JP7667103B2 - Demand forecasting device and demand forecasting method - Google Patents

Description

The present invention relates to a demand forecasting device and a demand forecasting method.

For make-to-stock products, which require a lead time from manufacturing to supply, it is necessary to predict product demand in advance and then procure parts, assemble the product, and transport it to the market based on the prediction results. In particular, for products that have the same basic structure but come in a variety of specifications based on differences in color, the presence or absence of additional functions, etc., the parts required and production methods differ for each specification, so it is essential to predict demand for each product specification.

Regarding product demand forecasting, for example, Patent Document 1 describes an "integrated demand forecasting device that constructs multiple demand forecasting models based on actual demand amounts stored in a demand actual amount storage means that stores actual demand amounts in the past, and calculates a predicted demand amount for each of the constructed demand forecasting models."

JP 2013-131259 A

The technology described in Patent Document 1 can predict the demand for a product group (a collection of products with the same basic structure. For example, in the case of refrigerators, a group of refrigerators made by the same manufacturer and with the same capacity and grade (high-end, standard, etc.)) based on past demand performance. However, the demand for products that belong to the same product group but have different specifications (for example, in the case of refrigerators, differences in color, door opening and closing direction, etc.) can be affected by the implementation or non-implementation of promotional (sales promotion) measures such as publication in flyers or e-mail magazines, discounts, and changes to in-store displays, and therefore cannot be predicted accurately based on past demand performance alone.

The present invention has been made in consideration of the above points, and aims to make it possible to accurately predict the demand for products that belong to the same product group but have different specifications, which may be affected by the implementation of promotional measures.

This application includes multiple means for solving at least some of the above problems, examples of which are as follows:

In order to solve the above problem, a demand forecasting device according to one embodiment of the present invention includes a convergence time calculation unit that refers to sales performance information indicating the sales performance of each product and parent-child item correspondence information indicating the correspondence between child items and parent items of a product to extract the sales performance of a parent item corresponding to a target product as the child item for which demand volume is to be predicted, and calculates a convergence time T indicating the time when the demand composition ratio of multiple products with different specifications belonging to a product group to which the target product belongs converges based on the sales performance of the parent item; a composition ratio calculation unit that determines a calculation method using the convergence time T as a threshold value and calculates the composition ratio of the target product based on at least one of the sales performance of the parent item and the sales performance of the target product according to the determined calculation method; and a composition ratio update unit that updates the calculated composition ratio using an update coefficient according to a sales promotion measure.

The present invention makes it possible to accurately predict the demand for products with different specifications that belong to the same product group and that may be affected by the implementation or non-implementation of promotional measures.

Problems, configurations, and advantages other than those described above will become clear from the description of the embodiments below.

FIG. 1 is a block diagram illustrating an example of the configuration of a demand forecasting system. FIG. 2 is a block diagram showing an example of the configuration of a typical computer. FIG. 3 is a diagram showing an example of a data structure of sales performance information. FIG. 4 is a diagram showing an example of a data structure of parent-child item correspondence information. FIG. 5 is a diagram showing an example of a data structure of the product exhibition information. FIG. 6 is a diagram showing an example of a data structure of the media publication information. FIG. 7 is a diagram showing an example of a data structure of the sales promotion measure information. FIG. 8 is a diagram illustrating an example of a data structure of the update coefficient information. FIG. 9 is a diagram illustrating an example of a data structure of the component ratio prediction information. FIG. 10 is a diagram showing an example of the data structure of the updated component ratio information. FIG. 11 is a diagram illustrating an example of a data structure of product group demand forecast information. FIG. 12 is a diagram illustrating an example of a data structure of the product master information. FIG. 13 is a diagram for explaining the concept of demand forecasting. FIG. 14 is a flowchart illustrating an example of a demand forecasting process. FIG. 15 is a flowchart illustrating an example of the process of step S3 in detail. FIG. 16 is a flowchart illustrating an example of the process of step S4 in detail. FIG. 17 is a diagram showing an example of the calculation result obtained by the process in step S45. FIG. 18 is a diagram for explaining an example of the process of step S47 in detail. FIG. 19 is a flowchart illustrating an example of the process of step S11 in detail. FIG. 20 is a diagram showing a display example of the output screen. FIG. 21 is a diagram showing an example of the use of the demand forecasting system.

Hereinafter, an embodiment of the present invention will be described with reference to the drawings. In all drawings for explaining the embodiment, the same components are generally given the same reference numerals, and repeated description will be omitted. In addition, in the following embodiments, the components (including element steps, etc.) are not necessarily essential, unless otherwise specified or considered to be obviously essential in principle. In addition, when it is said that "consists of A," "consists of A," "has A," or "includes A," it is needless to say that other elements are not excluded, except when it is specifically specified that only that element is included. Similarly, in the following embodiments, when referring to the shape, positional relationship, etc. of components, etc., it is intended to include those that are substantially similar or similar to the shape, etc., except when it is specifically specified or considered to be obviously not essential in principle.

<Configuration example of a demand forecasting system 1 according to an embodiment of the present invention>
1 shows an example of the configuration of a demand forecasting system 1 according to an embodiment of the present invention. The demand forecasting system 1 predicts the demand for a product that is uniquely determined by a product group and specifications. A product group is a collection of products that have the same basic structure. For example, in the case of refrigerators, the product group refers to a group of refrigerators manufactured by the same manufacturer, have the same capacity and grade (high-end, standard, etc.), and are assigned a common model number. The specifications refer to differences in color and door opening and closing directions of refrigerators assigned a common model number.

The demand forecasting system 1 includes a demand forecasting device 10, a database 20, and a user terminal 30, which are connected via a network 2. The network 2 is a two-way communication network such as the Internet.

The demand forecasting device 10 is realized by a general computer such as a personal computer.

Figure 2 shows an example of the configuration of a general computer 100 that realizes the demand forecasting device 10. The computer 100 includes a processor 101, a ROM 102, a RAM 103, an auxiliary storage device 104, a display device 105, an input device 106, a media reading device 107, and an information receiving/transmitting device 108.

The processor 101 is made up of a central processor unit (CPU) and executes various calculations according to a predetermined program. The read only memory (ROM) 102 stores programs required for starting the computer 100. The RAM 103 stores programs executed by the processor 101 and data required for executing the programs. The auxiliary storage device 104 is made up of, for example, a hard disk drive (HDD) or a solid state drive (SSD), and stores programs and various data. The display device 105 is made up of a liquid crystal display, an organic electroluminescence (EL) display, and displays various screens. The input device 106 is made up of a keyboard, a mouse, a microphone, and accepts input of commands, voice, and the like from the user. The media reader 107 reads data from a removable storage medium such as a semiconductor memory, and records data on a recording medium. The information receiver/transmitter 108 is made up of a communication module and the like, and transmits and receives data via the network 2.

In the computer 100, the processor 101 loads a specific program stored in the auxiliary storage device 104 into the RAM 103 and executes it to realize each functional block of the demand forecasting device 10.

Returning to FIG. 1, the demand forecasting device 10 includes the functional blocks of a memory unit 11, a processing unit 12, and a communication unit 13.

The memory unit 11 is realized by the RAM 103 and the auxiliary storage device 104 of the computer 100. The memory unit 11 stores sales performance information 111, parent-child item correspondence information 112, product exhibition information 113, media publication information 114, promotional measure information 115, update coefficient information 116, composition ratio forecast information 117, updated composition ratio information 118, product group demand forecast information 119, and product master information 1110.

The sales performance information 111 is information that represents the sales performance of the product. The sales performance information 111 is stored in advance in the database 20, and is obtained from the database 20 via the network 2 by the communication unit 13 and stored in the memory unit 11.

Figure 3 shows an example of the data structure of sales performance information 111. Sales performance information 111 includes cumulative sales performance 111a and composition ratio trend information 111b.

The cumulative sales performance 111a represents the sales performance of each product, which is uniquely identified by its product group (model) and specifications, from the time of its launch. Hereinafter, for example, a product belonging to the product group RS50 and with specifications LA will be referred to as product RS50-LA. The same applies to other products.

In the case of the figure, for example, the cumulative sales volume of the product RS50・LA is 20,000 units, and the cumulative sales volume of the product RS50・RA is 40,000 units. Note that the cumulative sales volume "-" indicates that the corresponding product has not been sold and is scheduled to be sold.

The composition ratio transition information 111b shows the weekly transition of the composition ratio of the products that make up the product group, calculated based on the cumulative sales performance 111a. In the case of the figure, the composition ratios of the products RS50-LA, RS50-RA, RS50-LB, and RS50-RB, which belong to the product group RS50 but have different specifications, are 50%, 30%, 20%, and 0% in the first week after the start of sales, 30%, 30%, 20%, and 20% in the second week, and 20%, 40%, 30%, and 10% in the 52nd week.

Returning to FIG. 1, parent-child item correspondence information 112 is information that indicates the correspondence between new products (child items) that are on sale or are scheduled to be on sale, and their previous generation old products (parent items). Parent-child item correspondence information 112 is stored in advance in database 20, and is obtained from database 20 via network 2 by communication unit 13 and stored in memory unit 11.

Figure 4 shows an example of the data structure of parent-child item correspondence information 112. In the case of this figure, for example, the parent item of the child item product RT50.LA is the product RS50.LA, and the parent item of the child item product RT50.RA is the product RS50.RA.

Returning to FIG. 1, the product display information 113 is information that indicates the display status of products in a store such as a consumer electronics retailer. The product display information 113 is stored in advance in the database 20, and is obtained from the database 20 via the network 2 by the communication unit 13 and stored in the memory unit 11.

Figure 5 shows an example of the data structure of product display information 113. In this figure, the date is omitted, and for example, it shows that product RT50・LA is not on display in store at store "X Electric" but is on display in store at store "Y Mall".

Returning to FIG. 1, media publication information 114 is information that indicates the publication ratio in media of products that belong to the same product group but have different specifications. Media publication information 114 is calculated in advance based on the number of images of each product published on EC (Electronic Commerce) sites, each web page, etc., and the publication ratio of each product (number of images published of each product / total number of images published of products belonging to the product group) and stored in database 20. Communication unit 13 obtains the information from database 20 via network 2 and stores it in memory unit 11. The number of images published of each product can be obtained, for example, by using a method such as image mining on web pages.

Figure 6 shows an example of the data structure of media publication information 114. In this figure, the dates are omitted, and the media publication ratios of four products RT50・LA, RT50・RA, RT50・LB, and RT50・RB that belong to the same product group are 30%, 50%, 15%, and 5%.

Returning to FIG. 1, the promotional measure information 115 is information indicating the implementation date (including the date that has already been implemented and the date that is planned) and the type of promotional measure for each product. The promotional measure information 115 is stored in advance in the database 20, and is obtained from the database 20 via the network 2 by the communication unit 13 and stored in the memory unit 11. Types of promotional measures include discounts, flyers, e-mail newsletters, in-store displays, and media coverage. The user can change the schedule for future promotional measures in the promotional measure information 115 (FIG. 7).

Figure 7 shows an example of the data structure of the promotional measure information 115. In this figure, for example, it shows that for product RT50・LA, an e-mail newsletter was sent out on August 16th, and for product RT50・RA, a flyer was published on August 9th and an e-mail newsletter was sent out on August 16th (or will be implemented). Note that a promotional measure "-" indicates that no promotional measure will be implemented for that product.

Returning to FIG. 1, the update coefficient information 116 is information that represents an update coefficient for updating the composition ratio depending on whether or not a promotional measure is implemented. Promotional measures include discounts, flyers, and e-mail newsletters, as exemplified in FIG. 7, as well as in-store displays and media coverage. The update coefficient information 116 is calculated by the update coefficient calculation unit 124 and stored in the memory unit 11.

Figure 8 is an example of update coefficient information 116. In the case of this figure, for example, the update coefficient is 1.8 when flyers are published as a promotional measure, and the update coefficient is 0.7 when they are not published. This means that the composition ratio of products that are published in flyers as a promotional measure is multiplied by 1.8, and the composition ratio of products that are not published in flyers is multiplied by 0.7.

Returning to FIG. 1, the composition ratio forecast information 117 is information that represents the composition ratio convergence week T at which the composition ratio of the demand amount of products belonging to the same product group but with different specifications is predicted to converge, the composition ratio convergence start week S at which the convergence is predicted to start, and the composition ratio forecast value. The composition ratio convergence week T is calculated by the convergence time point calculation unit 121, the composition ratio convergence start week S by the convergence start time point calculation unit 122, and the composition ratio forecast value by the composition ratio calculation unit 123, and are stored in the memory unit 11 as the composition ratio forecast information 117.

Figure 9 shows an example of the data structure of the composition ratio prediction information 117. In the case of this figure, for example, the convergence week T of the composition ratio of the product with different specifications belonging to the product group RT50 is the 10th week from the start of sales, the convergence start week S is the 5th week from the start of sales, and the predicted values of the composition ratio of the products with different specifications RT50-LA, RT50-RA, RT50-LB, and RT50-RB belonging to the product group RT50 are 32%, 38%, 25%, and 5%.

Returning to FIG. 1, the updated composition ratio information 118 is information that represents the trend of the predicted value of the composition ratio that is updated depending on the presence or absence of promotional measures for products that belong to the same product group but have different specifications. The updated composition ratio information 118 is calculated by the composition ratio update unit 125 and stored in the memory unit 11.

Figure 10 shows an example of the data structure of the updated composition ratio information 118. In the case of this figure, for example, the predicted values of the updated composition ratios of the products RT50-LA, RT50-RA, RT50-LB, and RT50-RB with different specifications belonging to the product group RT50 are 32%, 38%, 25%, and 5% on August 2nd, 20%, 61%, 16%, and 3% on August 9th, and 32%, 38%, 25%, and 5% on August 16th.

Returning to FIG. 1, product group demand forecast information 119 is information that represents a forecasted value of the demand amount for each product group. Product group demand forecast information 119 is stored in advance in database 20, and is acquired from database 20 via network 2 by communication unit 13 and stored in memory unit 11. The demand amount for each product group is forecast using techniques such as machine learning and autoregressive integrated moving average model based on statistical information such as the market size, economic situation, and population size of each country.

Figure 11 shows an example of the data structure of product group demand forecast information 119. In this figure, the demand forecast value for product group RT50 is 300 units on August 2nd, 500 units on August 9th, and 350 units on August 16th.

Returning to FIG. 1, product master information 1110 is information that indicates the product category, product group, specifications, sales start date, and sales end date of each product. In the case of household appliances, for example, product category refers to the distinction between refrigerators, washing machines, TVs, etc. Product master information 1110 is stored in advance in database 20, and is obtained from database 20 via network 2 by communication unit 13 and stored in memory unit 11.

Figure 12 shows an example of the data structure of product master information 1110. In this example, the sales start date of product RS50・LA, which belongs to product category A1, is April 1, 2020, and the sales end date is April 5, 2021.

Of the above-mentioned information, sales performance information 111, parent-child item correspondence information 112, product exhibition information 113, media coverage information 114, sales promotion policy information 115, product group demand forecast information 119, and product master information 1110 may not be stored in the memory unit 11, and each piece of information stored in the database 20 may be referenced as necessary.

Returning to FIG. 1, the processing unit 12 is realized by the processor 101 of the computer 100. The processing unit 12 has a convergence time point calculation unit 121, a convergence start time point calculation unit 122, a composition ratio calculation unit 123, an update coefficient calculation unit 124, a composition ratio update unit 125, a demand amount calculation unit 126, and an output control unit 127.

Based on the sales performance information 111, the parent-child item correspondence information 112, and the product master information 1110, the convergence time calculation unit 121 calculates the time (expressed in weeks starting from the sales start date; hereafter referred to as the composition ratio convergence week T) when the composition ratio of the demand for products that belong to the same product group but have different specifications is predicted to converge in accordance with the convergence time calculation process (FIG. 15) described below, and stores this in the composition ratio forecast information 117.

The convergence start time calculation unit 122 calculates the time (expressed in weeks starting from the sales start date; hereafter referred to as the composition ratio convergence start week S) when the composition ratio of the demand for products with different specifications that belong to the same product group is predicted to start converging, based on the sales performance information 111, parent-child item correspondence information 112, composition ratio forecast information 117, and product master information 1110, in accordance with the convergence start time calculation process (FIG. 16) described below, and stores the time in the composition ratio forecast information 117.

The composition ratio calculation unit 123 calculates a predicted value (predicted value α, β, or γ) of the composition ratio of the demand for products that belong to the same product group but have different specifications based on the sales performance information 111, parent-child item correspondence information 112, composition ratio forecast information 117, and product master information 1110 in accordance with the composition ratio calculation process (FIG. 14) described below, and stores this in the composition ratio forecast information 117 as the composition ratio forecast value.

The update coefficient calculation unit 124 calculates an update coefficient for each product based on the sales performance information 111, the product exhibition information 113, the media publication information 114, and the sales promotion policy information 115 in accordance with the update coefficient calculation process (FIG. 18) described below, and stores the calculated update coefficient in the update coefficient information 116.

The composition ratio update unit 125 multiplies the composition ratio of the demand for products that belong to the same product group but have different specifications by an update coefficient based on the product exhibition information 113, media publication information 114, sales promotion policy information 115, and update coefficient information 116 in accordance with the composition ratio update process (FIG. 13) described below, and further updates the composition ratio by normalizing it, and stores it in the updated composition ratio information 118.

The demand calculation unit 126 calculates the demand for each product by multiplying the product group demand forecast value by the composition ratio of each product based on the updated composition ratio information 118 and the product group demand forecast information 119.

The output control unit 127 causes the user terminal 30 to display an output screen showing the demand quantity of each product calculated by the demand quantity calculation unit 126.

The communication unit 13 is realized by the information receiving/transmitting device 108 of the computer 100. The communication unit 13 connects to the database 20 via the network 2, acquires various information, and stores it in the memory unit 11.

The database 20 is, for example, an ERP (Enterprise Resource Planning) system. It is assumed that the database 20 stores in advance sales performance information 111, parent-child item correspondence information 112, product exhibition information 113, sales promotion policy information 115, product group demand forecast information 119, and product master information 1110. The user terminal 30 is, for example, a general computer such as a personal computer, and is operated by a user who creates a production plan for a product based on the forecasted value of product demand. The user terminal 30 presents the user with an output screen that shows the forecasted value of product demand, etc. obtained from the demand forecasting device 10.

Next, FIG. 13 is a diagram explaining the concept of demand forecasting in this embodiment.

In this embodiment, the following characteristics (1) and (2) regarding product demand are utilized.
(1) The composition ratio of products (child items) that belong to the same product group but have different specifications is correlated with the composition ratio of the corresponding parent items.
(2) The composition ratio of products with different specifications that belong to the same product group will eventually converge due to factors such as ergonomics (for example, since more Japanese people are right-handed than left-handed, there is greater demand for refrigerator doors that open from the right rather than the left so that they can be opened and closed with the right hand), consumer psychology, living environment, etc.

The period after the product's sales start is divided into a first period from the sales start to composition ratio convergence start week S (excluding composition ratio convergence start week S), a second period from composition ratio convergence start week S to composition ratio convergence start week T (including composition ratio convergence start week S but excluding composition ratio convergence week T), and a third period after composition ratio convergence week T (including composition ratio convergence week T), and the method of calculating the composition ratio is changed depending on which of the first to third periods the planning date (target date for demand forecasting) belongs to.

Specifically, as shown in the upper part of the figure, if the planning date, week N, falls in the first period (if N<S), the composition ratio of the target product (child item) is the composition ratio α of the parent item after convergence, calculated based on the cumulative sales performance of the corresponding parent item.

Furthermore, as shown in the middle of the figure, when the Nth week, which is the planning date, belongs to the second period (S≦N<T), the composition ratio of the product (child item) to be forecasted is the composition ratio γ calculated using the composition ratio α and the composition ratio β in accordance with the following formula (1).
γ={α×(TN)+β×(NS+1)}/(TS+1)...(1)

Here, the composition ratio β is a composition ratio calculated based on the cumulative sales performance of the target product (child item) at the current time. The composition ratio β converges after the composition ratio convergence week T. In the second period, the composition ratio of the target product (child item) transitions from α to β, so the composition ratio γ is calculated by combining the composition ratio α and the composition ratio β. Specifically, when week N is close to the composition ratio convergence start week S, the proportion of the parent item's composition ratio α is high, and as week N approaches the composition ratio convergence week T, the proportion of the composition ratio β becomes higher.

Furthermore, as shown in the lower part of the figure, if the planning date, week N, falls in the third period (T≦N), the composition ratio of the target product (child item) is the composition ratio β, which is calculated based on the cumulative sales performance of the product.

<Demand forecasting process by the demand forecasting device 10>
Next, FIG. 14 is a flowchart illustrating an example of a demand forecasting process performed by the demand forecasting device 10.

The demand forecasting process, for example, predicts the demand for a group of products specified by the user using the user terminal 30, or for multiple products with different specifications that belong to the product group to which the product specified by the user belongs. The demand forecasting process is automatically executed at a predetermined timing, for example, by batch processing or the like.

First, the communication unit 13 connects to the database 20 via the network 2, acquires sales performance information 111, parent-child item correspondence information 112, product exhibition information 113, media publication information 114, sales promotion policy information 115, and product master information 1110, and stores them in the memory unit 11 (step S1).

Next, the processing unit 12 sets the product group specified by the user, or the product group to which the product specified by the user belongs, as the target product group for demand forecasting, and sequentially focuses on multiple products that belong to the target product group and have different specifications, and executes the following steps S3 to S10 as the target products for demand forecasting (step S2).

Specifically, the convergence time calculation unit 121 refers to the sales performance information 111, the parent-child item correspondence information 112, and the product master information 1110, and calculates the composition ratio convergence week T of the target product based on the sales performance of the parent item corresponding to the target product (child item) (step S3).

Figure 15 is a flowchart detailing an example of the processing of step S3. First, the convergence time calculation unit 121 refers to the parent-child item correspondence information 112 to identify the parent item that corresponds to the target product (child item) (step S21). If the parent item that corresponds to the target product (child item) is not recorded in the parent-child item correspondence information 112, the product master information 1110 can be referenced to determine, as the parent item, the product that belongs to the same product category as the target product (child item) and has similar codes representing the product group model and specifications, and whose sales start date is earlier than and closest to the sales start date of the target product (child item).

Next, the convergence time calculation unit 121 extracts sales performance information 111 of the parent item of the target product (child item) (step S22). Next, the convergence time calculation unit 121 determines the period from the start of sales of the parent item to the end of sales in weeks based on the product master information 1110 as the target week t (step S23).

Next, the convergence point calculation unit 121 refers to the composition ratio transition information 111b of the sales performance information 111, and calculates Δ shown in the following equation (2), that is, the square of the difference between the composition ratio of the parent item corresponding to the target product (child item) in the target week t and the composition ratio in the previous week (t-1) (step S24).
Δ = {(composition ratio in target week t) - (composition ratio in previous week (t-1))} ²
... (2)

Next, the convergence time point calculation unit 121 determines whether Δ is less than or equal to ^0.052 , that is, whether the change in the composition ratio of the parent item between the target week t and the previous week (t-1) is within ±5% (step S25).

Here, if it is determined that Δ is not ^0.05.2 or less (NO in step S25), the convergence time calculation unit 121 determines whether the target week t is the end week of the sales week (step S26), and if it is determined that the target week t is not the end week of the sales week (NO in step S26), it initializes the count variable n to 0 (step S27). Note that the count variable n represents the number of consecutive weeks in which Δ is ^0.05.2 or less, and its initial value is 0. Thereafter, the process returns to step S23, the target week t is incremented by 1, and steps S23 and after are repeated. Note that if the convergence time calculation unit 121 determines that the target week t is the end week of the sales week (YES in step S26), the process proceeds to step S30.

On the other hand, if it is determined that Δ is equal to or smaller than ^0.052 (YES in step S25), the convergence time point calculation unit 121 increments the count variable n by 1 (step S28).

Next, the convergence time calculation unit 121 determines whether the count variable n is 3, and whether the target week t is the sales end week (step S29). If it is determined that the count variable n is not 3 and the target week t is not the sales end week (NO in step S29), the process returns to step S23, the target week t is incremented by 1, and steps S23 and after are repeated.

Conversely, if the count variable n is 3 and/or it is determined that the target week t is the sales end week (YES in step S29), processing proceeds to step S30.

Next, the convergence time point calculation unit 121 determines the target week t as the composition ratio convergence week T of the target product (child item) and records this in the composition ratio forecast information 117 (step S30).

As described above, after the convergence time point calculation unit 121 determines the composition ratio convergence week T of the target product (child item) based on the sales performance of the parent item corresponding to the target product (child item), the process proceeds to step S4 in FIG. 14.

Next, the convergence start time calculation unit 122 calculates the convergence start week S when the composition ratio of the target product starts to converge based on the sales performance information 111, the parent-child item correspondence information 112, the composition ratio forecast information 117, and the product master information 1110 (step S4).

Figure 16 is a flowchart detailing an example of the processing of step S4. First, the convergence start time calculation unit 122 refers to the parent-child item correspondence information 112 to identify the parent item of the target product (child item) and its parent item (two generations before the child item) (step S41). Note that if the parent item of the target product (child item) is not recorded in the parent-child item correspondence information 112, as in step S21 of Figure 15, among products that belong to the same product category as the target product (child item) and have similar codes representing the product group model and specifications, the product whose sales start date is before the sales start date of the target product (child item) and is closest to the sales start date of the target product (child item) can be regarded as the parent item. The same applies to the parent item of a parent item.

Next, the convergence start time calculation unit 122 extracts the parent item of the target product (child item) and the sales performance information 111 of that parent item (step S42).

Next, the convergence start time calculation unit 122 refers to the composition ratio transition information 111b of the sales performance information 111 of the target product (child item) two generations ago, and identifies the composition ratio of the target product (child item) two generations ago after convergence as the composition ratio α of the parent item of the target product (child item). The convergence start time calculation unit 122 also refers to the composition ratio transition information 111b of the sales performance information 111 of the parent item of the target product (child item), and identifies the composition ratio β of the parent item of the target product (child item) after convergence (step S43).

Next, the convergence start time calculation unit 122 assumes the convergence start week S to be one week to T weeks from the parent item's sales start date to the composition ratio convergence week T calculated in step S3, based on the product master information 1110 (step S44).

Next, the convergence start time calculation unit 122 calculates the composition ratio of the target product (child item) for week N when the planning date, week N, is changed from week 1 to week T while keeping the convergence start week S fixed, according to the case classification shown in Figure 13 (step S45).

Figure 17 shows an example of the calculation results in step S45. However, this figure shows a case where the composition ratio convergence week T calculated in step S3 is 10, the composition ratio α of the parent item of the target product (child item) calculated in step S43 is 0.270, and the composition ratios β based on the cumulative sales performance for each of N = 1 to 10 are 0.381, 0.397, ..., 0.270.

For example, when the convergence start week S=1, S≦N when week N=1, and week N belongs to the second period shown in FIG. 13, so that the component ratio γ is calculated according to formula (1).
γ=(0.270×(10-1)+0.381×1)/10=0.281

Furthermore, for example, when the convergence start week S=2, when week N=1, N<S and week N belongs to the first period shown in Fig. 13, so the component ratio α=0.270 is adopted. When week N=2, S≦N and week N belongs to the second period shown in Fig. 13, so the component ratio γ is calculated according to formula (1) as follows:
γ=(0.270×(10-2)+0.397×2)/10=0.295

Furthermore, when N weeks = 10, T≦N, and N weeks belongs to the third period shown in FIG. 13, so the composition ratio β = 0.270 is adopted.

Furthermore, for example, when the convergence start week S = 10, when N week = 1 to 9, N < S, and N week belongs to the first period shown in Figure 13, so the composition ratio α = 0.270 is adopted, and when N week = 10, T ≦ N, and N week belongs to the third period shown in Figure 13, so the composition ratio β = 0.270 is adopted.

Returning to FIG. 16, the convergence start time calculation unit 122 next determines whether the current convergence start week S is the same as the composition ratio convergence week T (step S46). If it is determined that the current convergence start week S is not the same as the composition ratio convergence week T (NO in step S46), the process returns to step S44, S is incremented by 1, and steps S44 to S46 are repeated.

If it is determined that the current convergence start week S is the same as the composition ratio convergence week T (YES in step S46), the convergence start point calculation unit 122 then calculates the error between the composition ratio of the target product (child item) in week N of the target product (child item) when week N, the planning date, is changed from week 1 to week (T-1) for each possible value (1 to T) of the convergence start week S, and the composition ratio β based on the cumulative sales performance for M weeks (M = 1 to 8) after week N, and sets the S with the smallest sum as the composition ratio convergence start week S of the target product (step S47). Here, M = 1 represents the same week as week N, and M = 2 represents the week following week N.

Figure 18 is a diagram for explaining step S47 in detail. From Figure 17, for example, in the convergence start week S=1, the composition ratio for N=1 is 0.281, and the composition ratio β for M=1 is 0.381, so the error is |0.281-0.381| = 0.100. Also, the composition ratio β for M=2 is 0.397, so the error is |0.281-0.397| = 0.116.

For example, in the convergence start week S=1, the composition ratio for N=2 is 0.295, and the composition ratio β for M=1 is 0.397, so the error is |0.295-0.397| = 0.102. Also, the composition ratio β for M=2 is 0.372, so the error is |0.295-0.372| = 0.077.

Then, with S fixed, the value of S that minimizes the sum of the errors calculated for each combination of N and M is set as the composition ratio convergence start week S for the target product and is stored in the composition ratio forecast information 117.

As described above, after the convergence start time calculation unit 122 determines the composition ratio convergence start week S of the target product (child item), processing proceeds to step S5 in FIG. 14.

Then, the composition ratio calculation unit 123, for example, according to input from the user, identifies the planned target date (N weeks starting from the sales start date of the target product) for predicting the composition ratio of the target product (step S5). Next, the composition ratio calculation unit 123 references the composition ratio convergence start week S and the composition ratio convergence week T from the composition ratio prediction information 117, and determines to which of the first to third periods the planned target date (N weeks) belongs (step S6). Then, depending on the result of this determination, the process branches to one of steps S7 to S9.

In other words, if it is determined that N weeks belongs to the first period, the composition ratio calculation unit 123 refers to the composition ratio transition information 111b in the sales performance information 111, and uses the composition ratio of the parent item of the target product (child item) after convergence as the composition ratio α and records it in the composition ratio forecast information 117 (step S7).

If it is determined that week N belongs to the second period, the composition ratio calculation unit 123 refers to the sales performance information 111 and calculates the composition ratio α based on the cumulative sales performance of the parent item of the target product (child item), calculates the composition ratio β based on the cumulative sales performance of the target product (child item), calculates the composition ratio γ according to formula (1), and records it in the composition ratio forecast information 117 (step S8).

In addition, if it is determined that week N belongs to the third period, the composition ratio calculation unit 123 refers to the sales performance information 111, calculates the composition ratio β based on the cumulative sales performance of the target product (child item), and records it in the composition ratio forecast information 117 (step S9).

Then, the processing unit 12 determines whether or not there is a product that has not yet been focused on among multiple products that belong to the same product group as the target product but have different specifications (step S10). If it is determined that there is a product that has not yet been focused on (YES in step S10), the processing returns to step S2, and the next target product is determined, and steps S3 to S10 are then repeatedly executed.

Then, all products that belong to the same product group but have different specifications are treated as target products, and steps S3 to S10 are executed. If it is determined that there are no products that have not yet been considered (NO in step S10), the process proceeds to step S11.

Next, the update coefficient calculation unit 124 calculates an update coefficient based on the sales performance information 111, the product exhibition information 113, the media coverage information 114, and the sales promotion policy information 115 (step S11).

Figure 19 is a flowchart detailing an example of the processing of step S11. First, the update coefficient calculation unit 124 focuses on the promotional measures (discount, flyer publication, e-mail newsletter, in-store display, media publication) in order (step S51).

Next, the update coefficient calculation unit 124 refers to the product display information 113, media publication information 114, and promotional measure information 115, extracts the implementation date on which the promotional measure of interest was implemented for the products belonging to the target product group, and focuses on the extracted implementation dates in order (step S52). Note that if the promotional measure of interest is in-store display or media publication, the date on which each was changed will be regarded as the implementation date.

Next, the update coefficient calculation unit 124 refers to the sales performance information 111 and extracts the sales performance (cumulative sales performance) before and after the implementation of the promotional measure of interest for each product in the target product group (step S53). If the promotional measure of interest is in-store display or media coverage, the daily sales performance before and after the day on which the in-store display situation or the coverage rate changed is extracted. If the promotional measure of interest is discounts, flyer coverage, or e-mail newsletters, the weekly sales performance before and after the implementation date is extracted.

Next, the update coefficient calculation unit 124 calculates (sales performance after promotional measures are implemented)/(sales performance before promotional measures are implemented) for each product in the target product group that has undergone promotional measures and for each product that has not undergone promotional measures (step S54).

Then, the update coefficient calculation unit 124 determines whether or not there are any implementation dates among the implementation dates of the promotional measures that have not yet been focused on (step S55), and if it is determined that there are any implementation dates that have not yet been focused on (YES in step S55), the process returns to step S52 and steps S52 to S55 are repeated.

If it is determined there are no implementation dates that have not yet been focused on (NO in step S55), the update coefficient calculation unit 124 calculates the average of the values calculated in step S54 for each product in the target product group that has implemented a promotional measure for that product for the number of implementation dates, and stores this in the update coefficient information 116 as the update coefficient for when the promotional measure of interest has been implemented. In addition, the update coefficient calculation unit 124 calculates the average of the values calculated in step S54 for each product in the target product group that has not implemented a promotional measure for that product for the number of implementation dates, and stores this in the update coefficient information 116 as the update coefficient for when the promotional measure of interest has not been implemented (step S56).

Next, the update coefficient calculation unit 124 determines whether or not there are any promotional measures that have not yet been addressed (step S57), and if it determines that there are any promotional measures that have not yet been addressed (YES in step S57), the process returns to step S51, and steps S51 to S57 are repeated.

If it is subsequently determined that there are no promotional measures that have not yet been taken into consideration (NO in step S57), the update coefficients for when each promotional measure is implemented and when it is not implemented will have been obtained, and processing will proceed to step S12 in FIG. 14.

Next, the composition ratio update unit 125 updates the composition ratio of the products belonging to the target product group based on the update coefficient information 116 and the composition ratio prediction information 117, and stores the updated composition ratio information 118 (step S12).

More specifically, the composition ratio of the products belonging to the target product group stored in the composition ratio forecast information 117 is multiplied by the update coefficient stored in the update coefficient information 116 in accordance with the implementation schedule of the promotional measures represented by the promotional measure information 115. Furthermore, the composition ratio is updated by normalizing the values obtained by multiplying the composition ratio of the products belonging to the target product group by the update coefficients so that the sum of these values is 100%.

Specifically, for example, when forecasting demand for products belonging to product group RT50 on August 9, 2021, it can be seen from promotional measure information 115 (FIG. 7) that among the products with different specifications belonging to product group RT50, product RT50-RA will be featured in a flyer as a promotional measure, but not products RT50-LA, RT50-LB, and RT50-RB. Also, it can be seen from update coefficient information 116 (FIG. 18) that the adjustment coefficient when flyers are featured is 1.8, and when they are not featured is 0.7. Furthermore, it can be seen from composition ratio forecast information 117 (FIG. 9) that the composition ratios of LA for products with different specifications RT50-LA, RT50-RA, RT50-LB, and RT50-RB belonging to product group RT50 are 32%, 38%, 25%, and 5%, respectively. Therefore, the composition ratio after updating and normalizing the RT50 and RA products is {38% x 1.8 / (32% x 0.7 + 38% x 1.8 + 25% x 0.7 + 5% x 0.7)} = 61%.

Next, the demand calculation unit 126 calculates the demand forecast value of each product belonging to the target product group by multiplying the demand forecast value for each product group represented by the product group demand forecast information 119 by the updated composition ratio of each product stored in the updated composition ratio information 118 (step S13). Then, the output control unit 127 displays an output screen 200 (Figure 20) showing the calculated demand forecast value of each product belonging to the target product group on the user terminal 30 connected via the network 2 (step S14). This completes the demand forecast process by the demand forecasting device 10.

This demand forecasting process makes it possible to accurately predict the demand for each product in the same product group, taking into account the implementation of promotional measures.

Next, FIG. 20 shows an example of the display of the output screen 200 displayed on the user terminal 30. The output screen 200 displays the weekly demand forecast value for each product belonging to the same product group and the planned promotional measures. For example, the demand forecast values for the products RT50・LA, RT50・RA, RT50・LB, and RT50・RB, which are different specifications and belong to the product group RT50, for the week of August 9, 2021 are 80, 244, 64, and 12, respectively, and it can be confirmed that the publication of flyers is planned as a promotional measure only for the product RT50・RA. As described above, the user can change the implementation schedule of the promotional measures in the promotional measure information 115 (FIG. 7), and the effect of the change can be confirmed on the output screen 200.

Next, FIG. 21 shows an example of the use of the demand forecasting system 1 of this embodiment. The demand forecasting system 1 can be used in a supply chain 2000 that is composed of a manufacturing and distribution community 2001 and a market 2002.

The manufacturing and distribution community 2001 is a collection of companies that procure materials and parts, manufacture products, and transport, ship, and sell the products. For example, in the automobile industry, it is a collection of companies such as parts manufacturers, finished vehicle manufacturers, and vehicle mass retailers. The market 2002 orders products from the manufacturing and distribution community 2001, and the manufacturing and distribution community 2001 delivers the products to the market 2002.

In optimizing a supply chain consisting of a manufacturing and distribution community 2001 and a market 2002, a sales plan is generated using highly accurate demand forecasts for products with different specifications that belong to the same product group as input, and an MRP (Material Requirements Planning) system is applied in multiple stages. Inputting the sales plan into a supply and demand optimization system 2003 leads to a reduction in deviation from actual demand, making it possible to reduce inventory throughout the supply chain and prevent lost sales opportunities.

For example, the demand forecasting system 1 makes it possible to predict in advance orders from the market 2002 as the demand for each product group specification required as input to the supply and demand optimization system 2003. This enables the supply and demand optimization system 2003 to appropriately issue part procurement instructions, production instructions, transportation instructions, shipping instructions, etc. to suppliers, factories, warehouses, sales companies, etc. that are the bases of each manufacturing and distribution community 2001, thereby matching supply and demand and enabling the reduction of inventory and the prevention of lost sales opportunities.

The present invention is not limited to the above-described embodiment, and various modifications are possible. For example, in this embodiment, sales plans, promotional measures plans, composition ratio forecasts, product group demand forecasts, and demand forecasts by product group specifications are calculated using weekly values. However, daily or monthly values may be used depending on the frequency with which sales plans are created.

For example, the above-mentioned embodiment has been described in detail to explain the present invention in an easy-to-understand manner, and is not necessarily limited to having all of the described configurations. In addition, the above-mentioned configurations, functions, processing units, processing means, etc. may be realized in hardware, partly or in whole, for example, by designing them as integrated circuits. In addition, the above-mentioned configurations, functions, etc. may be realized in software by a processor interpreting and executing a program that realizes each function. Information such as programs, tables, files, etc. that realize each function can be placed in a memory, a recording device such as a hard disk or SSD, or a recording medium such as an IC card, an SD card, or a DVD. In addition, the control lines and information lines are shown as those that are considered necessary for explanation, and do not necessarily show all of the control lines and information lines in the product. In reality, it can be considered that almost all of the configurations are connected to each other.

1: Demand forecasting system, 2: Network, 10: Demand forecasting device, 11: Memory unit, 111: Sales performance information, 111a: Cumulative sales performance, 111b: Composition ratio transition information, 112: Parent-child item correspondence information, 113: Product exhibition information, 114: Media publication information, 115: Sales promotion policy information, 116: Update coefficient information, 117: Composition ratio forecast information, 118: Updated composition ratio information, 119: Product group demand forecast information, 1110: Product master information, 12: Processing unit, 121: Convergence time point calculation unit, 122: Convergence start time point calculation unit, 123: Composition ratio calculation output unit, 124... update coefficient calculation unit, 125... composition ratio update unit, 126... demand calculation unit, 127... output control unit, 13... communication unit, 20... database, 30... user terminal, 100... computer, 101... processor, 102... ROM, 103... RAM, 104... auxiliary storage device, 105... display device, 106... input device, 107... media reading device, 108... information receiving/transmitting device, 200... output screen, 2000... supply chain, 2001... manufacturing and distribution community, 2002... market, 2003... supply and demand optimization system

Claims (6)

a convergence time calculation unit that refers to sales record information that indicates the sales record for each product and parent-child item correspondence information that indicates the correspondence between child items and parent items of the product, extracts the sales record of a parent item that corresponds to a target product as the child item for which demand volume is to be predicted, and calculates a convergence time T that indicates the time when the demand composition ratio of multiple products with different specifications that belong to a product group to which the target product belongs will converge, based on the sales record of the parent item;
a composition ratio calculation unit that determines a calculation method using the convergence time point T as a threshold value, and calculates the composition ratio of the target product based on at least one of the sales record of the parent item and the sales record of the target product in accordance with the determined calculation method;
a composition ratio update unit that updates the calculated composition ratio by using an update coefficient corresponding to a sales promotion measure;
a demand calculation unit that calculates the demand for the target product by multiplying a demand prediction value for a product group to which the target product belongs by the updated composition ratio ,
The composition ratio calculation unit assumes that the first week to the Tth week from the sales start date of the parent item until the composition ratio convergence week T is a convergence start week S in order, and calculates the composition ratio of the target product in week N when week N, which is the planning date, is changed from week 1 to week T while the convergence start week S is fixed,
When it is determined that the target product belongs to the first period from the start of sales to the composition ratio convergence start week S, the composition ratio of the parent item of the target product after convergence is set as a composition ratio α,
When it is determined that the target product belongs to a second period from the composition ratio convergence start week S to the composition ratio convergence week T, a composition ratio α is calculated based on a cumulative sales record of the parent item of the target product, and a composition ratio β is calculated based on a cumulative sales record of the target product, and a composition ratio γ is calculated using the composition ratio α and the composition ratio β,
When it is determined that the target product belongs to a third period after the composition ratio convergence week T, the composition ratio β is calculated based on the cumulative sales performance of the target product,
For each of the possible values (1 to T) of the convergence start week S, the error between the composition ratio of the target product in week N when week N, which is the planning date, is changed from week 1 to week (T-1), and the composition ratio β based on the cumulative sales performance for M weeks (M = 1 to 8) after week N is calculated, and the S with the smallest sum is set as the composition ratio convergence start week S of the target product.
With reference to a composition ratio convergence start week S and a composition ratio convergence week T, it is determined whether a planning target date for predicting a composition ratio of the target product input by a user belongs to the first period, the second period, or the third period;
When it is determined that the target product belongs to the first period, the composition ratio of the parent item of the target product after convergence is set as a composition ratio α;
When it is determined that the target product belongs to the second period, a composition ratio α is calculated based on a cumulative sales record of the parent item of the target product, and a composition ratio β is calculated based on a cumulative sales record of the target product, and a composition ratio γ is calculated using the composition ratio α and the composition ratio β;
If it is determined that the target product belongs to the third period, the composition ratio β is calculated based on the cumulative sales performance of the target product.
Demand forecasting device. 2. A demand forecasting device according to claim 1,
an update coefficient calculation unit that refers to sales promotion policy information that indicates a status of implementation of the sales promotion policy for the product, and calculates the update coefficient based on sales results of the target product before and after the implementation of the sales promotion policy for the target product;
A demand forecasting device comprising: 2. A demand forecasting device according to claim 1,
The sales promotion measures include at least one of discounts, flyers, e-mail newsletters, in-store displays, and media coverage. 2. A demand forecasting device according to claim 1,
The component ratio update unit is a demand forecasting device that updates the component ratio by multiplying the calculated component ratio by the update coefficient and normalizing it. 2. A demand forecasting device according to claim 1 ,
A demand forecasting device comprising an output control unit that displays a screen showing the calculated demand for the target product and promotional measures for the target product. A demand forecasting method using a demand forecasting device, comprising:
a convergence time calculation step of extracting sales results of parent items corresponding to a target product as a child item for which demand is to be predicted, by referring to sales results information indicating sales results for each product and parent-child item correspondence information indicating the correspondence between child items and parent items of the product, and calculating a convergence time T indicating the time when the demand composition ratio of multiple products with different specifications that belong to a product group to which the target product belongs will converge, based on the sales results of the parent items;
a composition ratio calculation step of determining a calculation method using the convergence time point T as a threshold value, and calculating the composition ratio of the target product based on at least one of the sales record of the parent item and the sales record of the target product in accordance with the determined calculation method;
a composition ratio update step of updating the calculated composition ratio by using an update coefficient according to a sales promotion measure;
a demand calculation step of multiplying a demand prediction value for a product group to which the target product belongs by the updated composition ratio to calculate the demand for the target product ,
The composition ratio calculation step assumes that the first week to the Tth week from the sales start date of the parent item until the composition ratio convergence week T is a convergence start week S in order, and calculates the composition ratio of the target product in week N when week N, which is the planning date, is changed from week 1 to week T while the convergence start week S is fixed,
When it is determined that the target product belongs to the first period from the start of sales to the composition ratio convergence start week S, the composition ratio of the parent item of the target product after convergence is set as composition ratio α,
When it is determined that the target product belongs to a second period from the composition ratio convergence start week S to the composition ratio convergence week T, a composition ratio α is calculated based on a cumulative sales record of the parent item of the target product, and a composition ratio β is calculated based on a cumulative sales record of the target product, and a composition ratio γ is calculated using the composition ratio α and the composition ratio β,
When it is determined that the target product belongs to a third period after the composition ratio convergence week T, the composition ratio β is calculated based on the cumulative sales performance of the target product,
For each of the possible values (1 to T) of the convergence start week S, the error between the composition ratio of the target product in week N when week N, which is the planning date, is changed from week 1 to week (T-1), and the composition ratio β based on the cumulative sales performance for M weeks (M = 1 to 8) after week N is calculated, and the S with the smallest sum is set as the composition ratio convergence start week S of the target product.
With reference to a composition ratio convergence start week S and a composition ratio convergence week T, it is determined whether a planning target date for predicting a composition ratio of the target product input by a user belongs to the first period, the second period, or the third period;
When it is determined that the target product belongs to the first period, the composition ratio of the parent item of the target product after convergence is set as a composition ratio α;
When it is determined that the target product belongs to the second period, a composition ratio α is calculated based on a cumulative sales record of the parent item of the target product, and a composition ratio β is calculated based on a cumulative sales record of the target product, and a composition ratio γ is calculated using the composition ratio α and the composition ratio β;
If it is determined that the target product belongs to the third period, the composition ratio β is calculated based on the cumulative sales performance of the target product.
Demand forecasting methods.

Images