Sales forecasting device, sales forecasting method, and program

Abstract

We provide a system that accurately forecasts sales for new stores. [Solution] The system includes: a store location and attribute information acquisition unit that acquires the location and attributes of a target store for which sales forecasting is to be performed; a forecast area setting unit that sets a forecast area of ​​a first radius centered on the location of the target store, wherein the forecast area setting unit sets a first radius such that the number of existing stores for which sales data can be acquired within the forecast area satisfies a first condition; a survey area setting unit that sets a survey area of ​​a second radius having the same center, wherein the survey area setting unit sets a second radius by adding the radius of the core trading area of ​​stores of the same type as a boundary condition to the first radius of the forecast area; a customer traffic forecasting unit that forecasts the number of customers attracted to the target store and existing stores within the survey area; a regression model fitting unit that fits a regression model between the number of customers attracted to existing stores for which sales data can be acquired within the forecast area and sales; and a forecast result output unit that inputs the number of customers attracted to the target store into the fitted regression model to output the sales forecast result for the target store and outputs a customer traffic distribution map and a population pyramid of the customer traffic for the target store.

Description

[Technical Field]

[0001] This invention relates to a sales forecasting device, a sales forecasting method, and a program. [Background technology]

[0002] A system is known that uses GIS (Geographic Information System) to predict sales for new stores that are scheduled to open, based on information about existing stores in the area.

[0003] For example, Patent Document 1 describes setting a trade area on map data, determining the parameters of a probability model for sales forecasting based on statistical information about consumers and data such as the sales area and sales performance of stores within the trade area, and then forecasting sales for a new store. [Prior art documents] [Patent Documents]

[0004] [Patent Document 1] Japanese Patent Publication No. 2004-185539 [Overview of the project] [Problems that the invention aims to solve]

[0005] However, in creating sales forecasting models, it was not always possible to obtain sufficient data such as the sales area and sales performance of existing stores, which sometimes prevented the model from improving its forecasting accuracy.

[0006] Therefore, the present invention aims to provide a system that can accurately forecast sales for new stores. [Means for solving the problem]

[0007] A sales forecasting device according to one aspect of the present invention comprises: a store location and attribute information acquisition unit that acquires the location and attribute information of a target store for which sales forecasting is to be performed; a forecasting area setting unit that sets a forecasting area of ​​a first radius centered on the location of the target store, wherein the forecasting area setting unit sets the first radius such that the number of existing stores within the forecasting area from which sales data can be acquired satisfies a first condition; a survey area setting unit that sets a second radius centered on the location of the target store, obtained by adding a radius as a boundary condition to the first radius; a customer attraction population forecasting unit that forecasts the number of customers attracted to the target store and existing stores within the survey area; a regression model fitting unit that fits a regression model to forecast sales, with the number of customers attracted predicted for existing stores within the forecasting area from which sales data can be acquired as the explanatory variable and sales as the dependent variable; and a forecast result output unit that outputs a sales forecast result calculated by inputting the number of customers attracted to the target store into the regression model, and a map and graph.

[0008] A sales forecasting method according to one aspect of the present invention comprises: a store location and attribute information acquisition step in which a computer acquires the location and attribute information of a target store for which sales forecasting is to be performed; a forecast area setting step in which the computer sets a forecast area of ​​a first radius centered on the location of the target store, wherein the first radius is set such that the number of existing stores within the forecast area from which sales data can be obtained satisfies a first condition; a survey area setting step in which the computer sets a second radius centered on the location of the target store, by adding a radius as a boundary condition to the first radius; a customer attraction population forecasting step in which the computer forecasts the number of customers for the target store and existing stores within the survey area; a regression model fitting step in which the computer fits a regression model to predict sales, with the number of customers predicted for existing stores within the forecast area from which sales data can be obtained as the explanatory variable and sales as the dependent variable; and a forecast result output step in which the computer outputs a sales forecast result calculated by inputting the number of customers for the target store into the regression model, and a map and graph.

[0009] A program according to one aspect of the present invention causes a computer to function as a store location and attribute information acquisition unit that acquires the location and attribute information of a target store for which sales forecasting is to be performed; a forecast area setting unit that sets a forecast area of ​​a first radius centered on the location of the target store, wherein the forecast area setting unit sets the first radius such that the number of existing stores within the forecast area from which sales data can be acquired satisfies a first condition; a survey area setting unit that sets a second radius centered on the location of the target store by adding a radius as a boundary condition to the first radius; a customer attraction prediction unit that predicts the number of customers attracted to the target store and existing stores within the survey area; a regression model fitting unit that fits a regression model to predict sales, with the predicted number of customers attracted to existing stores within the forecast area from which sales data can be acquired as the explanatory variable and sales as the dependent variable; and a forecast result output unit that outputs sales forecast results calculated by inputting the number of customers attracted to the target store into the regression model, as well as maps and graphs. [Effects of the Invention]

[0010] According to the present invention, it is possible to provide a system that can predict the sales of new stores with high accuracy. [Brief explanation of the drawing]

[0011] [Figure 1] This figure shows the configuration of the sales forecasting system 1 according to this embodiment. [Figure 2] A block diagram showing the configuration of the sales forecasting device 10 according to this embodiment. [Figure 3] A block diagram showing the configuration of the user terminal 20 according to this embodiment. [Figure 4] This block diagram shows a functional module of a program executed by the processor 11 of the sales forecasting device 10 according to this embodiment. [Figure 5] A flowchart illustrating the sales forecasting method using the sales forecasting system 1 according to this embodiment. [Figure 6] A diagram illustrating the concepts of prediction area and survey area in this embodiment. [Figure 7]A diagram for explaining a method of predicting the number of customers attracted using a spatial interaction model by the sales prediction system 1 according to this embodiment. [Figure 8] A diagram exemplifying a method of acquiring the sales floor area by the sales prediction system 1 according to this embodiment. [Figure 9] A diagram showing in a three-dimensional graph the relationship between the combination of coefficients α and β of the spatial interaction model at the prefecture level, etc. in this embodiment and the coefficient of determination between the number of customers attracted and sales. [Figure 10] A diagram showing in a graph the decreasing rate (shopping probability) of the number of customers attracted according to the straight-line distance from the store for each coefficient β in this embodiment. [Figure 11] A diagram showing an example display of a new store, a predicted area and an investigation area, and an existing store for which sales data can be acquired, by the sales prediction system 1 according to this embodiment. [Figure 12] A diagram showing an example display of the fitting result of the regression model by the sales prediction system 1 according to this embodiment. [Figure 13] A diagram showing an example display of the number of customers attracted and the sales prediction result of a new store by the sales prediction system 1 according to this embodiment. [Figure 14] A diagram showing an example display of the customer attraction population distribution map of a new store by the sales prediction system 1 according to this embodiment. [Figure 15] A diagram showing an example display of the population pyramid of the customers attracted by a new store and the circular graph of the scale composition of the customer households by the sales prediction system 1 according to this embodiment.

Embodiments for Carrying Out the Invention

[0012] Next, embodiments for carrying out the present invention will be described in detail with reference to the drawings. FIG. 1 is a diagram exemplifying the configuration of a sales prediction system 1 including a sales prediction apparatus 10 according to an embodiment of the present invention. As shown in FIG. 1, the sales prediction system 1 includes a sales prediction apparatus 10 and a user terminal 20. The sales prediction apparatus 10 and the user terminal 20 are connected via a communication network N such as the Internet.

[0013] The sales forecasting device 10 may be a general-purpose computer, and may consist of a single computer or multiple computers distributed over a communication network N. Furthermore, the sales forecasting device 10 may be built on the cloud. In addition, the functions of the sales forecasting device 10 in this embodiment may be implemented on the user terminal 20.

[0014] Figure 2 is a block diagram showing the configuration of the sales forecasting device 10. As shown in Figure 2, the sales forecasting device 10 includes a processor 11, main memory 12, input / output interface 13, communication interface 14, and storage device 15. The storage device 15 is a computer-readable recording medium such as semiconductor memory (e.g., volatile memory or non-volatile memory) or disk media (e.g., magnetic recording medium or magneto-optical recording medium). The storage device 15 stores programs to be executed by the processor 11, as well as various data. The programs are read from the storage device 15 into the main memory 12, interpreted and executed by the processor 11, thereby executing various functions.

[0015] The user terminal 20 is a terminal used by the user to utilize the sales forecasting system 1. The user terminal 20 can be any terminal device capable of exchanging data with the sales forecasting device 10 via the communication network N, such as a personal computer (PC), notebook PC, tablet terminal, or smartphone. Figure 3 is a block diagram showing the configuration of the user terminal 20. As shown in Figure 3, the user terminal 20 includes a processor 21, input devices 22 such as a keyboard, mouse, various operation buttons, or touch panel, a display device 23 such as an LCD display, a communication interface 24 for connecting to the communication network N, and a storage device 25 such as a disk drive or semiconductor memory (ROM, RAM, etc.). The storage device 25 may store various programs executed by the processor 21 and various data. The user terminal 20 may have a dedicated application installed for using the sales forecasting system 1.

[0016] Figure 4 is a block diagram showing the functional modules of the program executed by the processor 11 of the sales forecasting device 10. As shown in Figure 4, the functional modules executed by the processor 11 of the sales forecasting device 10 include a store location / attribute information acquisition unit 101, a forecast area setting unit 102, a survey area setting unit 103, a customer attraction population forecast unit 104, a regression model fitting unit 105, and a forecast result output unit 106.

[0017] (Types of businesses, industries, and characteristics of stores targeted by the sales forecasting system) In this embodiment, the store type is a physical retail store. The target store for which sales are to be predicted using the sales forecasting system 1 of this embodiment is a newly opened store, and is therefore called a "new store" S. In contrast, stores that exist before the forecast are called "existing stores". As an example of the type of new store, a food supermarket (hereinafter referred to as a food supermarket) will be used to explain the sales forecasting method. The reasons for choosing a food supermarket are (1) it is an industry that handles many products that are purchased frequently, have a large volume, and are heavy, and there is a strong tendency for purchases to be made locally, so the demand source for such stores is mainly the permanent resident population, and (2) the store area stipulated by the Large-Scale Retail Store Location Law is 1000m². 2 This is due to the large number of stores mentioned above, the strong appeal of the stores, and the fact that data on sales and area of ​​existing stores is often published in data books, etc., making it relatively easy to obtain data on existing stores. These characteristics of food supermarkets make them an industry in which highly accurate sales forecasting is possible (References 1, 2). (Reference 1) Birkin, M., Boden, P., Williams, J., “Spatial decision support systems for petrol forecourts”, “Planning Support Systems in Practice” (Edited by Geertman, S. and Stillwell, J.), Springer, 2003 (Reference 2) Hiroyuki Kosaka, "Construction of an Effective Store Network Using GIS," "Annual Research Report of the Institute of Information Science, College of Humanities and Sciences, Nihon University," No. 3, 2004, pp. 3-33.

[0018] Next, the sales forecasting method using the sales forecasting system 1 will be explained using the flowchart in Figure 5. First, information on a new supermarket store S is entered via the user terminal 20 (step ST1). In this embodiment, the user (supermarket company, etc.) enters the location information (latitude and longitude) and attribute information (sales area) of the new store. Attribute information such as the number of parking spaces and whether the store is located in a shopping center may also be entered, but for the sake of simplicity, only the sales area, which has the greatest explanatory power, will be considered. The entered information is stored in the store location / attribute information acquisition unit 101.

[0019] The prediction area setting unit 102 reads the location information of the new store S from the store location / attribute information acquisition unit 101 and sets a "prediction area" with the location of the new store S as the center P of the circle (step ST2). The prediction area may be set as an area enclosed by a circle with radius R1 (first radius) with the location of the new store S as the center P of the circle. Radius R1 is set so that the number of existing stores (stores of the company and other companies in the same industry) that exist within the prediction area and for which sales data (e.g., annual sales) can be obtained is K or more (first condition). K is the number of data points required to fit the regression model, and in this embodiment, it is set to 15, for example. Specifically, first, R1 is set to 5km as an initial value. If there are 14 or fewer existing stores for which sales data can be obtained within the circle with a radius of 5km, the prediction area setting unit 102 expands R1 by 1km increments, for example, until the number of existing stores reaches 15 or more. Ultimately, R1 may be determined once the number of existing stores for which sales data can be obtained reaches 15 or more. However, even if R1 reaches 16km, if the first condition is not met, the sales forecasting system 1 will determine that it is "unpredictable" because there is not enough sales data for existing stores.

[0020] In the survey area setting unit 103, a "survey area" is set centered on the new store S (step ST3). The survey area is set as a "boundary condition" to improve the accuracy of predicting the number of customers attracted to existing stores within the prediction area. The survey area may be set as an area enclosed by a circle with radius R2 (second radius) with the location of the new store S as the center P of the circle.

[0021] In this embodiment, it is assumed that food supermarkets, which are the target of sales forecasting, form a "core trade area" in an area enclosed by a circle with a radius R3 (third radius) centered on the location of each store, regardless of the store's attributes, whether it is a new store or an existing store. For example, let's assume a core trade area of ​​5 km.

[0022] The radius R2 of the survey area is calculated from the following formula (1). Radius R2 of the survey area = Radius R1 of the forecast area + Radius R3 of the core market area (1)

[0023] Figure 6 shows an example where the radius of the prediction area is 5 km and the radius of the survey area is 10 km. For existing store A located within the prediction area, a core trade area with a radius of 5 km is formed centered on store A, and since this trade area is contained within the survey area, the total number of customers attracted to the core trade area can be predicted. In contrast, for existing store B located within the survey area, the core trade area with a radius of 5 km extends beyond the survey area, so only the number of customers attracted to a partial core trade area (within the survey area) can be predicted. However, existing stores located within the survey area have an impact as competing stores on existing stores and new stores located within the prediction area, so they must be considered as boundary conditions. For this reason, existing stores located within the prediction area are used for sales forecasting, but existing stores located within the survey area are not used for sales forecasting and are only considered in terms of the competitive impact on existing stores and new stores located within the prediction area as "boundary conditions".

[0024] (Prediction of the number of visitors using a spatial interaction model) Next, the customer attraction population prediction unit 104 predicts the number of customers for all stores (new stores and existing stores) located within the designated survey area (step ST4). The customer attraction population prediction unit 104 consists of a spatial interaction model unit 104a, a distance measurement unit 104b, and a coefficient calibration unit 104c.

[0025] Figure 7 illustrates the method for predicting the number of customers using the spatial interaction model unit 104a of this embodiment (Reference 3). In Figure 7, i represents an individual mesh, and j represents each store (supermarket). Pi is the total population of mesh i (and the number of customers by sex and age group, and the number of households by size). Wj represents the sales area of ​​store j. dij represents the straight-line distance from mesh i to store j, and Sij shows the number of customers attracted from mesh i to store j (and the number of customers attracted by sex and age group, and the number of households attracted by size). In this case, the number of customers attracted Sij can be expressed by the following equation (2). (Reference 3) Hiroyuki Kosaka, "Geobusiness: Location evaluation and customer traffic prediction for retail stores using GIS," Kokon Shoin, 2014, pp. 95-111.

[0026]

number

[0027] In equation (2), α is the attractive force coefficient of the store, and β is the resistance coefficient with respect to the distance from the mesh to the store. Ai is the equilibrium factor of mesh i, and can be expressed by the following equation (3).

[0028]

number

[0029] In equation (2), if multiple variables are included as variables of the store's attractiveness, such as the number of parking spaces C in the parking lot in addition to the sales area W, the group of variables related to the attractiveness can be expressed by the following equation (4). ΠX i ai =W α ×Cγ ×··· (4) However, X on the left side i The variables i and a represent the store's attractiveness. i θ represents the coefficient of the power of variable i. The right-hand side shows a specific example, where α is the attractive force coefficient for the store's sales area W, and γ is the attractive force coefficient for the number of parking spaces C. In this way, to incorporate multiple variables of the store's attractive force, it can be expressed by multiplying the attractive force variables X, which have power coefficients, as a product (total product). Equation (2) is a simple spatial interaction model used for explanation, and in order to improve the accuracy of predicting the number of customers, the scalability of such a model can be used for the store's attractive force.

[0030] In equation (2), a negative exponential function is used as the distance diminishing function related to the distance from the mesh to the store. Regarding the distance diminishing function in spatial interaction models, other functions such as power functions and step functions may also be used, demonstrating the model's extensibility in this respect as well.

[0031] In this embodiment, the sales forecasting system 1 is provided for areas of a certain size, such as prefectures. Therefore, the data required to predict the number of customers using a spatial interaction model is (1) population data for each mesh, (2) attribute data for each store, and (3) distance from each mesh to each store, which are collected at the prefecture level or similar. The first two data sets must be accompanied by location information (latitude and longitude). It is also necessary to identify the two coefficients α and β in equation (2).

[0032] The mesh population data represents a variable for measuring the demand for food supermarkets. It utilizes the total population, population by sex and age group, and number of households by size from the 1 / 8 or 1 / 4 regional mesh data with high spatial resolution from the census (the explanation below will use the 1 / 8 regional mesh, which has the highest spatial resolution). The 1 / 8 regional mesh population data allows us to obtain population data within a range of approximately 125m in both the latitude and longitude directions. Utilizing population data with high spatial resolution improves the accuracy of predicting the number of customers attracted to each store, which in turn leads to improved accuracy in sales forecasting for new stores.

[0033] Store attribute data includes data such as sales and sales area. Sales area data indicates a variable that measures the attractiveness of a food supermarket. Store area (total building floor area) can be used instead of sales area, but sales area is preferable because it has a more direct impact on sales. Means of obtaining sales and sales area data for food supermarkets include "company data" from users (food supermarket companies, etc.), commercially available data (e.g., "Food Supermarket Yearbook" (Ryutsu Kikaku Co., Ltd.)), data on the internet (e.g., Store Japan Co., Ltd.'s website), and "competitor data" such as data disclosed by food supermarket companies. In addition, sales area can also be obtained from the registered store area in the list of large retail stores (e.g., publications by local governments or "Store Opening Plan Information: Nationwide Edition" (Sankei-I Co., Ltd.)).

[0034] For stores where sales area data cannot be obtained using the method described above, as shown in Figure 8, aerial photographs of each store may be displayed on a GIS (Geographic Information System) to create building polygons G, and the area of ​​the building polygons G may be obtained as the store area. Furthermore, the sales area may be estimated by multiplying the obtained store area by the sales area ratio (for example, 0.6) or by considering the number of floors in the building. The sales area ratio may be the average of the sales area to store area ratios in food supermarkets where the sales area and store area are known. Obtaining sales area data for all existing food supermarkets at the prefectural level, etc., will ultimately lead to an improvement in the accuracy of sales forecasts for new stores.

[0035] To add location information to mesh population data, the regional mesh of the census is displayed on a GIS and the latitude and longitude of the mesh center point are calculated. To add location information to store sales and sales area data, the latitude and longitude of the stores are obtained by using an address matching service. By adding location information to two attribute data points—population, which indicates the demand for food supermarkets, and sales area, which represents the attractiveness of food supermarkets—and creating geospatial data where attributes and locations are paired, it becomes possible to map the competitive environment of food supermarkets, allowing for predictions of the number of people attracted to the store in a format closer to the real world, and as a result, the accuracy of sales forecasts for new stores is greatly improved.

[0036] The distance measurement unit 104b measures the distance from the mesh to the store. In this embodiment, the distance is measured as the "straight-line distance" from the mesh to the store, based on the location information (latitude and longitude) of the mesh and the store, using the great circle distance method (Reference 4). It is also possible to incorporate the extensibility of measuring the "road distance" from the mesh to the store using Dijkstra's algorithm (Reference 5) with the road network of OpenStreetMap. (Reference 4) Miura, Hidetoshi, "Three Distance Calculation Methods Using Latitude and Longitude," Operations Research, 2015, Vol. 60, No. 12, pp. 701-705. (Reference 5) Miyazaki, Shuichi, "Introduction to Graph Theory: Fundamentals and Algorithms," Morikita Publishing, 2015.

[0037] The coefficient calibration unit 104c calibrates and identifies the coefficients α and β in equation (2). Specifically, for example, the attractive force coefficient α is changed in increments of 0.2 between 0.4 and 2.0, and the distance resistance coefficient β is changed in increments of 0.01 between 0.01 and 0.08, performing 9 × 8 = 72 calibrations. For food supermarkets where annual sales data is available at the prefectural level, the unit identifies the combination of two coefficients that maximizes the coefficient of determination between the number of customers and annual sales calculated using equation (2).

[0038] Figure 9 is a three-dimensional graph showing the relationship between combinations of coefficients α and β and the coefficient of determination R². In the example in Figure 9, the coefficient of determination reaches its highest value of 0.3966 when α = 0.8 and β = 0.02.

[0039] Figure 10 is a graph showing the relationship between the straight-line distance from the supermarket (x) and the rate of decrease in the number of customers (shopping probability) (y) as the distance resistance coefficient changes. For example, in a certain prefecture, when β=0.02, y=e -0.02x Therefore, the probability of making a purchase decreases to about half, or 0.55, when you move 3km away from a supermarket.

[0040] In this embodiment, the sales forecasting system 1 predicts the number of customers attracted to stores (existing stores and new stores) within a survey area of ​​radius R2, using a spatial interaction model with coefficients α and β that are specified at the prefecture level or similar.

[0041] (Sales forecasting using regression models) Next, the regression model fitting unit 105 fits a regression model to predict sales (step ST5). In the regression model, a simple regression model is fitted to existing stores in the prediction area for which sales data (annual sales) can be obtained, with the number of customers predicted by the customer attraction population prediction unit 104 as the explanatory variable and annual sales as the dependent variable, and the regression coefficients (intercept and slope) are identified. Note that in equation (2), instead of the total population, a multiple regression model may be fitted with the number of customers by sex and age group and the number of customers by size predicted using the number of customers by sex and age group and the number of customers by size as explanatory variables, and annual sales as the dependent variable.

[0042] The prediction result output unit 106 predicts the sales of a new supermarket store S using a regression model with the regression coefficients identified in step ST5 (step ST6). The prediction result output unit 106 consists of a prediction output unit 106a and a map / graph output unit 106b.

[0043] The prediction output unit 106a predicts annual sales by inputting the number of customers attracted to the new supermarket store S into a regression model that has the regression coefficients identified in step ST5. It interval estimates the annual sales of the new supermarket store S with a 95% confidence interval and outputs the average and interval of the annual sales forecast. The prediction output unit 106a also outputs an image of a plot of the number of customers attracted and the actual annual sales values ​​of existing stores in the prediction area that were used to fit the regression model, as well as the regression line.

[0044] The map / graph output unit 106b outputs a map showing new supermarket stores as icons, predicted and surveyed areas with radii R1 and R2 centered on the new stores as circles, existing stores within the surveyed areas as circles of varying sizes according to their sales area, and existing stores within the predicted areas for which sales data can be obtained as circles of varying colors and sizes according to their sales area. It also outputs images of a population distribution map of target stores within the surveyed areas, and pie charts showing the population pyramid of those stores and the size composition of the households that visit them.

[0045] In this embodiment, when location data (latitude and longitude) and attribute data (sales area) of a new supermarket store S are input, a first prediction area with a radius of 5 km centered on the new store S and a second survey area with a radius of 10 km, which is created by adding a 5 km radius core trade area as a boundary condition, are first set. If there are a certain number (15 or more) of existing stores for which sales data can be obtained within this prediction area, the prediction area will have a radius of 5 km. If there are not a certain number or more, the radius is gradually increased by 1 km increments from 6 km to 15 km, and the radius containing the certain number of stores is set as the prediction area. Using equation (2) of a spatial interaction model with two coefficients specified by the prefecture where the new store S is located, the number of customers attracted to the new store and existing stores within the survey area with a radius obtained by adding the radius of the prediction area and the radius of the core trade area is predicted. Next, for existing stores for which sales data can be obtained within the prediction area, a regression model is fitted between the number of customers attracted and annual sales, and the regression coefficients are identified. Sales forecasting for the new store is performed by inputting the number of customers attracted to the new store into this regression model.

[0046] (Examples) An example of sales forecasting using the sales forecasting system 1 according to this embodiment is shown below. First, the user inputs location information (latitude and longitude) and attribute information (sales area) of a new supermarket store from the user terminal 20. The sales forecasting device 10 then creates a map of the area around the new store S, as illustrated in Figure 11, and displays it on the display device 23 or outputs it as an image to the storage device 25. As shown in Figure 11, a forecasting area R1 with a radius of 12 km and a survey area R2 with a radius of 17 km are set with the new store S, indicated by an icon, as the center. The forecasting area R1 is set to include a predetermined number (15 or more in this case) of existing stores from which sales data can be obtained. In the example in Figure 11, there are 16 existing stores (gray circles) from which sales data can be obtained within the forecasting area R1, satisfying the sample size requirement necessary for fitting the regression model. In the example in Figure 11, the size of the circles representing existing stores reflects the size of the sales area.

[0047] The sales forecasting device 10 predicts the number of customers attracted to existing stores within a set survey area with a radius of 17 km. For existing stores within a forecast area with a radius of 12 km from which sales data can be obtained, it fits a regression model using the predicted number of customers and the annual sales of each store. Figure 12 shows the regression line and 95% confidence interval for 16 existing stores within the forecast area from which sales data can be obtained, plotting the predicted number of customers on the horizontal axis (X) and the actual annual sales on the vertical axis (Y). This is displayed on the display device 23 or output as an image to the storage device 25. Equation (5) shows the regression equation representing the regression line in Figure 12. X is the number of customers, and Y is the annual sales (million yen). The coefficient of determination of equation (5) is 0.646, which is a high value. Y = 22.972 + 0.2059X (5)

[0048] The sales forecasting device 10 displays the predicted number of customers and annual sales for the new store S on the display device 23 or outputs the result as an image to the storage device 25, as shown in Figure 13. The predicted number of customers is 7,753, and the predicted annual sales by the regression model are 1,619.24 million yen. Furthermore, the interval estimation with a 95% confidence level shows that the upper limit of annual sales is 1,742.09 million yen and the lower limit is 1,496.38 million yen. In addition, the radius of the forecast area and the survey area, and the number of existing stores for which sales data can be obtained within the forecast area are also displayed.

[0049] As shown in Figure 14, the sales forecasting device 10 displays a population distribution map of the new store S in the survey area on the display device 23, or outputs it as an image to the storage device 25. The sales forecasting device 10 can display not only the number of customers but also their distribution.

[0050] As shown in Figure 15, the sales forecasting device 10 uses the number of customers by gender and age group and the number of customers by size to display a population pyramid of the customer base and a pie chart of the size composition of the customers for the new store S on the display device 23, or output it as an image to the storage device 25. Furthermore, by fitting a multiple regression model with the number of customers by gender and age group or the number of customers by size as the explanatory variable and annual sales as the dependent variable, it leads to a more accurate sales forecast that takes into account gender, age group, and household size.

[0051] (Display of Possible Store-Opening Areas) As another embodiment of the present invention, it may be configured to display possible store-opening areas. For example, when there is a store-opening plan for a new store with a sales floor area of 2000 m 2 , sales forecasts are executed for all 1 / 8 area meshes covering the prefecture. Further, on the map, each mesh may be colored according to the predicted sales. This provides information useful for a company or the like making a store-opening plan for a new store to strategically determine candidate locations.

[0052] (Sales Forecast with Reliability and High Accuracy) In the sales forecasting system according to this embodiment, as shown in Table 1, two items, namely sales and the number of customers attracted, are predicted, and attempts have been made to improve the prediction accuracy from various aspects. First, in the sales forecast, an evidence-based approach is adopted in which a store development policy is formulated based on objective factual evidence (evidence). The factual evidence in the sales forecast is the sales performance values of existing stores, and the concept of a "prediction area" is devised as the area for collecting such data.

Table 1

[0053] In collecting the sales data of existing stores, a dynamic collection method is adopted that determines the collection range by judging the situation, rather than a static collection method that determines the collection range in advance. In the present invention, "spatially variable data collection" is devised in which the radius of the prediction area is expanded until the number of data required for fitting the regression model (the first condition) is satisfied, and the sales performance value data is collected.

[0054] By fitting a regression model between the sales data of existing stores (stores of the company itself and other companies in the same industry) within the prediction area collected based on "spatially variable data collection" and the number of customers attracted to the corresponding existing stores predicted by the spatial interaction model, interval estimation with reliability for the sales prediction result of the new store is made possible.

[0055] In contrast, in predicting the number of visitors, various efforts were made to achieve high accuracy in the spatial interaction model, focusing on (1) data, (2) coefficients, and (3) boundary conditions. By using high-resolution regional mesh population data as demand data for the spatial interaction model, high-resolution visitor population predictions were achieved.

[0056] As supply data for the spatial interaction model, we attempted to obtain information on store sales area from our own sources, as well as from literature and online sources. When this was not possible, we created building polygons from aerial photographs of each store using a geographic information system and calculated the sales area. Obtaining sales area data for all stores within the survey area enabled a complete reproduction of competitive relationships and resulted in highly accurate predictions of customer traffic.

[0057] Furthermore, regarding the coefficients of the spatial interaction model, the coefficients were calibrated and identified so that the coefficient of determination between the number of visitors and sales was maximized at the prefectural level, enabling sales forecasts that reflect regional characteristics and leading to highly accurate predictions of the number of visitors.

[0058] Furthermore, by setting a "survey area" that is wider than the prediction area as a boundary condition, and considering the influence of existing stores located within the survey area, we were able to predict the number of customers attracted by existing stores within the prediction area, thereby improving the accuracy of the prediction.

[0059] While this embodiment of the sales forecasting system uses the forecasting of a new supermarket as an example, the type of store is not limited to supermarkets and can be applied to drugstores and other types of stores. In the case of supermarkets, there is an advantage in that sales data for existing stores is relatively easy to obtain, as store sales data is often publicly available in data books, etc.

[0060] The embodiments described above are provided to facilitate understanding of the present invention and are not intended to limit its interpretation. The flowcharts and specific examples of elements provided in the embodiments described are not limited to those exemplified and can be modified as appropriate. Furthermore, it is possible to partially substitute or combine the configurations shown in different embodiments. [Explanation of symbols]

[0061] 1...Sales forecasting system, 10...Sales forecasting device, 11...Processor, 12...Main memory, 13...Input / output interface, 14...Communication interface, 15...Storage device, 20...User terminal, 21...Processor, 22...Input device, 23...Display device, 24...Communication interface, 25...Storage device, 101...Store location / attribute information acquisition unit, 102...Forecasting area setting unit, 103...Survey area setting unit, 104...Customer population forecasting unit, 105...Regression model fitting unit, 106...Forecasting result output unit

Claims

1. A store location and attribute information acquisition unit acquires the location and attribute information of target stores (new stores) for which sales forecasts are to be made, A prediction area setting unit that sets a prediction area with a first radius centered on the location of the target store, the prediction area setting unit that sets the first radius such that the number of existing stores in the same industry as the target store and within the prediction area from which sales data can be obtained satisfies a first condition, A survey area setting unit for setting a second radius survey area centered on the location of the target store, wherein the second radius is set by adding a third radius of the core trading area of ​​stores of the same type as the target store as a boundary condition to the first radius of the prediction area, For the target stores and existing stores within the survey area, the customer attraction forecasting department predicts the number of customers, The system includes a regression model fitting unit that, for existing stores within the aforementioned forecast area where sales data can be obtained, fits a regression model to predict sales, with the predicted number of customers as the explanatory variable and sales as the dependent variable. The aforementioned customer population forecasting unit is, A sales forecasting device that predicts the number of customers attracted to a target store using a spatial interaction model, utilizing population data within a predetermined range as the demand for the target store and existing stores within the survey area.

2. The aforementioned customer population forecasting unit is, The sales forecasting device according to claim 1, which uses sales area data as the supply volume of existing stores within the survey area and predicts the number of customers attracted by the spatial interaction model.

3. The aforementioned customer population forecasting unit is, The sales forecasting device according to claim 2, wherein, if data on the sales floor area of ​​a store cannot be obtained, it displays an aerial photograph of the store on a geographic information system to create a building polygon, measures the store area from the area of ​​the building polygon, and estimates the sales floor area by multiplying the store area by a certain sales floor ratio and / or taking into account the number of floors in the building.

4. The aforementioned customer population forecasting unit is, The sales forecasting device according to claim 1, wherein the spatial interaction model has two coefficients: an attractive force coefficient related to the supply capacity corresponding to the sales area of ​​the store, and a distance resistance coefficient related to the probability of purchase corresponding to the distance from the regional mesh to the store.

5. The sales forecasting device according to claim 4, wherein the two coefficients are calibrated on a prefecture-by-prefecture basis to determine the coefficient of determination between the predicted number of customers and sales.

6. The aforementioned customer population forecasting unit is, The sales forecasting device according to claim 1, which predicts the number of customers attracted to the target store and the existing store within the survey area using a spatial interaction model having the specified coefficients.

7. The sales forecasting device according to claim 1, further comprising: inputting the number of customers of the target store predicted by the customer acquisition population prediction unit into a fitted regression model, outputting a sales forecast result for the target store, and outputting a map showing the target store, its forecast area, the survey area, and the distribution of existing stores of the same type within the survey area; a distribution map of the customer acquisition population of the target store in the survey area; and pie charts showing the population pyramid of the customer acquisition population and the size composition of customer households.

8. The computer acquires location and attribute information of target stores for which sales forecasting is performed, and A prediction area setting step in which a computer sets a prediction area with a first radius centered on the location of the target store, wherein the prediction area setting step sets the first radius such that the number of existing stores within the prediction area from which sales data can be obtained satisfies a first condition, A survey area setting step in which a computer sets a survey area with a second radius centered on the location of the target store, wherein the second radius is set by adding a third radius of the core trading area of ​​stores of the same type as the target store as a boundary condition to the first radius of the predicted area, A computer performs a customer attraction prediction process to predict the number of customers attracted to the target store and existing stores within the survey area, The computer performs a regression model fitting step in which it fits a regression model to predict sales for existing stores within the predicted area for which sales data can be obtained, using the predicted number of customers as the explanatory variable and sales as the dependent variable. The system includes a prediction result output step in which a computer inputs the number of customers attracted to the target store into the fitted regression model, outputs a sales forecast result for the target store, and outputs a distribution map of existing stores in the same industry within the survey area, a distribution map of the number of customers attracted to the target store, a population pyramid of the customer population, and a pie chart showing the size composition of the customer households. In the aforementioned process of predicting the number of visitors, A sales forecasting method that predicts the number of customers attracted to a target store using a spatial interaction model, utilizing population data within a predetermined range as the demand for the target store and existing stores within the survey area.

9. Computers, A store location and attribute information acquisition unit acquires the location and attribute information of target stores for which sales forecasting is performed, A prediction area setting unit that sets a prediction area with a first radius centered on the location of the target store, the prediction area setting unit that sets the first radius such that the number of existing stores within the prediction area from which sales data can be obtained satisfies a first condition, A survey area setting unit for setting a second radius survey area centered on the location of the target store, wherein the second radius is set by adding a third radius of the core trading area of ​​stores of the same type as the target store as a boundary condition to the first radius of the prediction area, For the aforementioned target stores and existing stores within the aforementioned survey area, there is a customer attraction forecasting unit that predicts the number of customers attracted, For existing stores within the aforementioned forecast area where sales data can be obtained, a regression model fitting unit is used to fit a regression model that predicts sales, with the predicted number of customers as the explanatory variable and sales as the dependent variable. The system functions as a prediction result output unit, inputting the number of customers attracted to the target store into the aforementioned fitted regression model, outputting the sales forecast results for the target store, and also outputting a distribution map of existing stores in the same industry within the survey area, a distribution map of the number of customers attracted to the target store, a population pyramid of the customer base, and pie charts showing the size composition of customer households. The aforementioned customer population forecasting unit is, This program uses population data within a specified range to predict the number of customers attracted to a target store and existing stores within the survey area, employing a spatial interaction model.

Images