JP4678879B2 - Sales prediction system, method and computer program - Google Patents

Description

  The present invention relates to a technique for recommending ordering / returning / exhibition at each individual store based on the individual characteristics of the individual store (one bookstore).

In the individual store, under the resale system, books are ordered and displayed regularly to the agency, and the remaining inventory is returned to the publisher through the agency in the form of returned goods.
In this case, the number and types of books issued at each individual store are enormous, and each individual store sells genres that sell well, for example, business books sell well in office districts and lifestyle-related books sell in residential districts. Etc. were different.

  Under such circumstances, individual stores usually depend on the intuition and experience of the store clerk to decide which books to order. For this reason, if the salesperson in charge of ordering is not expected, the book that has been ordered will not be sold, and there will be an increase in the number of books to be returned due to excess inventory.

  As a result, there is a problem that the distribution efficiency is deteriorated also in the agency.

In order to solve such problems, based on the sales performance and inventory status of each book managed in the sales / inventory information database, the sales prospect is calculated based on the sales volume in the past several years, etc., and the average sales volume is estimated. A mechanism for forecasting sales has been proposed (Patent Document 1).
There is also a mechanism for sharing various information such as the accurate sales situation and inventory status of the market through the network between individual stores, agency stores, and publishers, so that appropriate duplication, delivery, loss of opportunities, etc. can be prevented. It has been proposed (Patent Document 2).

JP2002-183548 JP 2002-7664 A

However, the inventions described in each of the above-mentioned prior art documents all count the number of volumes sold without considering the change in the sales trend due to the passage of time from the point in time in which the sales performance was obtained to the time the book ordered arrives. It was just a prediction.
In addition, all the inventions described in the prior literature perform sales predictions for all stores, but if this sales prediction for all stores is applied to sales predictions for individual stores, the influence of the characteristics of the individual stores and books is large. It was extremely difficult to predict sales at individual stores. In particular, since small individual stores sell fewer books in the first place, it is recommended that individual books be sold by chance. In small individual stores, individual books are recommended based on past sales performance. It was extremely difficult to predict sales.

Moreover, as for the individual store side, what can be sold at their own store even if it can refer to the information on the best selling of books in the whole country and the information on the current best-selling books at large major chain stores etc. The number of books to be ordered and the number of books to be returned when and how many were returned depend largely on the intuition and experience of the ordering clerk. For this reason, if the ordering and return are not performed properly, the proper books are not in stock, resulting in a loss of sales opportunities.
In addition, at individual stores, predictions based on the most recent sales results are often made, and sales forecasts that take into consideration the entire process from the start of sales at individual stores have not been made. For this reason, books were sometimes purchased despite the saturation of the retail store market.

  In addition, some books sold differently, such as those that were sold several months or years after their release, and books that were constantly sold. Such a feature often becomes more prominent as the number of days elapsed from the release date, and such a feature cannot be inferred from the past sales volume of individual books.

  In addition, a mechanism for sharing information among publishers, brokers, and individual stores was proposed, but a mechanism for forecasting and placing orders in consideration of time lags in information and logistics was not proposed. .

A mechanism for determining the number of books ordered based on the estimated number of sales was proposed, but the effect of the number of exhibitions on the number of sales was not considered.
For example, the last remaining book has a high probability of being unsold unless it is a top seller, so there is a minimum number of exhibits to sell.

  The present invention has been made to solve the above-described problems, and an object thereof is to provide a mechanism for recommending ordering / returning / exhibition for each individual store based on the characteristics of the individual store, book, and genre. And

A sales prediction system according to one aspect of the present invention is a system for predicting the number of books sold for each book for each individual store, and stores sales records collected from the sales management terminals of all the individual stores. With reference to the results storage means, the store sales forecast volume number storage means for storing the book sales forecast number of books in all stores, calculated according to the elapsed time since release for each book, and the sales results storage means, Store title performance coefficient calculation that calculates the store title performance coefficient that represents the sales scale at the individual store for the book, based on the ratio of the total sales volume at the individual store to the total sales volume at all stores for each book Stores that store the store genre scale factor representing the ratio of the average number of sales records of all stores classified into each genre and the number of sales records of the individual stores to be predicted for each genre for each store. Based on the estimated sales volume in all stores stored in the all-store sales predicted book volume storage means, at least one of the calculated store title performance coefficient and the store genre scale coefficient And a store sales forecast processing means for calculating the number of predicted sales at the store based on the coefficient.
Here, “all stores” refers to a set of stores with a certain unit, for example, a set of individual stores under a certain agency store, which is a set of a plurality of individual stores under that subordinate store. .
Moreover, the sales forecast book number means the sales forecast book number of any one store or all stores.
The genre is a book classification for each characteristic. For example, a genre includes not only a business book and a new book, but also a publisher.

An individual representing a weight using an elapsed time from the first delivery date when the book was delivered to the individual store for the first time, and a store genre scale factor and a store title performance factor at each individual store according to the elapsed time. It further has a store adjustment ratio parameter storage means for storing the store adjustment ratio parameter,
The individual store sales prediction processing means calculates the elapsed time from the first delivery date of the book for each individual store, refers to the individual store adjustment ratio parameter storage means, and corresponds to the calculated elapsed time. Even if the individual store adjustment ratio parameter is acquired, the store genre scale factor and the store title performance coefficient are weighted based on the acquired individual store adjustment ratio parameter, and the sales forecast volume at the individual store is calculated. Good.

  With reference to the above sales performance storage means, based on the past sales performance of the genre to which the book to be predicted belongs, the total sales from the first edition sales date to the first reference date (Nth day) of the book to be predicted The NM coefficient (t) is calculated from the ratio (q / p) with respect to the total number of books to be sold q up to the Mth day of the forecast date arbitrarily determined after the first reference date for the number of books p, and the first reference of the book The book sales forecast by the NM method for predicting the book sales volume based on the calculated NM coefficient (t) and the sales record storage means, and the forecast Based on the past sales performance of the genre to which the target book belongs, the error between the total sales forecast volume for the next predetermined period and the total sales volume for that period using the total sales volume for the past three predetermined periods Based on the incremental coefficient that minimizes All stores that predict the sales of all stores based on at least one of the sales prediction methods of the book sales by the incremental method, and store the total sales forecast book number of books in all stores in the store sales forecast number storage means described above You may further have a sales prediction process means.

Hybrid ratio parameter storage means for storing a hybrid ratio parameter, which is a weight when each of the sales forecast based on the NM method and the sales forecast based on the incremental method is used according to the elapsed time after the book is released Further comprising
The store sales forecast processing means refers to the hybrid ratio parameter storage means for the sales forecast based on the NM method and the sales forecast results based on the incremental method. Sales prediction based on the above two sales prediction methods may be performed by weighting with the hybrid ratio parameter according to time.

  The priority of the individual store shelf, the individual store display coefficient table storing the display coefficient and the maximum number of display books for each shelf, and the priority of the individual store display coefficient table in descending order of the predicted sales number of each book. Determine the recommended number of books and shelves for each book by assigning a high shelf and calculating the recommended number of books for the book based on the individual store exhibition coefficient of the assigned shelf by referring to the individual store exhibition coefficient table. In addition, an exhibition recommendation processing means for creating an exhibition recommendation list may be further included.

In the sales record storage means, in addition to the number of sales records for each book in the individual store, the number of inventory books, the number of delivered products, and the number of returned products are stored. Calculate the logical inventory up to the period N by adding the number of delivered books for each book from the point of inventory to the period N to the number of inventories in the store, and subtracting the number of returned books and the number of sales records for each book. By subtracting the number of N + 1 sales forecasts, the number of inventory forecasts for each book in period N + 1 is calculated. It may further include an output means for calculating the recommended number of books ordered by the individual store by subtracting the above-mentioned stock forecasted number of books of the period N + 1, and creating and outputting an order recommendation list based on the calculated recommended number of books ordered. .
Here, the aggregation date is a date for confirming stock for a predetermined period. For example, if it is weekly, it is the last day of the week.

  It further comprises inventory data storage means for storing book inventory data at the agency store, and the output means has been ordered from the individual store with reference to the sales performance storage means and has not been sent to the individual store. Referring to the number of book orders remaining and the above inventory data storage means, even if the book recommended for ordering is output in addition to the recommended ordering list, stock status information indicating whether the book is in stock or not is output. Good.

  The output means may create and output a return recommendation list from a book in which the recommended number of exhibited books calculated by the recommended exhibition processing means is 0 and a book in which the recommended number of ordered books is negative. .

  The sales prediction method according to one aspect of the present invention is calculated according to the sales record storage means for storing the sales record collected from the sales management terminals of all the individual stores, and the elapsed time since the release for each book. , Stored store sales forecast number storage means for storing the predicted number of books sold in all stores, and the average number of sales records of all stores classified into each genre for each book genre and the sales results of the above-mentioned individual stores to be predicted A computer having store genre scale coefficient storage means for storing a store genre scale coefficient representing the ratio of the number of books for each individual store, and a method for performing a process of predicting the number of books sold for each book for each individual store. , The computer refers to the sales record storage means, and for each book, the ratio of the total number of sales records at the individual store to the total number of sales records at the store, Store title performance coefficient calculation means processing for calculating a store title performance coefficient representing the model, and the calculated store title performance coefficient based on the predicted number of sales in all stores stored in the above-mentioned all-store sales prediction book number storage means Or, based on at least one of the store genre scale factors, the individual store sales prediction process for calculating the number of predicted sales at the individual store is performed.

  The computer program according to one aspect of the present invention is calculated according to the sales record storage means for storing the sales record collected from the sales management terminals of all the individual stores, and the elapsed time since the release for each book, Stored store sales forecast number storage means for storing the predicted number of books sold at all stores, and the average number of sales records of all stores classified into each genre for each book genre and the number of sales records of the above-mentioned individual stores to be predicted A store genre scale coefficient storing means for storing a store genre scale coefficient representing the ratio of each store, and a function for causing the computer to function as a sales prediction device, and Referring to the storage means, for each book, the ratio of the total sales volume at the relevant store to the total sales volume at all the stores represents the sales scale at that individual store. Based on the store title performance coefficient calculation means process for calculating the store title performance coefficient and the total sales forecast volume in all stores stored in the store sales forecast volume number storage means, the calculated store title performance coefficient or Based on at least one of the store genre scale coefficients, a store sales prediction process for calculating the number of predicted sales at the store is executed.

According to the present invention, it is possible to recommend ordering / returning / exhibition for each individual store based on the characteristics of the individual store, book, and genre. Therefore, ordering, returning, and display can be performed efficiently. As a result, the individual store does not have useless inventory, and it is possible to eliminate the loss of sales opportunities and the display of unsold books.
Even for a small individual store, sales prediction, display recommendation, and return recommendation can be made based on the characteristics of the individual store.

The number of books to be sold can be predicted in consideration of a change in the sales tendency with the passage of time from a certain point in the past when the sales performance is obtained to the point at which the book that has been ordered arrives.
In addition, it is possible to predict sales and place orders in consideration of time differences in information and logistics.

  By recommending the number of books required for display based on the expected number of sales, and recommending the number of orders and returns based on the number of books required for display, it is possible to prevent excessive display and shortage at stores. .

Next, embodiments of the present invention will be described with reference to the drawings.
FIG. 1 is a diagram showing the overall configuration of a system in which the present invention is applied to a system for ordering, sending and returning goods between a private store and a brokerage store.
In FIG. 1, there is an all-store management server 1, and an individual store terminal 2 and an agency store terminal 3 configured to be able to exchange data with the all-store management server 1 via a network or the like.

The individual store terminal 2 is a device that receives and outputs an exhibition / order recommendation list and a return recommendation list at each individual store transmitted from the all store management server 1. The individual store terminal 2 can be constituted by a so-called personal computer. This individual store terminal 2 is connected to the entire store management server 1 via the Internet, a telephone line, etc., acquires an exhibition / order recommendation list and a return recommendation list for the individual store, and displays them on a display or a printer. It can be output by controlling and printing. Further, the individual store terminal 2 can place an order for a book with respect to the all-store management server 1 by operation of a store clerk at the individual store.
The individual store terminal 2 can also be configured by a FAX apparatus. In this case, the exhibition / order recommendation list and the return recommendation list transmitted from the all-store management server 1 via the telephone line are received and printed. In addition, the store clerk can transmit the order / return request described in the exhibition / order recommendation list and the return recommendation list to the all-store management server 1 as a FAX. The individual store terminal 2 periodically reads the codes of all the books in stock at the individual store, so that the number of books to be inventoried and the number of the books as the number of individual store inventory books are communicated to the all store management server 1 It can be delivered by means.
In addition, each store has a POS (Point not shown).
A sales management terminal such as an (Off Sale) terminal is arranged to manage the actual number of books sold in a single store and to be transmitted to the all-store management server 1 as a sales record.

  The agency store terminal 3 can utilize the ordering, returning, and display recommendation information of the individual store created by the all-store management server 1 for use in sending goods to the individual store and for return operations.

  The all-store management server 1 is a device that creates and outputs order recommendation information, return recommendation information, and exhibition recommendation information at individual stores. This all-store management server 1 is constituted by a computer, and as shown in FIG. 2, an all-store database (DB) 11, an agency store inventory database (DB) 12, a coefficient determination processing unit 13, an all-store sales prediction processing unit 14, The store sales prediction processing unit 15, the individual store display recommendation processing unit 16, the recommendation list creation processing unit 17, and each table shown in FIG.

The all store database 11 storing past sales results for a predetermined period is a database of past sales results of all stores under the management of the store management server 1. Here, the predetermined period may be a period in which sufficient data can be stored to determine each coefficient used in a prediction method such as an incremental method described later.
The past sales results are the sales results aggregated by the sales management terminals located at each individual store, and the code, number of books, date, and first shipment date of the delivered, sold, and returned books for each individual store. Are stored.
Returned data is data such as the code of the book, the number of books, the date of return.
Further, the information on each individual store includes information such as the individual store identification number, the individual store name, and the location thereof.
The book information includes information such as title, genre, release date, and author.

  The agency store inventory database 12 is a database of book inventory data at an agency store. The agency store inventory database 12 stores the number of books held as inventory at the agency for each book.

The coefficient determination processing unit 13 performs processing for calculating coefficients and parameters used for predicting the number of books sold at all stores, coefficients and parameters used for predicting the number of books sold at individual stores, and coefficients used for calculating the recommended number of books displayed at individual stores.
The coefficients and parameters used for predicting the number of books sold at all stores include NM coefficients, incremental coefficients, and hybrid ratio parameters.
The NM coefficient is a coefficient used for sales prediction based on the NM method. Here, the NM method is based on the sales performance of books belonging to the genre of the book to be predicted, and the total sales volume p from the first edition sales date of the book to be predicted to the first reference date (Nth day), The NM coefficient (t) is calculated from the ratio (q / p) to the total number of sales volumes q up to the Mth day of the forecast date arbitrarily determined after the first reference date, and the total up to the first reference date of the book This is a method of predicting the total number of sales volumes of the book based on the number of sales records and the calculated NM coefficient (t).
For example, referring to the NM coefficient table in FIG. 3A, when it is desired to predict the total sales volume for one week based on the total sales volume of the first day, the first reference date N is “1 day” and M is “8 days. " Similarly, when it is desired to make a prediction based on the total number of sales records on the second day, the first reference date N is “2 days” and M is “9 days”.
As for the sales prediction method using the NM coefficient, the prediction method described in Japanese Patent Application No. 2006-201810 filed earlier by the present applicant can be applied. The contents described are included in the contents of this application.

The incremental coefficient is a coefficient used for a prediction method based on an incremental method. Based on the past sales performance of the genre to which the book to be predicted belongs, the total sales forecast volume for the next predetermined period and the total sales performance volume for that period using the total sales volume for the past three predetermined periods This is a method for predicting the total number of books sold based on an incremental coefficient that minimizes the error.
The hybrid ratio parameter is a parameter for weighting each prediction when the sales prediction based on the NM method and the sales prediction based on the incremental method are used in combination. Since the accuracy of the prediction method varies with the passage of time and the characteristics of the book, a highly accurate prediction value is obtained by weighting both prediction values with the hybrid ratio parameter. The hybrid ratio parameter can be set in advance by the store clerk and / or manager of the agency based on the past performance, or can be obtained by calculation using an optimization algorithm such as a hill-climbing method.
As coefficients and parameters used for predicting the number of books sold at individual stores, there are a store genre scale factor, a store title performance coefficient, and an individual store adjustment ratio parameter.
The store genre scale factor is the total of all the genres of each individual store with respect to the average of the total sales of all books of all the genres of all stores under the store management server 1 for each genre of past books for a predetermined period. This is a coefficient indicating the ratio of the total number of books sold in this book, and the strength (sales scale) of each individual store in that genre can be found.
The store title performance coefficient is a coefficient that represents the ratio of the total sales volume of individual stores to the total sales volume of all books of a given book. I understand.
The individual store adjustment ratio parameter is a parameter that weights each of the method of predicting using the store genre scale coefficient and the method of predicting using the store title performance coefficient for each book when forecasting the sales of the individual store. . The individual store adjustment ratio parameter can be set in advance by the store clerk and / or manager of the store based on the past results, or can be obtained by calculation using a predetermined algorithm such as the quasi-Newton method. it can.
In addition, as a coefficient used for calculating the display of a single store, there is a single store display coefficient. This is a coefficient for obtaining the recommended number of books to be displayed on the shelf of each book according to the size of the shelf of the individual store, and the number of books to be displayed can be set according to the shelf. This individual store display coefficient can be set by the administrator based on the shelf size and past performance, and can be calculated by a predetermined algorithm such as correlation analysis between the number of exhibited books and the number of sold books for each individual shelf. You can ask for it.

The all-store sales prediction processing unit 14 performs calculation processing for predicting the number of sales books in all stores for each book.
This processing is performed based on the NM method, the incremental method, and the hybrid ratio parameter.
Specifically, the sales prediction obtained by the NM method and the incremental method is weighted by a hybrid ratio parameter corresponding to the elapsed week from the first delivery date, and the store sales prediction is performed.

The individual store sales prediction processing unit 15 predicts sales for each individual store at regular intervals from the first delivery date to each individual store.
Specifically, the individual store sales prediction processing unit 15 obtains a store genre scale factor and a store title performance factor according to the elapsed time from the first delivery date, and responds to the elapsed time from the first delivery date. The individual store adjustment ratio parameter is used to weight both of them and to predict the store sales.

Here, as shown in FIG. 14, the recommendation for ordering / returning goods to a single store is based on the results of period N, and in order to optimize the exhibition in period N + 2, from period N to period N + 2 in period N + 1. Considering changes in sales trends over time, recommending orders and returns.
Furthermore, in order to make the exhibition work at the store of period N + 2 more efficient, display recommendation is performed.

The individual store recommendation processing unit 16 performs a process of calculating the recommended number of exhibits in the period N + 2 at the individual store.
Specifically, the individual store display recommendation processing unit 16 calculates the recommended number of exhibition books for the period N + 2 for each recommended book based on the display coefficient on the shelf of the target individual store.

The recommendation list creation processing unit 17 calculates the recommended number of books ordered in the period N + 1 and the recommended number of returned goods in the period N + 1 from the recommended number of books displayed at the store in the period N + 2 and the number of individual stores in the period N + 1, and recommends the display and order in the period N + 1 A list and recommended return list for period N + 1 are created and transmitted to each individual store terminal 2.
In this process, the recommended list creation processing unit 17 determines the number of recommended orders for the period N + 2 and the recommended number of exhibits for the period N + 2 for each individual store based on the stock status at the agency and the individual store and the recommended number of orders for the period N + 2. Process to calculate.
The recommended list creation processing unit 17 creates a return recommendation list for each individual store for the period N + 1. This process is recommended to return books for which the recommended number of books for exhibition in period N + 2 is zero and for books in which the recommended number of books ordered for period N + 2 is negative (books that are excessively stocked at the individual store) for period N + 1. Process to output as a list.

Next, an example of a processing method based on this embodiment will be described.
In this embodiment, as shown in FIG. 2, coefficients and parameters used for all store sales forecast, coefficients and parameters used for forecasting the number of individual store sales, and coefficients used for calculating the number of individual store exhibitions are calculated (S1). .
Then, an all-store sales prediction process is performed based on the calculated coefficients and parameters (S2), an individual store sales prediction process (S3), and an individual store display volume calculation process (S4). Then, when these calculations are completed, a recommended list creation process for ordering, display, and return at each store is performed, and this is transmitted to the individual store terminal 2 (S5), thereby completing a series of processing.

First, coefficient calculation processing as preprocessing will be described with reference to FIG.
In FIG. 3A, the coefficient determination processing unit 13 calculates NM coefficients for each genre. The NM coefficient calculation process is performed by the method described in Japanese Patent Application No. 2006-201810.
First, coefficients and parameters are set or input, and for each sales performance in the time series of the book population of the selected predetermined genre, the first from the first edition release date entered for each book of the genre. The total sales volume p up to the reference date (N) and the total sales volume q up to the forecast date (M) of the book are acquired.
Then, based on the sales performance of the books belonging to the genre of the book to be predicted, from the first reference date for the total number of books to be sold p from the first edition sales date of the book to be predicted to the first reference date (Nth day) The NM coefficient (t) is calculated from the ratio (q / p) to the total number of books to be sold q up to M days after the arbitrarily determined forecast date. Further, based on the distribution of the ratio (q / p), the NM coefficient (t) is calculated together with its accuracy (for example, frequency).
Then, based on the selected first reference date (N), the total number of sales volumes of the book to be predicted up to the first reference date (N), and the calculated NM coefficient (t), the predicted date of the book The process of calculating the predicted value of the total number of sales volumes (M) is performed.

In the incremental coefficient calculation process, as shown in FIG. 4A, first, the coefficient determination processing unit 13 refers to the all-store database 11 and calculates the total number of sales records for each book of the target genre (S101).
Thereby, as shown in FIG.4 (b), the total sales performance book volume according to the elapsed week after sale is calculated for every book. For example, as shown in FIG. 4B, for the book of title 1, the total number of sales records is calculated as 255 books in the first week, 542 books in the second week, and 902 books in the third week. .

Next, the coefficient determination processing unit 13 calculates an incremental coefficient (k, l, m) that minimizes the error between the prediction and the actual result by the quasi-Newton method (S102), and ends the process.
This calculation is performed by the following formula.
_{Incre (N + 1) = k} · R (N) + l · R (N-1) + m · R (N-2)
Here, Incre _{(N + 1)} represents the total number of predicted sales for the period N + 1, and R _(N) represents the total number of actual sales for the period N.
minΣ (| Incre _{(N + 1)} -R _{(N + 1)} | / | R _{(N + 1)} -R _(N) |)
Here, | Incre _{(N + 1)} −R _{(N + 1)} | represents the total sales prediction error in the period N + 1.
| R _{(N + 1)} −R _(N) | represents the number of sales records in the period N + 1.
Σ indicates the total of each book included in the genre.
Then, the incremental coefficient (k, l, m) that minimizes the error between the prediction and the actual result is stored in the incremental coefficient table of FIG.

  For the hybrid ratio parameter, the past sales performance is analyzed, and the hybrid ratio parameter value of the predicted value by the NM method and the predicted value by the incremental method is set in the hybrid ratio parameter table. In the example of FIG. 3C, since the weight of prediction by the NM method (NM) is 1.00 and the weight of the incremental method (Incre) is 0.00 in the first week and the second week, It becomes the prediction. On the other hand, after 3 weeks, since the weight of the NM method is 0.00 and the weight of the incremental method (Incre) is 1.00, the prediction is based on the incremental method.

  The calculation of the store genre scale coefficient in FIG. 3D is performed for each book genre: (total number of books sold in the genre for each individual store) / (sale of books in the genre of all stores under the store management server 1) The average of the average number of books per store is calculated, and this is stored as the store genre scale factor.

  The individual store adjustment ratio parameter is determined by analyzing past sales results.

Next, with reference to FIG. 5, a detailed process of the store sales forecast process will be described.
In FIG. 5, the all-store sales prediction processing unit 14 refers to the all-store database 11 and calculates an elapsed time after the book is released (S201). For example, when the prediction is performed by the incremental method on a weekly basis, the elapsed time from the sales date stored in the all-store database 11 is expressed in weekly units, and 18/7 when 18 days have passed since the sales date. = 2.5, so the third week from the release date is specified.
Further, the all-store sales prediction processing unit 14 refers to the all-store database and acquires the sales results of the individual stores (S202).
Note that the frequency of updating is not limited to a week.

The store sales prediction processing unit 14 refers to the NM coefficient table of FIG. 3A and acquires the NM coefficient of the genre corresponding to the elapsed time from the release date (S203).
For example, for a single store that places an order once a week, a factor for predicting sales for two weeks is required in consideration of the distribution time from the order from the individual store to the agency store to the delivery. In the example of FIG. 3 (a), a book of A genre that has passed 2 days after release, and if it is the first week from the release date (N = 2, M = 9 days), NM coefficient “9.2”, release Each coefficient “11.2” for the second week (N = 2 days, M = 16 days) from the date is acquired.
The all-store sales prediction processing unit 14 calculates the total sales forecast book volume based on the acquired NM coefficient (S204). This calculation process can be calculated by the total store total sales forecast NM _{(N + 1)} = tNM _(N) .
Here, t is the NM coefficient, and NM _(N) is the total number of sales records in period N.

Further, the all-store sales prediction processing unit 14 refers to the incremental coefficient table of FIG. 3B determined by the pre-processing, and acquires an incremental coefficient corresponding to the genre of the book (S205).
Then, the all-store sales prediction processing unit 14 performs calculation processing of the total sales forecast book volume by the incremental method (S206). This calculation process can be performed by Incre _{(N + 1)} = k · R _(N) + l · R _(N−1) + m · R _(N−2) .

When the total sales prediction by the NM method and the total sales prediction by the incremental method are completed, the store sales prediction processing unit 14 refers to the hybrid ratio parameter table in FIG. 3 (c) to determine the hybrid ratio corresponding to the elapsed time from the release date. A parameter is acquired (S207).
The all-store sales prediction processing unit 14 calculates the number of all-store sales forecasts based on the acquired hybrid ratio parameter (S208). In this process, the all-store sales prediction processing unit 14 sets the total store sales forecast volume to X _{(N + 1)} = a. (Incre _{(N + 1)} -Incre _(N) ) + (1-a). (NM _{(N + 1)} -NM _{( N)} Calculate according to).
Here, a is a hybrid ratio parameter, Incre _{(N + 1)} is the total number of predicted total sales of all stores based on the incremental method in the period N + 1, and NM _{(N + 1)} is the total number of predicted total sales of all stores based on the NM method in the period N + 1. In FIG. 3C, if it is not long after the release, the hybrid ratio parameter a = 0 and the predicted number of all-store sales is calculated based only on the NM method, and the value of a becomes 1 as the number of days elapses from the release date. By approaching, prediction based on the incremental method increases weight.
At the same time, the all-store sales prediction processing unit 14 also calculates the predicted number of all-store sales for the period N + 1 and the period N + 2, stores this in the all-store database 11 and ends the processing (S208).
As an example of the display of all store predictions, as shown in FIG. 6, the previous week ranking, ISBN code, book title, middle classification, publisher, first edition date, number of days after release, current week, number of all store sales in all stores (N Week, N-1 week, N-2 week), and the predicted number of all stores sold (N + 1 week, N + 2 week).

Next, a calculation method for sales prediction of a single store will be described with reference to FIG.
In FIG. 7, the individual store sales prediction processing unit 15 first refers to the entire store database 11 to identify the first delivery date of the book to the individual store, calculates the elapsed time from the first delivery date, A store adjustment ratio parameter is acquired with reference to the store adjustment ratio parameter table of FIG. 3E (S301, S302).
The elapsed time from the first delivery date represents, for example, the elapsed time from the first delivery date stored in the all-store database 11 in week units when the prediction is made on a weekly basis. Since 18/7 = 2.5 when 18 days have passed since the date, the third week is specified from the first delivery date.

  The individual store sales prediction processing unit 15 refers to the store genre scale coefficient table of FIG. 3D, and acquires a store genre scale coefficient corresponding to the elapsed time from the first delivery date (S303).

The individual store sales prediction processing unit 15 refers to the all store database 11 to acquire the number of individual store sales results and the number of all store sales results of a specific book after individual store delivery, and calculates a store title performance coefficient ( S304, S305).
In calculating the store title performance coefficient, the individual store sales prediction processing unit 15 calculates the store title performance coefficient based on the total number of individual store sales records / total number of total sales records of the book.

Then, the individual store sales prediction processing unit 15 performs calculation processing of the number of predicted sales of the individual store, stores the calculated number of predicted sales of the individual store in the memory (S306), and ends the sales prediction processing of the individual store. .
This process is calculated by the individual store sales prediction processing unit 15 based on the following formula.
Estimated number of individual store sales Y _{(N + 1)} = [b, store genre scale factor + (1-b), store title performance factor] X _{(N + 1)}
Here, X _{(N + 1)} is the number of predicted sales of the book at all stores in the period N + 1 predicted by the above-described processing, and b is the individual store adjustment ratio parameter.
At this time, the individual store sales prediction processing unit 15 performs not only the period N + 1 but also the period N + 2 in the same way.
As an example of calculation of this individual store prediction, as shown in FIG. 8, individual store prediction ranking, all store prediction ranking, N-1 week ranking, ISBN code, title, middle classification, publisher, publication date, actual sale date, The predicted sales volume for the period N + 1 and the period N + 2 is displayed.

Next, the display volume calculation process of the individual store will be described with reference to FIG.
In addition, when this personal store display book number calculation processing is performed, the personal store display coefficient table in FIG. 3F and the sales forecast book number table in FIG. In the individual store display coefficient table, the shelf priority for displaying books, the individual store display coefficient for each shelf (coefficient c, coefficient d), the maximum number of books displayed for each shelf, and the maximum number of display titles are displayed. It is remembered.
The priority order of the shelves can be determined in advance in the order in which the customers of the individual store are easily noticeable.
The maximum number of books to be displayed and the maximum number of titles may be determined, for example, depending on the type of shelf, the number of titles is 1 for flatbeds, the number of books is not limited, and the number of titles is limited for the insertion shelf.

In FIG. 9, the individual store display recommendation processing unit 16 acquires the number of predicted sales of the individual store calculated by the individual store sales prediction processing unit 15 from the memory, and sorts the books in descending order of the predicted number of sales at the individual store. (S401).
As a result, as shown in FIG. 3G, the memory stores the number order of books representing the order in which the number of predicted sales volumes is large, the title, and the number of predicted sales volumes as a sales predicted volume table.

The individual store display recommendation processing unit 16 assigns the books to the shelves with the highest use priority in the individual store display coefficient table in descending order of the predicted sales number from the sales predicted number table in FIG. 3G (S402).
As a result, books with a large number of predicted sales volumes are allocated in order to the shelves with the highest use priority.

The individual store display recommendation processing unit 16 refers to the individual store display coefficient table, acquires the individual store display coefficient of the assigned shelf, and uses the acquired individual store display coefficient to recommend the number Z of recommended exhibitions of the N + 1 week individual store. _{(N + 1) is obtained} by the following equation (S403).
Z _{(N + 1)} = c · Y _{(N + 1)} + d
Here, Y _{(N + 1)} is the predicted number of individual store sales in N + 1 week, c is a parameter that defines how many books to display on the shelf based on the predicted number of books determined in consideration of the situation of physical distribution, d represents the minimum number of books required for display on the shelf.

The private store display recommendation processing unit 16 stores the number of recommended display books of the books assigned to the shelves (S404).
The individual store display recommendation processing unit 16 determines whether or not the recommended number of books to display exceeds the upper limit of one of the two restrictions of the maximum number of display books or the maximum number of display titles in the individual store display coefficient table. (S405)

When it is determined that any one of the upper limits is exceeded, the individual store display recommendation processing unit 16 assigns the remaining recommended number of exhibitions of the same book to the shelf of the next priority of use. (S406)
In addition, if any restriction condition is not satisfied even after a book with a shelf is allocated, the individual store display recommendation processing unit 16 refers to the sales forecast book volume table, and selects a book with the next highest book volume forecast. Allocate (S407).
Then, the individual store display recommendation processing unit 16 determines whether or not the allocation of all the shelves is completed (S408), and if not completed, returns to the process of S402 described above.
If it is determined that all the shelves have been allocated, the process ends.

Next, the recommended list creation process will be described with reference to FIG.
In FIG. 10, the recommended list creation processing unit 17 refers to the agency store inventory database 12 to obtain the inventory results at the agency store, and refers to the entire store database 1 to obtain the inventory results at each individual store, Based on the actual shipment results, inventory return results, sales results, and inventory results after the inventory, the number of logical inventory books of the individual store in the period N is calculated (S501).
Specifically, the number of books shipped at each individual store is added to the number of delivered books per book, and the number of books returned at each book and the number of books sold are calculated to calculate the number of books at each individual store. You can calculate the number of logical inventory books for each.

  The recommended list creation processing unit 17 acquires the value of the predicted number of individual store sales of the target book predicted by the individual store sales prediction processing unit 15 stored in the memory, and stores the inventory as of the counting date of the period N + 1. Prediction is performed (S502). This is because the recommended list creation processing unit 17 subtracts the estimated sales volume of the individual store in the period N + 1 from the estimated inventory number of the individual store in the period N calculated in S501, thereby obtaining the inventory on the aggregation date in the period N + 1. You can calculate the expected number of books.

  Further, the recommended list creation processing unit 17 calculates the recommended number of orders for the period N + 2 (S503). In this process, the recommended list creation processing unit 17 calculates the period N + 1 calculated in the above-described process from the number of recommended exhibitions of the individual store in the period N + 2 calculated and stored in the memory by the individual store recommendation processing unit 16. Calculate by subtracting the number of daily inventory forecasts.

  The recommendation list creation processing unit 17 determines whether or not the calculated recommended number of books for the period N + 2 is negative (S504). Here, the case where the recommended number of books to be ordered is negative refers to the case where there is no recommended number of books to order or the recommended number of books to be ordered is negative (in the case of excessive inventory at a single store).

  As a result of the determination, if the recommended number of books to order is negative, the process proceeds to the recommended return list creation process of FIG.

If the recommended number of books ordered is not negative as a result of the determination, the recommendation list creation processing unit 17 performs a calculation process of the number of remaining orders received (S505).
This process can be calculated by the recommended list creation processing unit 17 subtracting the number of books to be delivered from the number of books ordered in the past with reference to the all-store database 11.

  In this state, the recommended list creation processing unit 17 refers to the agency store inventory DB 12 and acquires whether or not the agency store inventory exists for the book (S506).

  Further, as a result of the determination, for the book that is the target of the recommended order list, the information of the book and the recommended number of ordered books are temporarily stored in the memory for each individual store, and the recommended list creation processing unit 17 is stored in the memory. These books are sorted and listed in order of the recommended number of exhibited books (S507).

The recommended list creation processing unit 17 creates an exhibition / order recommendation list in which information on books stored in the memory, the recommended number of books to be ordered, and exhibition recommendation information are written. The data is transmitted to the terminal 2 (S508), and the process ends.
As this exhibition / order recommendation list, as shown in the example of FIG. 11, the individual store prediction order based on the sales prediction at each individual store, the all store prediction order based on the all store prediction, the order order considering the inventory at the agency store, ISBN code and title of the book, publisher name, recommended number of exhibitions, N-2 and N-1 weekly store sales forecasts, N-2 and N-1 weekly store sales volume, N weeks and N + 1 Weekly store sales forecast number, inventory forecast book number (N week actual stock volume and N + 1 week stock forecast book number), recommended order volume, remaining order volume, inventory status at the store, shelf information recommending shelf to display Is done.
As a result, the store clerk at the individual store can display and order the book with reference to the display / order recommendation list.
Further, the order recommendation list may be displayed on the screen and clicked on the screen, or the number of ordered books may be input, so that the order processing can be performed directly to the whole store management server 1 through the individual store terminal 2. .

Next, the return recommendation list creation process will be described with reference to FIG.
In FIG. 12, when the recommended order volume is not positive in the process of S503 of FIG. 10, the recommended list creation processing unit 17 displays the book for the period N + 2 calculated in the process of FIG. It is determined whether or not the recommended number of books is “0” (S601).

As a result of the determination, if the recommended number of exhibits is “0”, the recommendation list creation processing unit 17 describes the return as a complete return (S602).
As a result, it can be recognized that it is recommended that the individual stores return all items including the individual store inventory.

As a result of the determination, if the recommended number of exhibits is not “0”, that is, if the recommended number of exhibits is also “1”, it is listed in the return list as a partial return (S603).
Thereby, it can be recognized that the individual store is recommended to return the stock exceeding the recommended number of exhibits.

  Then, the recommended list creation processing unit 17 sorts the absolute value of the recommended number of books to be ordered as the recommended number of returned books in the descending order of the recommended number of returned books, and outputs this to the individual store terminal 2 as a recommended return list (S604).

As shown in the example of FIG. 13, the return recommendation list includes information indicating the ranking of N + 1 week, ISBN code of book, title, author name, publisher, classification, number of returned books, total return or partial return. Describe.
As a result, the store clerk can select the book to be returned with reference to the return recommendation list.

The figure which showed the whole structure of the individual store management system to which the sales prediction system concerning this invention is applied. The functional block diagram of the all the store management server concerning this embodiment. The figure which showed an example of the various tables concerning this embodiment. (A) The processing flow which showed the calculation process of the incremental coefficient concerning this embodiment. (B) A table representing the total number of sales records in accordance with the elapsed week after release. The processing flow which showed the store sales prediction process concerning this embodiment. The figure which showed the example of the output list of all the store sales prediction concerning this embodiment. The processing flow of the store sales forecast concerning this embodiment. The figure which showed the example of the output list of the store sales forecast concerning this embodiment. The processing flow for calculating the recommended number of individual store exhibits according to the present embodiment. A processing flow for creating an exhibition / order recommendation list according to the present embodiment. The figure which showed the example of the display and order recommendation list concerning this embodiment. The processing flow of the return recommendation list creation concerning this embodiment. The figure which showed the example of the return recommendation list | wrist concerning this embodiment. The figure showing the time table of the prediction process concerning this embodiment.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Store management server 2 Store terminal 3 Store store terminal 11 Store store database 12 Store store stock database 13 Coefficient determination processing part 14 Store sales forecast processing part 15 Store sales forecast processing part 16 Store display recommendation processing part 17 Recommendation list creation process Part

Claims (8)

A system for forecasting the number of books sold for each individual store,
Sales performance storage means for storing sales performance collected from sales management terminals of all individual stores,
Stored store sales forecast number storage means for storing the book sales forecast number of books in all stores, calculated according to the elapsed time since release for each book,
Store genre scale factor that stores the store genre scale factor that represents the ratio between the average number of sales records of all stores classified into each genre and the number of sales records of predicted individual stores for each genre. Storage means;
A hybrid ratio that stores a hybrid ratio parameter that is a weight for each of a forecast value based on a sales forecast method based on the NM method and a forecast value based on a sales forecast method based on an incremental method according to the elapsed time since the book was released Parameter storage means;
Referring to the sales record storage means, based on the past sales record of the genre to which the book to be predicted belongs, the total number of sales records from the first edition sales date to the first reference date (Nth day) (p) NM coefficient (t) is calculated from the ratio (q / p) of the total number of sales records (q) up to an arbitrarily determined forecast date (M day) after the first reference date (N day) The total number of books sold for the forecasted book based on the total number of books sold from the first edition release date to the first reference date (Nth day) and the calculated NM coefficient (t) for the forecasted book Sales forecasting method by NM method predicted by
With reference to the above sales record storage means, based on the past sales record of the genre to which the book to be predicted belongs, the number of predicted sales volumes in the next predetermined period using the past three sales record numbers in the predetermined period and its A sales forecasting method based on an incremental method that forecasts sales based on an incremental coefficient that minimizes the error from the number of books sold during the period;
Based on each sales forecasting method, a predicted value of the number of sales volumes is calculated, and the calculated predicted value is weighted with the hybrid ratio parameter with reference to the hybrid ratio parameter storage means, A store sales forecast processing means for calculating the sales forecast book number of the book and storing the calculated sales forecast book number in the store sales forecast book number storage means.
Referring to the above sales record storage means, for each book, the store title representing the sales scale at the individual store for the book, based on the ratio of the total number of sold records at the individual store to the total number of sold records at all stores. Store title performance coefficient calculation means for calculating the performance coefficient,
Based on the estimated sales number of books in all stores stored in the all-store sales predicted book number storage means, based on the calculated store title performance coefficient or at least one of the store genre scale coefficients, Individual store sales forecast processing means for calculating the number of predicted sales at the individual store,
A sales forecasting system characterized by comprising: An individual representing a weight using an elapsed time from the first delivery date when the book was delivered to the individual store for the first time, and a store genre scale factor and a store title performance factor at each individual store according to the elapsed time. It further has a store adjustment ratio parameter storage means for storing the store adjustment ratio parameter,
The individual store sales prediction processing means calculates the elapsed time from the first delivery date of the book for each individual store, refers to the individual store adjustment ratio parameter storage means, and corresponds to the calculated elapsed time. Obtain a store adjustment ratio parameter, and based on the acquired store adjustment ratio parameter, weight the store genre scale factor and store title performance coefficient to calculate the number of predicted sales at the store.
Sales prediction system according to the claim 1, characterized in that. A store-exhibition coefficient table for storing the priority of the shelves of the above-mentioned store, the display coefficient for each shelf, the maximum number of exhibition books, and the maximum number of display titles
The shelf with the higher priority of the individual store display coefficient table is allocated in descending order of the predicted sales volume of each book, and the book is determined according to the individual store display coefficient of the allocated shelf with reference to the individual store display coefficient table. The display recommended processing means for determining the recommended number of books and shelves for each book by calculating the recommended number of exhibition books and creating an exhibition recommended list.
The sales prediction system according to claim 1 or 2, wherein the system is a sales prediction system. In the above sales performance storage means, in addition to the number of sales records for each book in the individual store, the number of inventory books, the number of products shipped, the number of returned books is stored,
When the latest total period is N, add the number of deliverables for each book from the inventory point to the period N, and subtract the number of returned and sold books for each book. Then, the logical inventory up to the period N is obtained, and further, the estimated number of books for each book at each store in the period N + 1 is calculated by subtracting the number of sales forecasted books in the period N + 1. The recommended number of orders for each store is calculated by subtracting the above-mentioned number of inventory forecasted books for the next period N + 1 from the recommended number of books displayed at the next store N + 2, and the recommended order list is based on the calculated recommended number of books ordered. It further has an output means for creating and outputting,
The sales prediction system according to claim 3, wherein: It further has inventory data storage means for storing the inventory data of the book at the agency store,
The output means refers to the sales record storage means, and has received an order remaining number of books that have been ordered from the individual store and have not been sent to the individual store,
With reference to the inventory data storage means, the book recommended for ordering is output in addition to the order recommendation list indicating stock status information indicating whether or not the book is in stock at the agency.
The sales prediction system according to claim 4, wherein: The output means creates and outputs a return recommendation list from a book in which the recommended number of exhibited books calculated by the recommended exhibition processing means is zero and a book in which the recommended number of ordered books is negative.
The sales prediction system according to claim 4, wherein: Sales records storage means for storing sales results collected from the sales management terminals of all individual stores, and all stores that store the estimated number of books sold at all stores, calculated according to the elapsed time since release for each book The store genre scale coefficient that represents the ratio of the average number of sales records for all stores classified into each genre and the number of sales records for individual stores to be predicted for each book genre for each store. The store genre scale coefficient storage means stored in the store, and the forecast value based on the sales forecast method based on the NM method and the forecast value based on the sales forecast method based on the incremental method according to the elapsed time since the book was released A computer having a hybrid ratio parameter storage means for storing a hybrid ratio parameter which is weighting predicts the number of books sold for each book for each individual store. A method for performing a physical,
The computer
Referring to the sales record storage means, based on the past sales record of the genre to which the book to be predicted belongs, the total number of sales records from the first edition sales date to the first reference date (Nth day) (p) NM coefficient (t) is calculated from the ratio (q / p) of the total number of sales records (q) up to an arbitrarily determined forecast date (M day) after the first reference date (N day) The total number of books sold for the forecasted book based on the total number of books sold from the first edition release date to the first reference date (Nth day) and the calculated NM coefficient (t) for the forecasted book Sales forecasting method by NM method predicted by
With reference to the above sales record storage means, based on the past sales record of the genre to which the book to be predicted belongs, the number of predicted sales volumes in the next predetermined period using the past three sales record numbers in the predetermined period and its A sales forecasting method based on an incremental method that forecasts sales based on an incremental coefficient that minimizes the error from the number of books sold during the period;
Based on each sales forecasting method, a predicted value of the number of sales volumes is calculated, and the calculated predicted value is weighted with the hybrid ratio parameter with reference to the hybrid ratio parameter storage means, A store sales forecasting process for calculating the sales forecast book number of the book, and storing the calculated sales forecast book number in the store sales forecast book number storage means,
Referring to the above sales record storage means, for each book, the store title representing the sales scale at the individual store for the book, based on the ratio of the total number of sold records at the individual store to the total number of sold records at all stores. Store title performance coefficient calculation processing for calculating the performance coefficient,
Based on the estimated sales number of books in all stores stored in the all-store sales predicted book number storage means, based on the calculated store title performance coefficient or at least one of the store genre scale coefficients, Individual store sales forecast processing to calculate the number of sales forecast book at the above-mentioned store,
A sales forecasting method characterized by: Sales records storage means for storing sales results collected from the sales management terminals of all individual stores, and all stores that store the estimated number of books sold at all stores, calculated according to the elapsed time since release for each book The store genre scale coefficient that represents the ratio of the average number of sales records for all stores classified into each genre and the number of sales records for individual stores to be predicted for each book genre for each store. The store genre scale coefficient storage means stored in the store, and the forecast value based on the sales forecast method based on the NM method and the forecast value based on the sales forecast method based on the incremental method according to the elapsed time since the book was released A program for functioning a computer having hybrid ratio parameter storage means for storing weight ratio hybrid ratio parameters as a sales forecasting device. There is,
For the above computer
Referring to the sales record storage means, based on the past sales record of the genre to which the book to be predicted belongs, the total number of sales records from the first edition sales date to the first reference date (Nth day) (p) NM coefficient (t) is calculated from the ratio (q / p) of the total number of sales records (q) up to an arbitrarily determined forecast date (M day) after the first reference date (N day) The total number of books sold for the forecasted book based on the total number of books sold from the first edition release date to the first reference date (Nth day) and the calculated NM coefficient (t) for the forecasted book Sales forecasting method by NM method predicted by
With reference to the above sales record storage means, based on the past sales record of the genre to which the book to be predicted belongs, the number of predicted sales volumes in the next predetermined period using the past three sales record numbers in the predetermined period and its A sales forecasting method based on an incremental method that forecasts sales based on an incremental coefficient that minimizes the error from the number of books sold during the period;
Based on each sales forecasting method, a predicted value of the number of sales volumes is calculated, and the calculated predicted value is weighted with the hybrid ratio parameter with reference to the hybrid ratio parameter storage means, A store sales forecasting process for calculating the sales forecast book number of the book, and storing the calculated sales forecast book number in the store sales forecast book number storage means,
Referring to the above sales record storage means, for each book, the store title representing the sales scale at the individual store for the book, based on the ratio of the total number of sold records at the individual store to the total number of sold records at all stores. Store title performance coefficient calculation processing for calculating the performance coefficient,
Based on the estimated sales number of books in all stores stored in the all-store sales predicted book number storage means, based on the calculated store title performance coefficient or at least one of the store genre scale coefficients, Individual store sales forecast processing to calculate the number of sales forecast book at the above-mentioned store,
A computer program for executing