures in order of expected effect. Fig. 11 shows an example of the list display in this case.

[0059] As described above, by outputting support information in various modes, it is possible to provide a display that is easy for the user (planner) to use.

[0060] Furthermore, the output unit 25 may accept an actual campaign effect input by the user after the campaign is implemented and output an evaluation result comparing the expected campaign effect with the actual campaign effect. Specifically, the output unit 25 calculates an evaluation of the expected campaign effect, such as the ratio of the actual campaign effect to the expected campaign effect, or the difference between the expected campaign effect and the actual campaign effect. For example, assume that the actual campaign effect is 57,000 yen when a set of coffee and bread is sold at the time of purchase for the campaign target segment "morning, coffee." In this case, the output unit 25 outputs an evaluation result such as that shown in FIG. 12 using the actual campaign effect of 57,000 yen input by the user. In the example of FIG. 12, the achievement rate of the expected campaign effect is calculated as the evaluation. In this example, the achievement rate is calculated as "actual campaign effect / expected campaign effect," and the achievement rate is displayed as 0.88.

[0061] By reflecting this evaluation result in the expected effect of a policy when a similar policy is formulated next time, it becomes possible to calculate the expected effect of the policy according to the actual policy. Therefore, the output unit 25 may redefine the response rate, for example, by multiplying the response rate included in the definition of the calculation method of the expected effect of the policy calculated by the calculation unit 24 by the achievement rate, which is the evaluation result. For example, if the original response rate is 0.5 and the achievement rate is 0.88, the output unit 25 may redefine the next response rate as 0.44 = 0.5 × 0.88.

[0062] Next, the operation of the policy formulation support device 10 will be described. Fig. 13 is a flowchart showing the flow of policy formulation support processing by the policy formulation support device 10. Fig. 16 is a flowchart showing the flow of evaluation processing by the policy formulation support device 10. The policy formulation support processing and evaluation processing are performed by the CPU 11 reading out a policy formulation support program from the ROM 12 or storage 14, expanding it into the RAM 13, and executing it. Note that the policy formulation support processing and evaluation processing are examples of the policy formulation support method disclosed herein.

[0063] First, the policy formulation support process shown in FIG. 13 will be described.

[0064] In step S10, the CPU 11 functions as the acquisition unit 21 to acquire the policy information including the policy target type and the policy KPI input by the user. Here, it is assumed that the policy target type is "purchase" and the policy KPI is "purchase amount."

[0065] Next, in step S20, the CPU 11, functioning as the acquisition unit 21, acquires store data recorded in units of "purchases" for a predetermined period (for example, one month) in accordance with the acquired policy information from the store data DB 31. Then, the CPU 11, functioning as the acquisition unit 21, passes the acquired policy information and store data in units of "purchases" to the aggregation unit 22.

[0066] Next, in step S30, the CPU 11, as the aggregation unit 22, aggregates the values ​​of the policy KPIs for each segment, which is a combination of the classification attribute values ​​of each combination of two classification attributes included in the store data for each "purchase." Then, the CPU 11, as the aggregation unit 22, stores the aggregation results in the aggregation result DB 32 and transfers the policy direction information to the extraction unit 23.

[0067] Next, in step S40, the CPU 11 executes extraction processing as the extraction unit 23. The extraction processing will now be described in detail with reference to FIG.

[0068] In step S41, the CPU 11, functioning as the extraction unit 23, selects one combination of classification attributes from the combinations of classification attributes linked to the store data in units of "purchase." Next, in step S42, the CPU 11, functioning as the extraction unit 23, acquires the tabulation results for the selected combination of classification attributes from the tabulation result DB 32. Then, the CPU 11, functioning as the extraction unit 23, selects one segment set from the acquired tabulation results.

[0069] Next, in step S43, the CPU 11, as the extraction unit 23, sets, for example, a segment with a low median purchase amount as the policy target segment and a segment with a high median purchase amount as the policy target segment based on the statistical value of the policy KPI "purchase amount" among the selected segment group.

[0070] Next, in step S44, the CPU 11, functioning as the extraction unit 23, determines whether the selected segment pair satisfies the extraction conditions. The extraction conditions include statistical conditions based on the statistical difference in the policy KPIs between the segments, and characteristic conditions related to the similarity between the segments. If the extraction conditions are satisfied, the process proceeds to step S45; if the extraction conditions are not satisfied, the process proceeds to step S46. In step S45, the CPU 11, functioning as the extraction unit 23, extracts pairs of the policy target segment and the policy target segment set in step S43, and stores them in the extraction result DB 33.

[0071] In step S46, the CPU 11, functioning as the extraction unit 23, determines whether all segment sets have been selected from the segments included in the counting result. If there are unselected segment sets, the process returns to step S42, and if all segment sets have been selected, the process proceeds to step S47.

[0072] In step S47, the CPU 11, functioning as the extraction unit 23, determines whether or not all combinations of classification attributes of store data in units of "purchase" have been selected. If there are any unselected combinations, the process returns to step S41, and if all combinations have been selected, the CPU 11, functioning as the extraction unit 23, passes the policy direction information to the calculation unit 24, ends the extraction process, and returns to the policy formulation support process (FIG. 13).

[0073] If a combination of classification attributes is specified in the policy information, the processes of steps S42 to S46 are executed for that combination of classification attributes.

[0074] Next, in step S50, the CPU 11, functioning as the calculation unit 24, acquires extraction results corresponding to the policy direction information from the extraction result DB 33. Then, for each pair of a policy target segment and a policy target segment included in the acquired extraction results, the CPU 11, functioning as the calculation unit 24, calculates an expected policy effect that is expected when a policy targeting the policy KPI for the policy target segment is implemented on the policy target segment. For all pairs of a policy target segment and a policy target segment included in the extraction results, for each combination of classification attributes, the CPU 11, functioning as the calculation unit 24, associates the calculated expected policy effect with the pair of the policy target segment and the policy target segment, and stores the result as support information in the support information DB 34. Then, the CPU 11, functioning as the calculation unit 24, passes the policy direction information to the output unit 25.

[0075] Next, in step S60, the CPU 11 executes output processing as the output unit 25. The output processing will now be described in detail with reference to FIG.

[0076] In step S61, the CPU 11 functions as the output unit 25 to acquire support information corresponding to the policy direction information from the support information DB 34 and display a list of the acquired support information. Next, in step S62, the CPU 11 functions as the output unit 25 to calculate representative values ​​such as the maximum value, average value, and total value of the expected policy effect included in the support information for each combination of classification attributes as the expected policy effect for that combination of classification attributes.

[0077] Next, in step S63, the CPU 11, as the output unit 25, generates policy content for each combination of classification attributes in accordance with a predetermined policy content generation definition. Next, in step S64, the CPU 11, as the output unit 25, ranks the combinations of classification attributes in descending order of the expected policy effect calculated in step S62, and displays a list with the policy content generated in step S63 attached.

[0078] Next, in step S65, the CPU 11, as the output unit 25, acquires support information corresponding to the combination of classification attributes selected by the user from the list displayed in step S64 from the support information DB 34. Next, in step S66, the CPU 11, as the output unit 25, generates a policy example for each support information for each pair of a policy target segment and a policy target segment included in the acquired support information, in accordance with a predetermined policy example generation definition.

[0079] Next, in step S67, the CPU 11, as the output unit 25, ranks the pairs of the policy target segment and the policy target segment in descending order of the expected effect of the policy contained in the support information acquired in step S65, and displays a list of the rankings together with the example policy generated in step S66. Then, the extraction process is terminated, and the process returns to the policy formulation support process (FIG. 13), whereupon the policy formulation support process is terminated.

[0080] Although a list is displayed in each of steps S61, S64, and S67, it is not necessary to display all of them, and at least one of them may be selectively displayed. After the display of step S67, the process may return to step S64 in response to a user instruction, or the ranking of the classification attribute combinations in step S64 may be always displayed, and a classification attribute combination designation may be accepted in step S65. In other words, a classification attribute combination different from the previous one may be accepted.

[0081] Next, the evaluation process shown in FIG. 16 will be described.

[0082] In step S71, the CPU 11, as the output unit 25, accepts the actual policy effect input by the user after the policy is implemented. Next, in step S72, the CPU 11, as the output unit 25, calculates an evaluation such as the ratio of the actual policy effect to the expected policy effect, the difference between the expected policy effect and the actual policy effect, etc. Next, in step S73, the CPU 11, as the output unit 25, displays the evaluation result, and the evaluation process ends.

[0083] As described above, the policy formulation support device according to this embodiment acquires store data, which is data related to purchases at a store, and policy policy information, which includes policy target types and policy KPIs. The policy formulation support device also aggregates the store data for each segment, which is a combination of classification attribute values ​​corresponding to a combination of classification attributes of the store data, in accordance with the policy policy information. The policy formulation support device also extracts, from the aggregated segments, pairs of policy target segments for which policies are to be implemented and policy target segments, which are the targets of the policies. For each pair of policy target segment and policy target segment, the policy formulation support device then calculates the expected policy effect that is expected when a policy targeting the policy KPI for the policy target segment is implemented for the policy target segment, and displays these as support information in association with each other. This makes it possible to support the selection of policy targets with high policy effects according to the direction of the policy.

[0084] The policy development support device according to this embodiment also outputs the policy content and example policy together with the support information. This allows easy selection of the policy content and example policy for a policy targeting "purchasing morning snacks," such as whether to sell snacks in a set with something else, or to reduce the price of snacks.

[0085] Furthermore, after a policy is implemented, the policy formulation support device according to this embodiment evaluates the expected effect of the policy based on the expected effect and the actual effect of the policy, and displays the evaluation result. This allows the evaluation result to be reflected in the calculation of the expected effect of the policy for the next and subsequent similar policies, and as a result, it becomes possible to calculate the expected effect of the policy in accordance with the situation of the store.

[0086] In the above embodiment, a case where a list of support information is displayed on the display unit has been described, but the output form of the support information is not limited to this. For example, the various lists described in the above embodiment may be printed by a printer.

[0087] <Other Application Examples> In the above embodiment, an example was described in which the direction of the sales promotion measures is "to increase purchase sales," but the technology of the present disclosure can also be applied to other directions of sales promotion measures. Here, a case in which the technology of the present disclosure is applied to a case in which the direction of the sales promotion measures is "to increase the number of customers visiting the store" will be described. Note that in this application example, some descriptions of parts that are similar to the above embodiment will be omitted.

[0088] The acquisition unit 21 acquires policy policy information including "customer" as the policy target type and "number of store visits" as the policy KPI. Furthermore, the acquisition unit 21 acquires store data recorded in units of "customer" as the policy target type included in the acquired policy policy information from the store data DB 31. FIG. 17 shows an example of store data recorded in units of "customer." In the example of FIG. 17 , each row (each record) corresponds to customer data for one customer, and each customer data includes information on "customer ID," "gender," "age group," "visit interval," "number of store visits," "number of purchased items," and "purchase amount."

[0089] The "customer ID" is an identifier for customer data. The "time slot" is the time slot during which the customer identified by the customer ID frequently visits the store, and is classified into classes (e.g., four time slot classes: morning, noon, afternoon, and evening) based on the number of visits by time slot during a predetermined period (e.g., one month). The classification method may be any method based on the time of visit. For example, the customer may be classified by the time slot during which the customer visited the store most frequently in a month, or multiple customers may be clustered based on the similarity of the frequency distribution of visit times for each customer, and each class may be classified as a single time slot class. The "visit interval" is classified into classes (e.g., four visit interval classes: once a day, once every three days, once every five days, and once every six days) based on the number of visits per predetermined period (e.g., one month). As with the method for classifying classes of time periods, the method for classifying the number of visits into classes can be any method that classifies based on the visit interval, such as clustering multiple customers based on the similarity of the frequency distribution of visit intervals per customer and classifying each class as a single visit interval class.

[0090] The aggregation unit 22 aggregates store data for each "customer" based on the policy information and stores the aggregation results in the aggregation result DB 32. For example, based on the policy KPI being "number of store visits," the aggregation unit 22 aggregates values ​​related to the "number of store visits" for the same segment for each combination of classification attributes of store data for each "customer." The aggregation unit 22 may calculate, for example, the median number of store visits or the frequency of customers (number of customers) as the aggregated value of the "number of store visits." In addition, the aggregation unit 22 also calculates the median purchase amount in order for the calculation unit 24 to calculate the expected effect of the policy.

[0091] FIG. 18 shows an example of the aggregation result for a combination of the classification attributes "time period" and "visit interval" in store data by "customer." In FIG. 18, for example, the segment "morning, once a day" indicates that the classification attribute value of the classification attribute "time period" is "morning" and the classification attribute value of the classification attribute "visit interval" is "once a day." Furthermore, the example of FIG. 18 indicates that, for example, when store data by "customer" for one month is aggregated for customers in the segment "morning, once a day," the median purchase amount is "3,700 yen" and the customer frequency (total number of customers) is "100 people."

[0092] The extraction unit 23 extracts pairs of a policy target segment and a policy target segment from the segments included in the tabulation results based on the policy policy information, and stores the extraction results in the extraction result DB 33. Referring to FIG. 19 , a process for extracting pairs of a policy target segment and a policy target segment from the tabulation results for combinations of the classification attributes "time period" and "visit interval" will be described. Here, an example of extraction processing will be described under extraction conditions in which the statistical conditions are that there is a significant difference and the effect size is greater than 0.15, and the characteristic conditions are that the classification attribute values ​​for "time period" are the same and the classification attribute value for "visit interval" is shorter for the policy target segment than for the policy target segment (higher visit frequency).

[0093] In this example, the extraction unit 23 first extracts a pair of two segments from the segments included in the counting result shown in Figure 18. For example, if the extraction unit 23 extracts a pair of segments "morning, once a day" and "morning, once every three days," the segment "morning, once every three days" with the lowest median of the measure KPI (here, the number of store visits) in this pair of segments is set as the measure target segment, and the segment "morning, once a day" with the highest median is set as the measure target segment. In this example, the statistical condition is determined to have "no significant difference," and the extraction unit 23 does not extract this pair of segments.

[0094] Furthermore, when the extraction unit 23 extracts a segment pair of "morning, once every 3 days" and "morning, once every 5 days" from the aggregation results, it sets the measure target segment to "morning, once every 5 days" and the measure target segment to "morning, once every 3 days" based on the measure KPI. In this case, there is a significant difference, and the effect size of 0.4 is greater than or equal to the threshold, so the statistical condition is met. Furthermore, the classification attribute value of "time period" is the same, "morning," and the classification attribute value of "visit interval" indicates that the visit interval is shorter (visit frequency is higher) in the measure target segment than in the measure target segment, so the characteristic condition is also met. Therefore, the extraction unit 23 extracts this segment pair and stores it in the extraction result DB 33.

[0095] Furthermore, it is assumed that the extraction unit 23 extracts a segment pair of "morning, once every three days" and "afternoon, once every five days" from the counting result. In this case, the classification attribute values ​​of "time period" are not the same and do not satisfy the characteristic condition, so the extraction unit 23 does not extract this segment pair.

[0096] The calculation unit 24 calculates the expected effect of the measure for each pair of a measure target segment and a measure target segment included in the extraction result by the extraction unit 23. For example, it is assumed that the calculation method of the expected effect of the measure is defined as shown below. Expected effect of the measure = a x b x d - e a: Frequency value of the segment with a low median number of store visits b: Median number of store visits of the segment with a high median number of store visits - Median number of store visits of the segment with a low median number of store visits c: Median purchase amount of the segment with a high median number of store visits - Median purchase amount of the segment with a low median number of store visits d: Response rate to the measure (or similarity) e: Cost of the number of store visits related to the measure (cost amount / c) x b

[0097] The above definition is used to calculate the expected effect of a campaign from the perspective of whether the median number of store visits by customers belonging to the campaign target segment, divided by their frequency (number of customers), will approach the median number of store visits for the campaign target group as a result of the campaign. Assuming that the campaign will not necessarily bring all customers belonging to the campaign target segment closer to the median number of store visits for the campaign target segment, the expected effect of the campaign is calculated by dividing the result by the response rate d to the campaign, which is set to a value of 0 < d < 1, and subtracting the cost of the number of store visits related to the campaign. Note that, because costs are often defined on a monetary basis, the cost e is calculated by converting the monetary cost into the number of store visits.

[0098] Furthermore, the response rate may be a constant based on past implementation of measures, or may be the similarity between the measure target segment and the measure target segment if the similarity between them can be calculated. For example, the similarity may be considered to be the similarity in the distribution of the time periods during which customers visit the store. In this case, the similarity between the segments can be calculated by calculating the centroid vectors of the measure target segment and the measure target segment based on a vector whose elements are the number of customers visiting the store by time period, and then calculating the cosine similarity between the centroid vectors as the similarity between the segments.

[0099] Furthermore, the cosine similarity is not necessarily used as the response rate; any method may be used as long as it allows for setting a response rate according to the similarity between segments. For example, a threshold value may be set as the response rate, such as 1.0 when the cosine similarity is 0.9 or greater, or 0 when the cosine similarity is less than 0.5. Alternatively, the similarity value may be input into a sigmoid function to emphasize the response rate of the segment with greater similarity and, conversely, weaken the response rate of the segment with less similarity. This allows the response rate of the top pair of the most similar segment and the bottom pair to be increased and decreased, respectively, among all pairs of the target segment and the target segment. If the response rate is set to a small value (e.g., 0), the expected effect of the campaign will be small, and therefore support information including these two segments may be excluded from the support information presented to the user.

[0100] In an example using the above calculation definition, the calculation unit 24 first extracts, for example, a pair of "morning, once every five days" as the measure target segment and "morning, once every three days" as the measure target segment from the extraction result shown in FIG. 19 . The calculation unit 24 calculates the expected effect of the measure for this pair in accordance with the definition of the above-mentioned calculation method for the expected effect of the measure. The calculation unit 24 uses values ​​stored in the extraction result DB 33 as the median number of store visits, median purchase amount, and frequency for each of the measure target segment and the measure target segment included in the above definition. Furthermore, the calculation unit 24 uses data that is predefined for each combination of classification attributes and for each segment group for the response rate and cost amount included in the above definition.

[0101] When implementing a campaign with the target segment "morning, once every 5 days" and the target segment "morning, once every 3 days," and assuming a response rate of 0.3 and a cost of 50,000, the expected effect of the campaign is calculated as 85 times, as shown below. Expected effect of campaign = a x b x c - d = 250 x (6 - 4) x 0.3 - (50,000 / (2,800 - 1,250) x (6 - 4)) ≒ 85

[0102] As shown in FIG. 20, the calculation unit 24 associates the calculated expected effect of the measure with a pair of the measure target segment and the measure target segment, and stores the pair in the support information DB 34 as support information.

[0103] The output unit 25 displays a list of the support information. Fig. 21 shows an example of a list of support information for a combination of the classification attributes "time period" and "visit interval."

[0104] Furthermore, the output unit 25 may tally the expected effect of the measures for each combination of classification attributes, rank the expected effect of the measures for each combination of classification attributes in descending order, and output a list with the details of the measures attached to each combination of classification attributes. This case will be specifically described. For example, in the case of support information for the combination of the classification attributes "time period" and "visit interval" shown in FIG. 21, if the representative value is set to the maximum value, the expected effect of the measures for the combination of the classification attributes "time period" and "visit interval" is 600 times.

[0105] The output unit 25 generates and assigns a policy content for each combination of classification attributes in accordance with a predetermined policy content generation definition. As shown in Fig. 22, the generation of the policy content is defined based on the combination of the policy target type, the policy KPI, and the classification attributes. In the example of Fig. 22, when the policy target type is "customer," the policy KPI is "number of store visits," and the combination of classification attributes is "time period, store visit interval," the policy content is "coupon distribution." Fig. 23 shows an example in which policy content is assigned to each combination of classification attributes and displayed in a list in descending order of expected policy effect.

[0106] The output unit 25 may also output a list in which the policy target segments, policy target segments, and expected policy effects corresponding to the combination of classification attributes selected by the user from the list shown in Fig. 23 are ranked in descending order of expected policy effects, and further includes examples of policy implementations. This case will be specifically described.

[0107] For example, when the combination of the first-ranked classification attributes "time period" and "visit interval" in Fig. 23 is specified, the output unit 25 retrieves support information corresponding to this combination from the support information DB 34. Furthermore, the output unit 25 generates a policy example for each piece of retrieved support information in accordance with a predetermined definition for generating a policy example. As shown in Fig. 24, the policy example is generated using a template defined based on the combination of the policy target type, policy KPI, and classification attribute, and the support information.

[0108] The output unit 25 generates a campaign example by inserting the classification attribute values ​​included in each of the campaign target segment and campaign target segment included in the support information into a template in accordance with the generation definition. For example, FIG. 24 shows an example of a generation definition of a campaign example under the conditions that the campaign target type is "customer," the campaign KPI is "number of visits," and the combination of classification attributes is "time period, visit interval." In this example, the template defined is, "To customers of <a set of classification attribute values ​​of the campaign target segment>, distribute coupons after the visit interval reaches ." In the example of FIG. 24, the <a set of classification attribute values ​​of the campaign target segment> is "daytime, once every five days" (underlined in FIG. 24), and the is "once a day" (dashed line in FIG. 24). In this case, the campaign example generated is, "To customers who visit once a day in the daytime, distribute coupons after the visit interval reaches once a day."

[0109] The output unit 25, for example, adds the example of the measures generated as described above to each piece of support information shown in Fig. 21 and outputs a list ranking the measures in descending order of expected effect. Fig. 25 shows an example of the list display in this case.

[0110] Furthermore, the output unit 25 may accept the actual campaign effect input by the user after the campaign is implemented and output an evaluation result comparing the expected campaign effect with the actual campaign effect. For example, suppose a campaign is implemented targeting customers whose visit interval is "once every five days during the day," in which a customer visits the store approximately once a day throughout a month. The campaign distributes coupons, such as discount coupons or product exchange coupons, when the visit interval reaches once a day. Assume that the actual campaign effect for this campaign is a total of 500 visits over the course of a month. In this case, the output unit 25 outputs an evaluation result, such as that shown in FIG. 26 , using the 500 actual campaign effects input by the user. In the example of FIG. 26 , the achievement rate of the expected campaign effect is calculated as the evaluation. In this example, the achievement rate is calculated as "actual campaign effect / expected campaign effect," and the achievement rate is displayed as 0.83.

[0111] In this way, the present application example also provides the same effects as the above embodiment.

[0112] In the above embodiment, the policy formulation support process and the evaluation process, which are executed by the CPU by reading software (programs), may be executed by various processors other than the CPU. Examples of such processors include programmable logic devices (PLDs) (such as field-programmable gate arrays (FPGAs)) whose circuit configuration can be changed after manufacture, and dedicated electrical circuits, such as application-specific integrated circuits (ASICs), which are processors with circuit configurations designed specifically to execute specific processes. Furthermore, the policy formulation support process and the evaluation process may be executed by one of these various processors, or by a combination of two or more processors of the same or different types (e.g., multiple FPGAs, or a combination of a CPU and an FPGA). Furthermore, the hardware structure of these various processors is, more specifically, an electrical circuit that combines circuit elements such as semiconductor devices.

[0113] In the above embodiment, the policy formulation support program is pre-stored (installed) in the storage 14, but the present invention is not limited to this. The program may be provided in a form stored on a non-transitory storage medium such as a CD-ROM (Compact Disk Read Only Memory), a DVD-ROM (Digital Versatile Disk Read Only Memory), or a USB (Universal Serial Bus) memory. The program may also be downloaded from an external device via a network.

[0114] The following additional notes are provided regarding the above-described embodiments.

[0115] (Supplementary clause 1) A policy formulation support device including: an acquisition unit that acquires store data, which is data related to purchases made at a store, and instruction information that indicates the direction of a sales promotion policy, including the type of policy target and indicators to be improved by implementing the policy; an aggregation unit that aggregates the store data for each segment, which is a combination of classification attribute values ​​corresponding to a combination of classification attributes of the store data, in accordance with the instruction information; an extraction unit that extracts from the aggregated segments pairs of policy target segments, which are the targets of the policy implementation, and policy target segments, which are the goals of the policy; a calculation unit that calculates, for each pair, an expected policy effect that is expected when a policy is implemented for the policy target segment; and an output unit that outputs the calculated expected policy effect in association with the pair of the policy target segment and the policy target segment.

[0116] (Appendix 2) The extraction unit is a policy development support device described in appendix 1, which extracts, from the aggregated sets of segments, sets of segments that satisfy extraction conditions including statistical conditions based on the statistical differences in the indicators between the segments and characteristic conditions regarding the similarity of the classification attribute values, as sets of the policy target segment and the policy target segment.

[0117] (Appendix 3) A policy development support device as described in appendix 1 or appendix 2, wherein the extraction unit extracts the segment with the lower index from the set of segments as the policy target segment and the segment with the higher index as the policy target segment.

[0118] (Appendix 4) A policy development support device described in any one of appendixes 1 to 3, wherein the calculation unit calculates the expected effect of the policy as the change in the index of the policy target segment that is expected when a policy is implemented to bring the index of the policy target segment closer to the index of the policy target segment.

[0119] (Appendix 5) A policy formulation support device described in any one of appendixes 1 to 4, wherein the output unit aggregates the expected policy effects for each group of segments for each combination of classification attributes, ranks the aggregated expected policy effects in order of highest to lowest, and displays a list of the combinations of classification attributes.

[0120] (Supplementary Item 6) The policy formulation support device according to Supplementary Item 5, wherein the output unit displays a list of ranked combinations of classification attributes with policy contents corresponding to the combinations of classification attributes.

[0121] (Appendix 7) A policy development support device as described in appendix 5 or appendix 6, in which the output unit displays a list of the policy target segments, policy target segments, and expected policy effects that correspond to a combination of classification attributes specified from the ranked combinations of classification attributes, ranked in order of the highest expected policy effects.

[0122] (Appendix 8) The output unit generates an example of a policy implementation based on the direction of the sales promotion policy and the difference between the classification attribute value of the policy target segment and the classification attribute value of the policy target segment, and displays a list of the example of a policy implementation attached to the policy target segment, the policy target segment, and the expected effect of the policy.

[0123] (Appendix 9) A policy formulation support device described in any one of appendixes 1 to 8, wherein the output unit obtains the actual policy effect after the policy is implemented and displays an evaluation result comparing the expected policy effect with the actual policy effect.

[0124] (Supplementary Item 10) A policy formulation support method executed by a policy formulation support device including an acquisition unit, a counting unit, an extraction unit, a calculation unit, and an output unit, wherein the acquisition unit acquires store data, which is data related to purchases at a store, and instruction information indicating the direction of a sales promotion policy, which includes the type of policy target and an indicator to be improved by implementing the policy; the counting unit aggregates the store data for each segment, which is a combination of classification attribute values ​​corresponding to a combination of classification attributes of the store data, in accordance with the instruction information; the extraction unit extracts from the aggregated segments pairs of policy target segments, which are the targets of the policy implementation, and policy target segments, which are the goals of the policy; the calculation unit calculates, for each pair, an expected policy effect that is expected when a policy is implemented for the policy target segment; and the output unit outputs the calculated expected policy effect in association with the pair of the policy target segment and the policy target segment.

[0125] (Supplementary Item 11) A policy development support program for causing a computer to function as each part of the policy development support device according to any one of Supplementary Items 1 to 9.

[0126] (Supplementary Item 12) A policy formulation support device comprising: a memory; and at least one processor connected to the memory, wherein the processor is configured to: acquire store data, which is data related to purchases at a store, and instruction information indicating the direction of a sales promotion policy, including the type of policy target and indicators to be improved by implementing the policy; aggregate the store data by segment, which is a combination of classification attribute values ​​corresponding to combinations of classification attributes of the store data, in accordance with the instruction information; extract from the aggregated segments pairs of policy target segments, which are the targets of the policy implementation, and policy target segments, which are the goals of the policy; calculate, for each pair, an expected policy effect that is expected when a policy is implemented for the policy target segment; and output the calculated expected policy effect in association with the pair of the policy target segment and the policy target segment.

[0127] (Appendix 13) A non-transitory storage medium storing a program executable by a computer to execute a policy formulation support process, wherein the policy formulation support process: acquires store data, which is data related to purchases at a store, and instruction information indicating the direction of a sales promotion policy, including the type of policy target and indicators to be improved by implementing the policy; aggregates the store data by segment, which is a combination of classification attribute values ​​corresponding to a combination of classification attributes of the store data, in accordance with the instruction information; extracts from the aggregated segments pairs of policy target segments, which are the targets of the policy implementation, and policy target segments, which are the goals of the policy; calculates, for each pair, an expected policy effect that is expected when a policy is implemented for the policy target segment; and outputs the calculated expected policy effect in association with the pair of the policy target segment and the policy target segment.

[0128] REFERENCE SIGNS LIST 10 Policy formulation support device 11 CPU 12 ROM 13 RAM 14 Storage 15 Input unit 16 Display unit 17 Communication I / F 19 Bus 21 Acquisition unit 22 Counting unit 23 Extraction unit 24 Calculation unit 25 Output unit 31 Store data DB 32 Counting result DB 33 Extraction result DB 34 Support information DB

Claims

1. A policy formulation support device comprising: an acquisition unit that acquires store data, which is data related to purchases made at a store, and instruction information that indicates the direction of a sales promotion policy, including the type of policy target and indicators to be improved by implementing the policy; an aggregation unit that aggregates the store data for each segment, which is a combination of classification attribute values ​​corresponding to a combination of classification attributes of the store data, in accordance with the instruction information; an extraction unit that extracts from the aggregated segments pairs of policy target segments for which policies are to be implemented and policy target segments that are the goals of the policies; a calculation unit that calculates, for each pair, an expected policy effect that is expected when a policy is implemented for the policy target segment; and an output unit that outputs the calculated expected policy effect in association with the pair of policy target segment and policy target segment.

2. The policy development support device of claim 1, wherein the extraction unit extracts, from the aggregated sets of segments, sets of segments that satisfy extraction conditions including statistical conditions based on the statistical differences in the indicators between the segments and characteristic conditions regarding the similarity of the classification attribute values, as sets of the policy target segment and the policy target segment.

3. A policy development support device as described in claim 1 or claim 2, wherein the extraction unit extracts the segment with the lower index from the set of segments as the policy target segment, and the segment with the higher index as the policy target segment.

4. A policy development support device as described in claim 1 or claim 2, wherein the calculation unit calculates the expected effect of the policy as the change in the index of the policy target segment that would be expected if a policy were implemented to bring the index of the policy target segment closer to the index of the policy target segment.

5. A policy development support device as described in claim 1 or claim 2, wherein the output unit aggregates the expected policy effects for each group of segments for each combination of classification attributes, ranks the aggregated expected policy effects in order of highest to lowest, and displays a list of the combinations of classification attributes.

6. The policy formulation support device according to claim 5, wherein the output unit displays a list of ranked combinations of classification attributes, with policy contents corresponding to the combinations of classification attributes.

7. A policy formulation support method executed by a policy formulation support device including an acquisition unit, a counting unit, an extraction unit, a calculation unit, and an output unit, wherein the acquisition unit acquires store data, which is data related to purchases at the store, and instruction information indicating the direction of a sales promotion policy, including the type of policy target and an indicator to be improved by implementing the policy; the counting unit aggregates the store data for each segment, which is a combination of classification attribute values ​​corresponding to a combination of classification attributes of the store data, in accordance with the instruction information; the extraction unit extracts from the aggregated segments pairs of policy target segments for which policies are to be implemented and policy target segments that are the goals of the policies; the calculation unit calculates, for each pair, an expected policy effect that is expected when a policy is implemented for the policy target segment; and the output unit outputs the calculated expected policy effect in association with the pair of policy target segment and policy target segment.

8. A policy formulation support program for causing a computer to function as each part of the policy formulation support device according to claim 1 or claim 2.

Images