Management apparatus

Abstract

To support advertising activities using multiple advertising media operated by different entities, in a web advertisement.SOLUTION: A management apparatus 1 of web advertisement includes: a management unit 120 which manages an advertisement using an advertising medium, for each of a plurality of advertising media operated by different entities; a past performance acquisition unit 111 which acquires past performance related to each of the advertisements; a first machine learning unit 112 which executes machine learning related to prediction of correspondence between an advertisement distribution condition and past performance, for each of the plurality of advertising media; a relationship prediction unit 116 which predicts correspondence between the advertisement distribution condition and past performance for each of the plurality of advertising media, based on the machine learning; and an estimation unit 117 which estimates, based on the prediction, an appropriate advertisement medium and an appropriate advertisement distribution condition related to requirements and a goal.SELECTED DRAWING: Figure 1

Description

Technical Field

[0001] The present invention relates to a management device for web advertisements.

Background Art

[0002] As an advertising means, so-called web advertisements are widely used. In web advertisements, advertisements of advertisers are posted on web pages via various advertising media. In various advertising media that provide web advertisements, bidding is performed using a unit price per performance exemplified by the number of clicks, conversion, number of impressions, etc. Note that conversion refers to the number of actions defined by the advertiser as "results", exemplified by purchase of goods, application for samples, request for materials, inquiry, etc.

[0003] Such bidding is automatically performed based on specified advertisement delivery conditions. Advertisement delivery conditions in web advertisements include various conditions exemplified by the type of performance used for calculating the unit price (number of clicks, conversion, number of impressions, etc.), upper limit value and target value of the unit price, period during which the advertisement is performed, budget per certain period, keywords related to the advertisement, content of the advertisement, etc. However, web advertisements are distributed via various advertising media. Therefore, advertising agencies, advertisers, etc. may have difficulty setting appropriate advertisement delivery conditions according to the advertising media in order to obtain a desired advertising effect. For such reasons, in web advertisements, means for assisting advertising agencies, etc. to obtain a desired advertising effect are required.

[0004] Regarding means for assisting an advertiser in obtaining a desired advertising effect in web advertising, Patent Document 1 discloses an advertising frame selection system that selects a plurality of desired acquisition advertising frames from a plurality of advertising frames across a plurality of advertising media. An advertising frame database stores, for each of the advertising media, an acquisition cost for acquiring each of the advertising frames and data indicating a contact probability of an advertisement displayed in each of the advertising frames. Input means is provided for inputting an upper limit budget. For each of the advertising media, a first selection processing means performs one or more times a process of sequentially selecting the desired acquisition advertising frames from the plurality of advertising frames based on the contact probability, thereby deriving one or more sets of desired acquisition advertising frame groups such that the total of the acquisition costs is within a predetermined ratio of the upper limit budget. A second selection processing means derives a combination of the desired acquisition advertising frame groups such that the total of the acquisition costs is within the upper limit budget by selecting one desired acquisition advertising frame group for each of the advertising media. Output means outputs each of the desired acquisition advertising frames included in the derived combination of the desired acquisition advertising frame groups. The advertising frame selection system according to Patent Document 1 may enable efficient selection of a plurality of desired acquisition advertising frames from a plurality of advertising frames across a plurality of advertising media.

Prior Art Documents

Patent Documents

[0005]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0006] By the way, the advertising effect of web advertisements depends on the advertisement delivery conditions. If it is an advertising medium under one's own management, an advertising agency or the like can easily obtain sufficient data directly showing the correspondence between the advertisement delivery conditions and the advertising effect. However, it may be difficult to obtain such data from advertising media managed by others. In addition, the correspondence between the advertisement delivery conditions and the advertising effect changes daily. Therefore, it is not easy to obtain such data for each of a plurality of advertising media with different operating entities.

[0007] The technology of Patent Document 1 is only capable of efficiently selecting a plurality of desired acquisition advertising frames from advertising frames purchased at a given acquisition cost, which are a plurality of advertising frames across a plurality of advertising media. Therefore, the technology of Patent Document 1 has room for further improvement in terms of supporting advertising activities that make good use of a plurality of advertising media with different operating entities in today's web advertisements where the correspondence between the advertisement delivery conditions and the advertising effect changes daily.

[0008] The present invention has been made in view of such circumstances. An object of the present invention is to support advertising activities that make good use of a plurality of advertising media with different operating entities in web advertisements.

Means for Solving the Problem

[0009] As a result of intensive studies to solve the above problems, the present inventors have found that the above object can be achieved by making a prediction based on past performance for each of a plurality of advertising media with different operating entities and renewing the contract with an appropriate advertising media based on the prediction. And the present inventors have completed the present invention. Specifically, the present invention provides the following.

[0010] The present invention provides a management device for web advertisements, comprising: a management unit that manages advertisements for each of a plurality of advertising media with different operating entities; a performance acquisition unit that acquires the performance of each of the advertisements; a machine learning unit that performs machine learning related to prediction of the correspondence between advertising delivery conditions and performance for each of the plurality of advertising media; a requirement acquisition unit that acquires requirements for the advertising delivery conditions and a target for the performance of the advertisements; a relationship prediction unit that predicts the correspondence between the advertising delivery conditions and the performance for each of the plurality of advertising media based on the machine learning; and an estimation unit that estimates an appropriate advertising media, which is an advertising media suitable for satisfying the requirements and achieving the target among the plurality of advertising media, based on the prediction. The machine learning unit performs machine learning including the association between the advertising media, the advertising delivery conditions, and the performance in learning data, and the estimation unit further estimates appropriate advertising delivery conditions, which are advertising delivery conditions suitable for achieving the target in the appropriate advertising media.

[0011] When using different advertising media with different operating entities, at least one of the plurality of advertising media is considered not to be under the control of the person making the selection. The present invention centrally manages advertisements for each of a plurality of advertising media with different operating entities. As a result, users such as advertising agencies can use a plurality of advertising media with different operating entities without being aware of differences in the user interfaces of the advertising media.

[0012] By the way, those who use multiple advertising media with different operating entities need to predict the relationship between the advertising delivery conditions and the advertising performance for at least one advertising media operated by an operating entity different from themselves in order to conduct advertising suitable for achieving the goal. The present invention obtains the advertising performance for each of the multiple advertising media with different operating entities. Thereby, the present invention can accumulate data on the association between the advertising media, the advertising delivery conditions, and the performance for each of the multiple advertising media with different operating entities. Then, the present invention performs machine learning including the association between the advertising media, the advertising delivery conditions, and the performance in the learning data. By this machine learning, the present invention can predict the correspondence relationship between the advertising delivery conditions and the performance for each of the multiple advertising media.

[0013] In order to perform machine learning including the above-mentioned association in the learning data, the present invention can predict the correspondence relationship with higher accuracy than the method using various score calculation formulas based on the performance and the method using the performance itself. Since the learning data includes the association between the advertising media, the advertising delivery conditions, and the performance regarding multiple advertising media, the present invention can perform machine learning and prediction that contribute to appropriate comparison between multiple advertising media.

[0014] Based on the above-mentioned machine learning, the present invention predicts the correspondence relationship between the advertising delivery conditions and the performance for each of the multiple advertising media. Therefore, based on this prediction, the present invention can estimate an appropriate advertising media that is suitable for realizing the goal aimed by the advertiser among the multiple advertising media. And the present invention can provide the appropriate advertising media and the appropriate advertising delivery conditions in the appropriate advertising media. In addition, the present invention can provide the prediction result of the performance in the appropriate advertising media. Therefore, the present invention can provide these information to the user and support the advertising activities that use multiple advertising media with different operating entities in web advertising.

Effect of the Invention

[0015] The present invention can support advertising activities that use multiple advertising media with different operating entities in web advertising.

Brief Description of the Drawings

[0016]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Mode for Carrying Out the Invention

[0017] The following is a detailed description of an example of an embodiment of the present invention with reference to the drawings.

[0018] <System S> Figure 1 is a block diagram showing the hardware configuration and software configuration of the system S of the present embodiment. The following is an explanation using Figure 1 regarding an example of a preferred aspect of the hardware configuration and software configuration in the system S of the present embodiment. The system S includes a management device 1 for web advertisements and a terminal T configured to be communicable with each other via a network N.

[0019] 〔Advertising Medium〕 The advertising medium in this embodiment refers to an advertising platform that distributes web advertisements. Hereinafter, "web advertisement" is also simply referred to as "advertisement". The advertising medium of this embodiment includes, for example, two or more of Google Adsense (registered trademark), Facebook Ads (registered trademark), TikTok Ads (registered trademark), X Ads (registered trademark), etc. The advertising medium of this embodiment preferably includes an advertising medium that distributes web advertisements to social network services (SNS) exemplified by Facebook (registered trademark), TikTok (registered trademark), X (registered trademark), etc. Thereby, the management device 1 can predict the performance including engagement exemplified by the number of "likes" attached to posts on the SNS, the number of "shares", etc.

[0020] 〔Management Device 1〕 The management device 1 includes a control unit 11, a storage unit 13, a communication unit 14, etc. The type of the management device 1 is not particularly limited. Examples of the type include a server, a cloud server, etc. In order to reduce the burden on the user in terms of money, labor, etc., the management process executed by the management device 1 is preferably provided as a service (Software as a Service, SaaS) in a provision form in which necessary functions can be used as services only as much as necessary.

[0021] [Control Unit 11] The control unit 11 includes a CPU (Central Processing Unit), a RAM (Random Access Memory), a ROM (Read Only Memory), etc.

[0022] The control unit 11 cooperates with the storage unit 13 and / or the communication unit 14 as necessary. And the control unit 11 realizes the software components of the program of the present embodiment executed by the management device 1. The software components include a performance acquisition unit 111, a first machine learning unit 112, a feature discrimination unit 113, a second machine learning unit 114, a requirement acquisition unit 115, a relationship prediction unit 116, an estimation unit 117, a feature prediction unit 118, an advertisement generation unit 119, a management unit 120, etc. These software components will be described later together with the preferred flow of the management process executed by the management device 1 of the present embodiment.

[0023] [Storage Unit 13] The storage unit 13 is a device in which data and / or files are stored, and has a storage unit that non-temporarily stores data by means of a hard disk, a semiconductor memory, a recording medium, a memory card, etc.

[0024] The storage unit 13 may have a mechanism that enables connection to a storage device or a storage system such as a NAS (Network Attached Storage), a SAN (Storage Area Network), a cloud storage, a file server, and / or a distributed file system via the network N.

[0025] The storage unit 13 stores an advertisement database 131, a performance database 132, etc. In addition to what is shown in the figure, the storage unit 13 stores a program executed by the control unit 11, a large language model, a recognition model, a generation neural network, learning data related to the first machine learning unit 112, learning data related to the second machine learning unit 114, etc. The recognition model includes, for example, a large language model, a voice recognition model, an image recognition model, etc. The generation neural network includes, for example, a neural network related to a large language model, a voice generation neural network, an image generation neural network, etc.

[0026] (Advertisement Database 131) The advertising database 131 stores data on web advertisements using advertising media. The advertising media include a plurality of advertising media with different operating entities. The data includes at least information for identifying the advertising media and advertising delivery conditions. The advertising delivery conditions include, for example, types of performance used for calculating unit prices (number of clicks, conversions, number of impressions, etc.), upper limit values and target values of unit prices, periods during which the advertisements are conducted, budgets per certain period, overall budgets for the advertisements, keywords related to the advertisements, and various conditions exemplified by the content of the advertisements. The content of the advertisements includes, for example, advertisement texts, advertisement images, advertisement voices, advertisement videos, etc. For the sake of convenience regarding data search, acquisition, and storage, it is preferable that the advertisement data be stored in association with an advertisement ID for identifying the advertisement.

[0027] By storing data on web advertisements using a plurality of advertising media with different operating entities in the advertising database 131, the management device 1 can manage advertisement delivery using an appropriate advertising medium among the plurality of advertising media. By storing data including advertising delivery conditions in the advertising database 131, the management device 1 can manage the advertising delivery conditions for each of the plurality of advertising media. By storing data including the content of the advertisements in the advertising database 131, the management device 1 can manage the content of the advertisements for each of the plurality of advertising media.

[0028] Figure 2 shows an example of the advertisement database 131. In the example shown in Figure 2, as data for the first advertisement identified by the advertisement ID "C0001", the advertisement medium is "Search service △△", the advertisement delivery conditions are "Type of performance: Number of clicks, Upper limit of unit price: 20 yen or less per click, Advertising period: 20 days from △△ month △△ day to △△ month △△ day, Budget per day: within 10,000 yen, Keywords: Beauty salon, Perm, Coloring, Tokyo metropolitan area, Tama, Tachikawa, Hachioji", and the advertisement content is "Advertising text: If you want a haircut in the Tama area, then Hair Salon △△, Advertising image: logo0001.png". Also, in the example shown in Figure 2, as data for the second advertisement identified by the advertisement ID "C0002", the advertisement medium is "Real-name SNS △△", the advertisement delivery conditions are "Type of performance: Number of impressions, Upper limit of unit price: 200 yen or less per 1,000 impressions, Advertising period: 1 week from △△ month △△ day to △△ month △△ day, Budget per day: within 30,000 yen, Keywords: Regardless of gender, Single, 30s - 40s", and the advertisement content is "Advertising text: Raise debris to create the earth! Advertising video: 30-second video 0002.mp4".

[0029] By storing these data in the advertisement database 131, the management device 1 can manage the advertisement medium, advertisement delivery conditions, and advertisement content for the first advertisement and the second advertisement.

[0030] (Performance database 132) The performance database 132 stores data for identifying the advertisement medium and data indicating the performance of the advertisement associated with the advertisement delivery conditions. The performance of the advertisement includes, for example, one or more of various indicators of web advertisements such as the number of impressions (Imp), cost per mille (CPM), number of clicks (Click), cost per click (CPC), click-through rate (CTR), number of conversions (CV), cost per action (CPA), conversion rate (CVR), acquisition cost (Cost), etc.

[0031] Note that CPM is the cost of web advertisements per 1,000 display times in actual delivery. CPC is the cost of web advertisements per click in actual delivery. CTR is the ratio of the number of clicks to the number of display times. CV is the number of times a person who has viewed a web advertisement has taken a specific action exemplified by accessing the advertiser's website, making an inquiry regarding a product or service, purchasing a product or service, etc. CPA is the cost of web advertisements per conversion. CVR is the ratio of the number of conversions to the number of display times.

[0032] By storing data indicating these various achievements in the performance database 132 in association with data identifying the advertising medium and the advertisement delivery conditions, the management device 1 can execute machine learning that includes the association between the advertising medium, the advertisement delivery conditions, and the data indicating these various achievements in the learning data. Therefore, the management device 1 can predict the correspondence between these various achievements and the advertisement delivery conditions. Therefore, the management device 1 can estimate an appropriate advertising medium and appropriate advertisement delivery conditions suitable for achieving the goals of various achievements.

[0033] In addition, the performance of advertisements may include various indicators related to engagement in a social networking service (SNS). Engagement is, for example, the number of "likes" attached to a post on the SNS, the number of "shares", etc. Thereby, the management device 1 can estimate an appropriate advertising medium and appropriate advertisement delivery conditions suitable for achieving the goals of various achievements in the advertising medium that is an SNS.

[0034] It is preferable that the performance database 132 stores the history of performance before and after the change of the advertisement delivery conditions. Thereby, the management device 1 can perform machine learning that includes the history in the learning data. It is preferable that the performance database 132 stores the association between the prediction by the relationship prediction unit 116 described later and the performance corresponding to the prediction. Thereby, the management device 1 can perform machine learning based on the comparison between the prediction model and the performance model. The effects brought about by these machine learnings will be described later together with the explanation of the management process.

[0035] For convenience in searching, acquiring, and storing data, it is preferable that the performance data be stored in association with a performance ID that identifies the performance. Further, for performing machine learning based on the correspondence with the advertisements stored in the advertisement database 131, it is preferable that the performance data be stored in association with data indicating the correspondence.

[0036] FIG. 3 is an example of the performance database 132. In the example shown in FIG. 3, as data of the first performance identified by the performance ID "R0001", the advertising medium is "Search service △△", the corresponding advertisement is "C0001", the advertisement delivery conditions are "Type of performance: Number of clicks, upper limit of unit price: 20 yen or less per click, advertisement period: 10 days from △△ month △△ day to △△ month △△ day, budget per day: within 10,000 yen, keyword: beauty salon, perm, coloring, within Tokyo, Tama, Tachikawa, Hachioji", and the performance is "Imp: 100,000, CPM: 980, Click: 7,000, CPC: 14, CTR: 7%, CV: 2,000, CPA: 49, CVR: 2%, Cost: 98,000". Also, in the example shown in FIG. 3, as data of the second performance identified by the performance ID "R0002", the advertising medium is "Search service △△", the corresponding advertisement is "C0001", the advertisement delivery conditions are "Type of performance: Number of clicks, upper limit of unit price: 30 yen or less per click, advertisement period: 10 days from △△ month △△ day to △△ month △△ day, budget per day: within 10,000 yen, keyword: beauty salon, perm, coloring, within Tokyo, Tama, Tachikawa, Hachioji", and the performance is "Imp: 80,000, CPM: 1,125, Click: 4,000, CPC: 14, CTR: 5%, CV: 2,400, CPA: 37.5, CVR: 3%, Cost: 90,000".

[0037] By storing these data in the performance database 132, the management device 1 can execute machine learning that includes, in the learning data, the association between the data indicating the first performance, the second performance, and other performances, and the advertising medium and the advertising distribution conditions. Also, in the example shown in FIG. 3, the second performance is such that, 10 days after the start of the advertising distribution, the upper limit of the unit price of the first performance is changed from "20 yen or less per click" to "30 yen or less per click". Therefore, the first performance and the second performance constitute the history of the performance before and after the change in the advertising distribution conditions. The history of the performance has many parts that are common or similar to the advertising distribution conditions, which are explanatory variables in machine learning. Thereby, the management device 1 can perform machine learning such that the relationship prediction unit 116 can predict the correspondence between the advertising distribution conditions and the performance with higher accuracy.

[0038] (Large Language Model) The large language model of the present embodiment is not particularly limited as long as it is a language model that executes natural language processing for generating a response to an input prompt. The "language model" mentioned here is a type of probability model used in natural language processing, and is a model for probabilistically predicting how likely a given word or sentence is to occur as natural language. Specifically, the language model predicts the next sentence by calculating the occurrence probability of a given sentence or the like, comparing the occurrence probabilities of multiple sentences, etc. Thereby, the language model can automatically generate the most likely sentence based on the context related to the given sentence or the like.

[0039] The large language model of the present embodiment is preferably a large language model that is trained with at least 500 GB or more of a large amount of text, such as ChatGPT-3.5 of OpenAI (registered trademark), ChatGPT-4, and LLaMa of Meta AI (registered trademark), and has at least 10 billion or more parameters. These large amounts of text include the text of the advertisement, the text related to the relationship between the advertisement text and its purpose / effect, and the like. Therefore, with such a large language model, the management device 1 can generate an advertisement text having appropriate features, which are the features of the advertisement suitable for achieving the goal.

[0040] It is preferable that the large language model of this embodiment has been pre-trained with the learning data including the association between the advertisement text and the characteristics of the advertisement text. Thereby, the large language model of this embodiment can generate a text having the appropriate characteristics in the process of inputting the appropriate characteristics.

[0041] (Speech recognition model) The speech recognition model is a model used for the process of inputting speech and generating a text indicating the characteristics of the speech. The type of the speech recognition model of this embodiment is not particularly limited. The speech recognition model of this embodiment preferably includes a hidden Markov model in order to suppress the amount of calculation.

[0042] It is preferable that the speech recognition model of this embodiment has been pre-trained with the learning data including the association between the advertisement speech and the characteristics of the advertisement speech. Thereby, the speech recognition model of this embodiment can generate a text appropriately indicating the characteristics of the advertisement speech in the process of inputting the advertisement speech.

[0043] In addition, the speech recognition model of this embodiment preferably includes an acoustic model related to the recognition of the characteristics of the speech in order to perform speech recognition based on the characteristics of the speech.

[0044] (Image recognition model) The image recognition model is a model used for the process of inputting an image and generating a text indicating the characteristics of the image. The type of the image recognition model of this embodiment is not particularly limited. The type may be, for example, a model related to the k-nearest neighbor method, the k-means method, the support vector machine, the convolutional neural network (CNN), the recurrent neural network (RNN), etc.

[0045] It is preferable that the image recognition model of the present embodiment has been pre-trained with the learning data including the association between the advertisement image and the features of the advertisement image. Thereby, the image recognition model of the present embodiment can generate text that appropriately represents the features of the advertisement image in the process of inputting the advertisement image.

[0046] (Voice generation neural network) The voice generation neural network is a neural network used in the process of generating voice having the features given in the input. The type of the process is not particularly limited. The process may be, for example, a process related to a variational auto-encoder (VAE), generative adversarial networks (GANs), or the like.

[0047] It is preferable that the voice generation neural network of the present embodiment has been pre-trained with the learning data including the association between the advertisement voice and the features of the advertisement voice. Thereby, the voice generation neural network of the present embodiment can generate an advertisement voice having the appropriate features in the process of inputting the appropriate features.

[0048] (Image generation neural network) The image generation neural network is a neural network used in the process of generating an image having the features given in the input. The type of the process is not particularly limited. The process may be, for example, a process related to a variational auto-encoder (VAE), generative adversarial networks (GANs), a diffusion model, or the like.

[0049] Among them, the processing related to the image generation neural network is preferably the processing related to the diffusion model. The diffusion model can learn the latent structure of the data set by modeling the behavior of each point of the data diffusing on the latent space. Therefore, the diffusion model can generate an image that well reflects the latent structure related to the features given as input.

[0050] It is preferable that pre-training in which the correspondence between the advertisement image and the features of the advertisement image is included in the learning data is performed for the image generation neural network of the present embodiment. Thereby, the image generation neural network of the present embodiment can generate an advertisement image having the appropriate features in the process of inputting the appropriate features.

[0051] [Communication unit 14] Returning to FIG. 1. The communication unit 14 is not particularly limited as long as it connects the management device 1 to the network N and enables communication with the terminal T and the like. Examples of the communication unit 14 include a wireless device compatible with a mobile phone network, a device connectable to a wireless LAN, and a network card compatible with the Ethernet standard.

[0052] [Network N] The type of the network N is not particularly limited as long as it enables communication between the management device 1 and the terminal T and the like. Examples of the type of the network N include the Internet, a mobile phone network, a wireless LAN, and the like.

[0053] [Terminal T] The terminal T is, for example, a personal computer, a laptop computer, a smartphone, a tablet terminal, or the like. The terminal T can execute processing for displaying information provided from the management device 1, processing for providing the management device 1 with requirements required for advertisement distribution conditions and target advertisement performance, and the like.

[0054] [Main flowchart of management processing] FIG. 4 is a main flowchart showing an example of a preferred flow of management processing executed by the management device 1 of the present embodiment. FIG. 5 is a figure following FIG. 4. FIG. 6 is a figure following FIG. 5. FIG. 7 is a figure following FIG. 6. The following is an example of a preferred flow of management processing executed by the management device 1 of the present embodiment using FIGS. 4 to 7.

[0055] Prior to estimating the appropriate advertisement delivery conditions and the appropriate advertisement medium, the management device 1 performs machine learning related to the management processing. The machine learning includes machine learning related to predicting the correspondence between the advertisement delivery conditions and the results in each of the plurality of advertisement media. The machine learning related to predicting the correspondence between the advertisement delivery conditions and the results is executed by the first machine learning unit 112 or the like. Note that the first machine learning unit 112 is also simply referred to as the "machine learning unit". Steps S1 to S5 are an example of a series of processing flows related to machine learning.

[0056] [Step S1: Determine whether to perform machine learning] The control unit 11, in cooperation with the storage unit 13 and the communication unit 14, executes a process of determining whether to perform machine learning (Step S1, machine learning necessity determination step). If it is determined to perform, the control unit 11 moves the process to Step S2. If it is not determined to perform, the control unit 11 moves the process to Step S6.

[0057] To enable users to perform machine learning at will, the machine learning necessity determination step preferably includes a procedure for determining to perform machine learning when an instruction to execute machine learning is received from a terminal T or the like. Further, to enable machine learning according to the update of results, the machine learning necessity determination step preferably includes a procedure for determining to perform machine learning when data has been updated in the result database 132 since the previous machine learning. At this time, to suppress an increase in the processing load due to machine learning, the machine learning necessity determination step preferably includes a procedure for determining to perform machine learning when the data updated since the previous machine learning satisfies a given condition. The given condition includes, for example, a condition that the delivery of any advertisement has been completed, a condition that a certain period of time has elapsed, a condition that the amount of updated data exceeds a given amount, and the like.

[0058] [Step S2: Obtain advertisement results for a plurality of advertising media] The control unit 11 cooperates with the storage unit 13 and executes the result acquisition unit 111. Then, the control unit 11 executes, by the result acquisition unit 111, a process of obtaining advertisement results for a plurality of advertising media (step S2, result acquisition step). The control unit 11 stores the results in the result database 132 and transfers the process to step S3.

[0059] The result acquisition step preferably includes a procedure for obtaining data related to the correspondence between the advertisement delivery conditions and the results under the conditions from an external database or the like. The data obtained from an external database or the like is, for example, data provided by an advertising medium operated by an operation entity different from the administrator of the management device 1 via the API of the advertising medium, or market data collected in a database operated by an operation entity different from the administrator of the management device 1. Thereby, the management device 1 can perform highly accurate machine learning using, for example, its own data and market data. Therefore, the management device 1 can predict the results of web advertisements with higher accuracy. Therefore, the management device 1 can estimate advertising media and advertisement delivery conditions suitable for achieving the target with higher accuracy.

[0060] [Step S3: Execute Machine Learning for Predicting the Corresponding Relationship] The control unit 11 collaborates with the storage unit 13 and executes the first machine learning unit 112. Then, the control unit 11 executes a process of executing machine learning related to the prediction of the corresponding relationship by the first machine learning unit 112 (Step S3, the first machine learning step (machine learning step)). The control unit 11 transfers the process to Step S4.

[0061] The machine learning related to the first machine learning step (machine learning step) is not particularly limited as long as it predicts the corresponding relationship between the advertisement delivery conditions and the results for each of a plurality of advertisement media. The machine learning includes, for example, correlation rule learning, Support Vector Machine (SVM), Extreme Learning Machine (ELM), learning using the error backpropagation method (backpropagation) related to neural networks, and the like. The machine learning preferably includes, in particular, machine learning that uses the advertisement medium and the results as target variables and the advertisement delivery conditions as explanatory variables. Thereby, the management device 1 can predict the advertisement delivery conditions suitable for achieving the target by a process using machine learning that takes the advertisement medium and the target results as inputs.

[0062] In order to reduce the labor related to the prediction, the machine learning preferably includes unsupervised learning. Further, in order to further improve the accuracy of the prediction, the machine learning may include supervised learning using an external evaluation or the like related to the validity of the prediction.

[0063] When the performance database 132 includes the performance history before and after the change of the advertisement delivery conditions, the machine learning in the first machine learning step preferably executes machine learning that includes the history in the learning data.

[0064] The relationship prediction unit 116 described below predicts a correspondence relationship in which a plurality of conditions included in the advertisement delivery conditions are explanatory variables and the results including a plurality of types of indicators such as CPM and CPA are objective variables, based on machine learning in the first machine learning unit 112. In machine learning for predicting such a multi-parameter to multi-parameter correspondence relationship, the prediction accuracy can be improved by learning data including explanatory variables with many common or similar parts. This is because such learning data well represents the relationship between data that only differs in some of the explanatory variables.

[0065] By the way, when changing the advertisement delivery conditions, often only a part of the advertisement delivery conditions is changed. Such changes are, for example, changes such as changing only the budget amount or only the unit price. Therefore, it is expected that the performance history before and after the change of the advertisement delivery conditions includes learning data with many common or similar parts in the advertisement delivery conditions that are explanatory variables. Therefore, in the first machine learning step, by performing machine learning that includes the performance history before and after the change of the advertisement delivery conditions in the learning data, the relationship prediction unit 116 can predict the correspondence relationship between the advertisement delivery conditions and the performance with higher accuracy.

[0066] Also, the machine learning in the first machine learning step preferably executes machine learning that includes in the learning data the association between the prediction by the relationship prediction unit 116 described below and the performance corresponding to the prediction. Machine learning based on the comparison between the prediction model and the performance model is known to improve the accuracy of the prediction model. Therefore, in the first machine learning step, by performing machine learning that includes in the learning data the association between the prediction by the relationship prediction unit 116 and the performance corresponding to the prediction, the relationship prediction unit 116 can predict the correspondence relationship between the advertisement delivery conditions and the performance with higher accuracy.

[0067] In order to reflect the differences for each advertising medium, the machine learning in the first machine learning step preferably makes the machine learning model different for each advertising medium learn to predict the correspondence relationship between the advertisement delivery conditions and the performance.

[0068] A series of processes related to machine learning preferably includes a procedure of executing machine learning (second machine learning) related to prediction of appropriate features, which are features of advertisements suitable for achieving a goal. The machine learning (second machine learning) related to prediction of appropriate features is executed by the second machine learning unit 114 or the like. Steps S4 to S5 are an example related to the flow of a series of processes related to the second machine learning.

[0069] [Step S4: Discriminate features of advertisement] The control unit 11 cooperates with the storage unit 13 and executes the feature discrimination unit 113. Then, the control unit 11 executes, by the feature discrimination unit 113, a process of discriminating the features of the advertisements stored in the advertisement database 131 (step S4, feature discrimination step). The control unit 11 moves the process to step S5.

[0070] To discriminate the features of various data constituting the advertisement, the feature discrimination step preferably includes a procedure of discriminating the features of the advertisement by a process according to the various data constituting the advertisement. The process includes, for example, a procedure of discriminating the features of the voice constituting the advertisement by a voice recognition model. The process includes, for example, a procedure of discriminating the features of the image and / or video constituting the advertisement by an image recognition model. Also, in order to suppress the amount of input data related to the second machine learning step described later, the feature discrimination step preferably includes a procedure of summarizing the features of the advertisement by a process using a large language model. To reflect the features of data other than text in the summary, the process using the large language model preferably includes, in the input data, text indicating the features obtained by procedures related to voice, image, etc. and the advertisement text.

[0071] [Step S5: Execute machine learning related to prediction of appropriate features] The control unit 11 cooperates with the storage unit 13 and executes the second machine learning unit 114. Then, the control unit 11 executes, by the second machine learning unit 114, a process of executing machine learning related to prediction of appropriate features (step S5, second machine learning step). The control unit 11 moves the process to step S6.

[0072] The machine learning related to the second machine learning step is not particularly limited as long as it is related to predicting appropriate features, which are features of advertisements suitable for achieving the target in each of a plurality of advertising media. The machine learning includes, for example, correlation rule learning, Support Vector Machine (SVM), Extreme Learning Machine (ELM), learning using the error backpropagation method (backpropagation) related to neural networks, and the like. The machine learning preferably includes, in particular, machine learning with the advertising medium and performance as objective variables and the appropriate features as explanatory variables. Thereby, the management device 1 can predict appropriate features suitable for achieving the target by means of processing using machine learning with the advertising medium and the target performance as inputs.

[0073] In order to reduce the labor involved in prediction, the machine learning preferably includes unsupervised learning. Also, in order to further improve the accuracy of prediction, the machine learning may include supervised learning using external evaluation related to the validity of prediction and the like.

[0074] In order to reflect the differences between advertising media, in the second machine learning step, it is preferable to cause different machine learning models for each advertising medium to perform machine learning related to the prediction of appropriate features.

[0075] After the machine learning is performed, the management device 1 executes a series of processes related to the estimation of appropriate advertisement delivery conditions and appropriate advertising media. Steps S6 to S11 are an example of the flow of a series of processes related to the estimation of appropriate advertisement delivery conditions and appropriate advertising media.

[0076] [Step S6: Determine whether to obtain requirements for advertisement delivery, etc.] The control unit 11 cooperates with the storage unit 13 and executes the requirement acquisition unit 115. Then, the control unit 11 executes a process of determining whether to acquire requirements for advertisement distribution by the requirement acquisition unit 115 (step S6, requirement acquisition necessity determination step). If it is determined that acquisition is required, the control unit 11 acquires the requirements and moves the process to step S7. If it is not determined that acquisition is required, the control unit 11 returns the process to step S1 and repeats the processes from step S1 to step S19.

[0077] The procedure for determining whether to acquire requirements for advertisement distribution in the requirement acquisition necessity determination step is not particularly limited. The procedure preferably includes a procedure for determining that requirements for advertisement distribution are acquired when the requirements for advertisement distribution are received from the terminal T in order to enable estimation of an appropriate advertisement medium or the like based on the user's wishes.

[0078] The requirements acquired in the requirement acquisition necessity determination step include the requirements required for advertisement distribution conditions and the target of advertisement performance.

[0079] The requirements referred to here include one or more of the type of unit price, the upper limit value of the unit price, the target value of the unit price, the period during which the advertisement is conducted, the budget per certain period, the overall budget for the advertisement, the keywords related to the advertisement, the content of the advertisement, etc.

[0080] By including the type of unit price and the upper limit value and / or the target value of the unit price in the requirements, the management device 1 can make a prediction regarding the requirement to keep the unit price within a desired range. By including the period during which the advertisement is conducted in the requirements, the management device 1 can make a prediction based on the requirements regarding the period and time. By including the budget per certain period and / or the overall budget for the advertisement in the requirements, the management device 1 can make a prediction regarding the requirement to keep the cost within a desired range. By including the keywords related to the advertisement and / or the content of the advertisement in the requirements, the management device 1 can make a prediction based on the content, field, etc. of the advertisement.

[0081] In addition, the advertising performance target mentioned here includes the target of making one or more of various indicators of web advertising better than a specific value. Various indicators of web advertising include, for example, the number of ad impressions (Imp), cost per mille (CPM), number of clicks (Click), cost per click (CPC), click-through rate (CTR), number of conversions (CV), cost per action (CPA), conversion rate (CVR), acquisition cost (Cost), etc.

[0082] By including the target of making Imp and / or CPM better than a specific value, the management device 1 can make a prediction suitable for the cognitive expansion phase of enhancing the popularity of products, services, etc. By including the target of making Click, CPC, CTR, CV, CPA, and / or CVR better than a specific value, the management device 1 can make a prediction that emphasizes the results of advertising such as clicks and conversions. By including the target of making CPM, CPC, CPA, and / or Cost better than a specific value, the management device 1 can make a prediction that emphasizes cost-effectiveness and / or cost.

[0083] [Step S7: Predict the correspondence between advertising delivery conditions and performance] The control unit 11 cooperates with the storage unit 13 and executes the relationship prediction unit 116. Then, the control unit 11 executes, by the relationship prediction unit 116, a process of predicting the correspondence between advertising delivery conditions and performance for each of a plurality of advertising media based on machine learning in the first machine learning step (step S7, relationship prediction step). The control unit 11 transfers the process to step S8.

[0084] The relationship prediction step preferably includes a process using machine learning in a first machine learning step that takes as input the advertising medium and the performance target obtained in the requirement acquisition necessity determination step and outputs the advertising distribution conditions for achieving the target. Thereby, the management device 1 can estimate the appropriate advertising distribution conditions, which are the advertising distribution conditions for achieving the above-described target, based on the output. Further, the relationship prediction step preferably includes a process using machine learning in a first machine learning step that outputs the advertising distribution conditions for achieving each of one or more neighboring targets that are in the vicinity of the above-described target, for each of the neighboring targets. Thereby, the management device 1 can estimate the advertising distribution conditions not only for the advertising distribution conditions for achieving the target itself, but also for more preferable or more conservative performance in the vicinity of the target.

[0085] The management process preferably includes a series of processes for notifying that it is necessary to correct the requirements and the like when there is no correspondence relationship that satisfies both the fulfillment of the above-described requirements and the realization of the above-described target among the correspondence relationships predicted in the relationship prediction step. Thereby, the management device 1 can prompt correction for requirements and the like that are predicted to be difficult to realize. Steps S8 to S9 are an example of the flow of a series of processes related to the notification.

[0086] [Step S8: Determine whether there is a correspondence relationship that satisfies both the fulfillment of the requirements and the realization of the target] The control unit 11, in cooperation with the storage unit 13, executes a process of determining whether there is a correspondence relationship that satisfies both the fulfillment of the requirements related to the requirements and the like obtained in the requirement acquisition necessity determination step and the realization of the target among the correspondence relationships predicted in the relationship prediction step (step S8, correspondence relationship existence determination step). If it is determined that there is, the control unit 11 moves the process to step S10. If it is not determined that there is, the control unit 11 moves the process to step S9.

[0087] [Step S9: Notify that it is necessary to correct the requirements and the like] The control unit 11 cooperates with the storage unit 13 and the communication unit 14, and executes a process of notifying the terminal T or the like that a modification such as a requirement is necessary (step S9, necessary modification notification step). If it is determined that there is, the control unit 11 returns the process to step S1, and repeats the processes from step S1 to step S19.

[0088] [Step S10: Estimate appropriate advertisement delivery conditions] The control unit 11 cooperates with the storage unit 13 and executes the estimation unit 117. Then, the control unit 11 executes a process of estimating, by the estimation unit 117, appropriate advertisement delivery conditions, which are advertisement delivery conditions suitable for fulfilling requirements and achieving goals, for each of a plurality of advertisement media based on the prediction in the relationship prediction step (step S10, appropriate advertisement delivery condition estimation step). The control unit 11 moves the process to step S11.

[0089] The appropriate advertisement delivery condition estimation step estimates the advertisement delivery conditions most suitable for fulfilling requirements and achieving goals by extrapolation, selection, etc. based on the prediction in the relationship prediction step. The appropriate advertisement delivery condition estimation step estimates, for example, for each of a plurality of advertisement media, the advertisement delivery conditions corresponding to the prediction with the best performance among the predictions in the relationship prediction step as the appropriate advertisement delivery conditions.

[0090] To enable advertisements at appropriate costs, the appropriate advertisement delivery condition estimation step preferably estimates various conditions related to the consideration of advertisements exemplified by the type of performance (number of clicks, conversion, number of impressions, etc.) used for calculating the unit price, the upper limit value and target value of the unit price, the budget per certain period, the budget for the entire advertisement, etc. Also, to enable advertisements at appropriate timings, the appropriate advertisement delivery condition estimation step preferably estimates the conditions related to the timing of advertisements exemplified by the period during which the advertisement is performed, etc.

[0091] [Step S11: Estimate appropriate advertisement media] The control unit 11 cooperates with the storage unit 13 and executes the estimation unit 117. Then, based on the prediction in the relationship prediction step and the estimation in the appropriate advertisement delivery condition estimation step by the estimation unit 117, the control unit 11 executes a process of estimating an appropriate advertisement medium that is an advertisement medium suitable for meeting the requirements and achieving the goal among a plurality of advertisement media (step S11, appropriate advertisement medium estimation step). The control unit 11 moves the process to step S12.

[0092] The appropriate advertisement medium estimation step estimates the advertisement medium most suitable for meeting the requirements and achieving the goal by extrapolation, selection, etc. based on the prediction in the relationship prediction step and the estimation in the appropriate advertisement delivery condition estimation step. The appropriate advertisement medium step estimates, for example, among a plurality of advertisement media, the advertisement medium corresponding to the most excellent appropriate advertisement delivery condition as the appropriate advertisement medium. The most excellent appropriate advertisement delivery condition is, for example, the condition in which any one of the conditions corresponding to the goal is the most excellent, the condition in which the evaluation value using a plurality of conditions corresponding to the goal is the most excellent, etc.

[0093] The management process preferably includes a series of processes of predicting appropriate features that are features of advertisements suitable for achieving the goal and generating advertisements having the appropriate features. Steps S12 to S14 are an example of the flow of a series of processes related to the prediction and generation. Note that the advertisements related to the generation of the process include various forms of advertisements exemplified by banners, which are images or videos for advertisements displayed in a certain area, advertisement articles, SEO articles, etc.

[0094] [Step S12: Determine whether to generate an advertisement] The control unit 11 cooperates with the storage unit 13 and executes a process of determining whether to generate an advertisement (step S12, advertisement generation necessity determination step). If it is determined to generate, the control unit 11 moves the process to step S13. If it is not determined to generate, the control unit 11 moves the process to step S15.

[0095] In order to generate an advertisement according to the user's request, the advertisement generation necessity determination step preferably includes a procedure for determining whether to generate an advertisement when a generation instruction for the advertisement is received from the terminal T.

[0096] [Step S13: Predicting appropriate features] The control unit 11 cooperates with the storage unit 13 to execute the feature prediction unit 118. Then, the control unit 11 executes a process of predicting appropriate features, which are features of an advertisement suitable for achieving the target obtained in the requirement acquisition necessity determination step, by the feature prediction unit 118 (step S13, feature prediction step). The control unit 11 moves the process to step S14.

[0097] The feature prediction step predicts appropriate features based on the machine learning performed in the second machine learning step. The machine learning includes, as learning data, the features of the advertisement discriminated by the feature discrimination unit 113 in the feature discrimination step and the results related to the advertisement. Thereby, the management device 1 can predict appropriate features based on the features and results of advertisements performed in the past.

[0098] Specifically, the feature prediction step predicts appropriate features, which are features of an advertisement suitable for achieving the target in the appropriate advertisement medium, by a process based on the machine learning that takes, as inputs, the requirements and target obtained in the requirement acquisition necessity determination step and the appropriate advertisement medium estimated in the appropriate advertisement medium estimation step.

[0099] In the feature prediction step, the management device 1 predicts appropriate features, which are features of an advertisement suitable for achieving the target, for various features such as, for example, BGM, voice tone, presence or absence of characters, features of characters exemplified by gender, age, appearance, clothing, behavior, etc., type of background exemplified by shooting location, etc., type of props, diction, turns of phrase, etc.

[0100] [Step S14: Generating an advertisement having appropriate features] The control unit 11 collaborates with the storage unit 13 and executes the advertisement generation unit 119. Then, the control unit 11 executes a process of generating an advertisement having the appropriate features predicted in the feature prediction step by the advertisement generation unit 119 (step S14, advertisement generation step). The control unit 11 moves the process to step S15.

[0101] In the advertisement generation step, an advertisement having appropriate features is generated by the generation neural network. The advertisement generation step may include a procedure of generating a plurality of components according to the type of advertisement by another generation neural network and concatenating the generated plurality of components. Further, in the advertisement generation step, an advertisement article or an SEO article may be generated based on the appropriate features predicted in the feature prediction step. An SEO article is a text aimed at resolving the intention of a searcher and is content optimized for a search engine so as to have a high probability of being displayed on the top page of search results. Thereby, the problem that time and labor are required for creating the advertisement article and the SEO article is solved. The generation of the advertisement article and the SEO article may be the same as the generation of the advertisement except that the article includes the above-mentioned text.

[0102] The following are examples of means for generating each of the plurality of components. In the advertisement generation step, for example, an advertisement text constituting the advertisement is generated by a large language model. In the advertisement generation step, for example, the voice constituting the advertisement is generated by a voice generation neural network. In the advertisement generation step, for example, an image or a video such as video material constituting the advertisement is generated by an image generation neural network. In addition, in the advertisement generation step, for example, keywords, captions, etc. related to the advertisement may be generated by a large language model. Further, in the advertisement generation step, an effect to be applied to the image, video, voice, etc. of the advertisement may be generated by a neural network related to the generation of the effect.

[0103] In web advertising, characteristics related to various elements of an advertisement, such as an advertisement text, advertisement voice, advertisement layout, advertisement image, etc., can have a great impact on the performance of the advertisement. However, advertisers who are not proficient in the management of web advertisements in a specific advertising medium may have difficulty setting appropriate advertisement characteristics in order to obtain a desired advertising effect in the advertising medium.

[0104] In the appropriate characteristic prediction step, the characteristic prediction unit 118 predicts an appropriate characteristic, which is an advertisement characteristic suitable for achieving a goal, based on machine learning in the second machine learning step. The machine learning is machine learning that includes the advertisement characteristics discriminated by the characteristic discrimination unit 113 in the characteristic discrimination step in the learning data. Then, in the advertisement generation step, the advertisement generation unit 119 generates an advertisement having the predicted appropriate characteristics by a generation neural network. Thereby, the management device 1 can support the advertiser in setting appropriate advertisement characteristics in order to obtain a desired advertising effect.

[0105] In the advertisement generation step, the management device 1 can generate an advertisement having appropriate features, which are features of an advertisement suitable for achieving the target, for various features exemplified by, for example, BGM, voice tone, presence or absence of characters, characteristics of characters exemplified by gender, age, appearance, clothing, behavior, etc., types of backgrounds exemplified by shooting locations, etc., types of props, diction, turns of phrase, etc. Note that the advertisement generation step preferably includes a procedure for improving various elements constituting the advertisement, etc. through dialogue with the user via a so-called chatbot. Thereby, in automatic generation using AI such as a generative neural network, the problem that the quality of generation parameters such as prompts related to the automatic generation depends on the user's proficiency can be solved. The procedure is realized by a large language model in which machine learning is performed using learning data including questions regarding generation parameters of various elements as explanatory variables and appropriate answers to the questions as target variables. That is, the procedure is, for example, to receive a question regarding generation parameters of various elements from the terminal T, cause the large language model in which the above-mentioned machine learning has been performed to generate an answer to the question, transmit the answer to the terminal T, and when receiving a generation instruction for various elements including generation parameters from the terminal T, perform generation of the above-mentioned various elements using the generation parameters, etc.

[0106] [Step S15: Predict performance related to appropriate advertising media and appropriate advertising distribution conditions] The control unit 11 cooperates with the storage unit 13 and executes the relationship prediction unit 116. Then, the control unit 11 executes a process of predicting the performance related to the appropriate advertising media and the appropriate advertising distribution conditions based on the machine learning in the first machine learning step by the relationship prediction unit 116 (step S15, performance prediction step). The control unit 11 moves the process to step S16.

[0107] The prediction related to the performance prediction step may be the same as the prediction related to the relationship prediction step, except that it predicts the performance related to the appropriate advertising media and the appropriate advertising distribution conditions.

[0108] [Step S16: Issue a command to display appropriate advertising media, etc.] The control unit 11 cooperates with the storage unit 13 and the communication unit 14 to execute a process of instructing the terminal T or the like to display an appropriate advertising medium or the like (step S16, display instruction step). The control unit 11 moves the process to step S17.

[0109] More specifically, the display instruction step instructs the display of the appropriate advertising medium estimated in the appropriate advertising medium estimation step, the appropriate advertising distribution condition estimated in the appropriate advertising distribution condition estimation step, and the performance predicted in the performance prediction step. Thereby, the management device 1 can recommend to the user the desired requirements related to the advertising distribution conditions and the preferable advertising medium and advertising distribution conditions related to the desired performance automatically generated by the above-described process.

[0110] By the recommendation of the management device 1, the user of the management device 1 can grasp the desired requirements related to the advertising distribution conditions, the appropriate advertising medium that satisfies the desired performance, and the appropriate advertising distribution conditions in the medium. Further, the user can grasp the performance predicted under the medium and the conditions. Thereby, the user can determine whether to start the distribution of the advertisement based on the predicted performance or the like.

[0111] The display instruction step can display predicted values of various performances such as the number of ad displays (Imp), cost per mille (CPM), number of clicks (Click), cost per click (CPC), click-through rate (CTR), number of conversions (CV), cost per action (CPA), conversion rate (CVR), acquisition cost (Cost), etc. in a user interface that is easy to view. Thereby, the management device 1 can easily grasp the predicted values of the performance related to the advertisement.

[0112] When an advertisement is generated in the advertisement generation step, the display instruction step preferably further instructs the display of the advertisement. Thereby, the user can determine whether to start the distribution of the advertisement based on the generated advertisement. Further, the user can appropriately modify the generated advertisement and make a determination to start the distribution of the advertisement.

[0113] [Step S17: Determine whether there is a command to start advertising distribution] The control unit 11 cooperates with the storage unit 13 and the communication unit 14 to execute the management unit 120. Then, the control unit 11 executes, through the management unit 120, a process of determining whether there is a command to start advertising distribution for the advertisement related to the display command step (step S17, determination step for presence / absence of distribution start command). If it is determined that there is a command, the control unit 11 moves the process to step S19. If it is determined that there is no command, the control unit 11 moves the process to step S18.

[0114] The command related to the determination step for presence / absence of distribution start command may be a command to start advertising distribution according to the appropriate advertising distribution conditions estimated in the appropriate advertising distribution condition estimation step. Also, the command related to the determination step for presence / absence of distribution start command may be a command to start advertising distribution according to the advertising distribution conditions specified by the user. Since the appropriate advertising distribution conditions are displayed in the display command step, the user can specify the desired advertising distribution conditions in a relatively easy procedure for modifying the appropriate advertising distribution conditions.

[0115] [Step S18: Determine whether there is a command to change requirements, etc.] The control unit 11 cooperates with the storage unit 13 and the communication unit 14 to execute the management unit 120. Then, the control unit 11 executes, through the management unit 120, a process of determining whether there is a command to change requirements, etc. for the advertisement related to the display command step (step S18, determination step for presence / absence of requirement change command). If it is determined that there is a command, the control unit 11 moves the process to step S6. If it is determined that there is no command, the control unit 11 moves the process to step S17.

[0116] [Step S19: Start the advertisement on the appropriate advertising medium] The control unit 11 cooperates with the storage unit 13 and the communication unit 14 to execute the management unit 120. Then, the control unit 11 executes, by the management unit 120, a process of starting an advertisement in the appropriate advertising medium estimated in the appropriate advertisement medium estimation step (step S19, advertisement distribution start step). The control unit 11 returns the process to step S1 and repeats the processes from step S1 to step S19.

[0117] In the advertisement distribution start step, the advertisement distribution is started based on the appropriate advertisement distribution conditions estimated in the appropriate advertisement distribution condition estimation step or the conditions modified by the user. The procedure for starting the advertisement distribution includes, for example, a step of instructing a server or the like related to the appropriate advertising medium to start the advertisement distribution under the specified advertisement distribution conditions.

[0118] [Advertisement stop step] As a process related to the management unit 120, the management process preferably includes a step of stopping the advertisement related to the prediction (advertisement stop step) when the correspondence relationship predicted by the relationship prediction unit 116 does not include a correspondence relationship that satisfies both the requirements obtained in the requirement acquisition necessity determination step and the achievement of the target.

[0119] An advertiser who is not proficient in managing web advertisements in a specific advertising medium may have difficulty setting the requirements for appropriate advertisement distribution conditions in order to obtain a desired advertising effect in the advertising medium. Therefore, it may be difficult to achieve the initially expected performance target with the advertisement distribution conditions that satisfy the initially given requirements. If the advertisement is automatically continued in such a case, the advertiser may pay an advertising fee for an advertisement different from the expectation.

[0120] In the advertisement stop step, when the correspondence relationship predicted in the relationship prediction step does not include a correspondence relationship that satisfies both the requirements acquired in the requirement acquisition necessity determination step and the achievement of the target related to the requirements, the management unit 120 stops the advertisement related to the prediction. Thereby, when it is difficult for the management device 1 to achieve the target performance initially expected under the advertisement distribution conditions that satisfy the initially given requirements, it prevents the advertiser from paying advertising fees for advertisements different from the expectations.

[0121] The timing at which the advertisement stop step is executed preferably includes after the start of the advertisement. Thereby, when the management device 1 can predict after the start of the advertisement that it is difficult to achieve the target performance initially expected under the advertisement distribution conditions that satisfy the initially given requirements, it prevents the advertiser from paying advertising fees for advertisements different from the expectations.

[0122] [Conversion display step] The management process preferably includes a series of processes in which the actual results acquisition unit 111 acquires the conversion actual results related to the advertisement during or after distribution from a server or the like that conducts online transactions for the advertisement target managed by the advertiser, and causes the terminal T or the like to display the actual data of the acquired conversion. Thereby, the user can easily grasp the actual data of the actual results related to the conversion exemplified by the CV or the like.

[0123] [Effect of management process] Web advertisements are carried out through various advertising media. These advertising media may differ in the fields they are good at, price ranges, etc. for each advertising media. Therefore, in order to conduct effective web advertisements, the selection of appropriate advertising media is important.

[0124] By executing the above-described management process, the management device 1 of the present embodiment can perform machine learning and prediction that contribute to appropriate comparison among a plurality of advertising media. Specifically, the management device 1 performs machine learning that contributes to appropriate comparison among a plurality of advertising media in the first machine learning step and the like. Further, the management device 1 performs prediction that contributes to appropriate comparison among a plurality of advertising media in the relationship prediction step and the like.

[0125] In the above management process, based on machine learning, for each of a plurality of advertising media, the correspondence between the advertisement delivery conditions and the results is predicted. Therefore, the management device 1 can estimate an appropriate advertising media, which is an advertising media suitable for achieving the goal aimed by the advertiser among the plurality of advertising media, based on this prediction. And the management device 1 can provide the appropriate advertising media and the appropriate advertisement delivery conditions in the appropriate advertising media. In addition, the management device 1 can provide the prediction result of the results in the appropriate advertising media in the display instruction step. Therefore, the management device 1 can assist in the start and management of advertisements including the selection of appropriate advertising media in web advertisements.

[0126] In the management process, the management device 1 predicts the results and estimates the appropriate advertisement delivery conditions and the appropriate advertising media for various results. Thereby, the management device 1 can assist in the start and management of advertisements including the selection of appropriate advertising media for goals corresponding to various desires exemplified by desires such as wanting to maximize profits, wanting to maximize the number of conversions, and wanting to lower the acquisition unit price.

[0127] By the way, creating an advertisement requires not only knowledge about products and services to be advertised, etc., but also extensive knowledge about texts, voices, images, etc. related to the advertisement. Therefore, it is not easy to create an advertisement. In the above management process, based on machine learning, an advertisement suitable for achieving the target results in an appropriate advertising media can be automatically generated. Therefore, the management device 1 can assist in creating an advertisement suitable for achieving the target results in the appropriate advertising media as well as only selecting the appropriate advertising media in web advertisements.

[0128] The user interfaces related to the management of advertisements of the plurality of advertising media are different from each other. The management device 1 can centrally manage the advertisements in the plurality of advertising media by the management unit 120. And the management device 1 enables a user to submit and deliver an advertisement using a plurality of advertising media without being aware of the differences in the advertising media through a user interface that displays the appropriate advertising media and the appropriate advertisement delivery conditions.

[0129] Therefore, the management device 1 that executes the above-described management process can support an advertising activity that uses a plurality of advertising media with different operating entities in web advertising.

[0130] <Usage Example> The following is a usage example of the management device 1.

[0131] [Machine Learning] The administrator of the management device 1 performs procedures related to machine learning.

[0132] [Provision of Learning Data] The administrator of the management device 1 causes the advertising distribution conditions and the data on the results under such conditions to be acquired. The data includes data related to the results of the advertisements managed by the management device 1. Further, in order to improve the accuracy of prediction and estimation, it is preferable that the data further includes data acquired from external advertising media, external databases, etc.

[0133] [Execution of Machine Learning] The administrator issues an instruction to execute machine learning to the management device 1. The management device 1 executes machine learning related to the prediction of the correspondence between the advertising distribution conditions and the results based on the results stored in the results database 132. In order to generate an appropriate advertisement, it is preferable that the management device 1 further executes machine learning related to the prediction of appropriate features, which are the features of an advertisement suitable for the achievement of the goal.

[0134] [Distribution of Web Advertisements] The user of the management device 1 starts the distribution of web advertisements via the terminal T.

[0135] [Specification of Requirements, etc.] The user of the management device 1 provides the management device 1 with the requirements and the results target (requirements, etc.) related to the advertising distribution conditions of the web advertisement via the terminal T. The management device 1 acquires the provided requirements, etc.

[0136] [Estimation of Appropriate Advertising Media, etc.] The management device 1 estimates an appropriate advertising medium that meets the provided requirements and the like and is suitable for achieving the goal, and appropriate advertising distribution conditions suitable for achieving the goal in the said medium based on machine learning. Further, the management device 1 predicts the results to be predicted under the said conditions.

[0137] [Generation of advertisements, etc.] The user may instruct the management device 1 to generate advertisements such as banners, advertising articles, etc. or SEO articles, etc. (advertisements, etc.) via the terminal T or the like. The management device 1 generates advertisements, etc. that are suitable for achieving the goal in the estimated appropriate advertising medium. The user can instruct the generation for any type of advertisement, etc., such as text advertisements, image advertisements, video advertisements, etc. Further, the user can partially or entirely modify the generated advertisements, etc. Thereby, the user can create and distribute desired advertisements, etc. with less effort than creating advertisements, etc. from scratch. In addition, the user can improve the generation parameters of advertisements, etc. via the chatbot. Thereby, even a user who is not familiar with generation using AI can create and distribute desired advertisements, etc.

[0138] [Confirmation of estimation results, etc.] The management device 1 causes the terminal T to display the estimated appropriate advertising medium and appropriate advertising distribution conditions, and the results to be predicted in the said medium and the said conditions. The user confirms the appropriate advertising medium, appropriate advertising distribution conditions, and predicted results displayed on the terminal T. Then, the user considers whether to distribute the advertisement under the said medium and conditions.

[0139] [Start web advertisement distribution] Based on the above display, the user instructs the management device 1 to start the distribution of the advertisement. The management device 1 causes the appropriate advertising medium to start the distribution of the advertisement.

[0140] (Distribution under appropriate advertising distribution conditions) The user can instruct the management device 1 to start the distribution of the advertisement under the displayed appropriate advertising distribution conditions. The management device 1 instructs the appropriate advertising medium to start the distribution of the advertisement under the displayed appropriate advertising distribution conditions.

[0141] (Delivery under conditions with modified appropriate advertisement delivery conditions) The user can modify the displayed appropriate advertisement delivery conditions. Then, the user can instruct the management device 1 to start delivering advertisements under the modified advertisement delivery conditions. The management device 1 instructs the appropriate advertisement medium to start delivering advertisements under the modified advertisement delivery conditions.

[0142] In the scope of the idea of the present invention, those skilled in the art can conceive of various modification examples and correction examples. Therefore, those modification examples and correction examples are understood to belong to the scope of the present invention. For example, for the above-described embodiments, those obtained by appropriately adding, deleting, or changing the design of components by those skilled in the art, or those obtained by adding, omitting, or changing conditions of steps, are also included in the scope of the present invention as long as they have the gist of the present invention.

Explanation of Signs

[0143] S System 1 Management device 11 Control unit 111 Performance acquisition unit 112 First machine learning unit 113 Feature discrimination unit 114 Second machine learning unit 115 Requirement acquisition unit, etc. 116 Relationship prediction unit 117 Estimation unit 118 Feature prediction unit 119 Advertisement generation unit 120 Management unit 13 Storage unit 131 Advertisement database 132 Performance database 14 Communication unit N Network T Terminal

Claims

1. For each of a plurality of advertising media with different operating entities, a management unit that manages advertisements using the advertising media, A performance acquisition unit that acquires the performance related to each of the advertisements, For each of the plurality of advertising media, a machine learning unit that executes machine learning related to predicting the correspondence between advertising delivery conditions and performance, A requirement and other acquisition unit that acquires the requirements required for the advertising delivery conditions and the goals of the advertisement performance, For each of the plurality of advertising media, a relationship prediction unit that predicts the correspondence between the advertising delivery conditions and the performance based on the machine learning, Based on the prediction, an estimation unit that estimates an appropriate advertising media that is an advertising media suitable for satisfying the requirements and achieving the goals among the plurality of advertising media, comprising The machine learning unit executes machine learning that includes the association of the advertising media, the advertising delivery conditions, and the performance in the learning data, The estimation unit further estimates appropriate advertising delivery conditions that are advertising delivery conditions suitable for achieving the goals in the appropriate advertising media, A management device for web advertisements.

2. The performance includes the history of the performance before and after the change of the advertising delivery conditions, The machine learning unit executes machine learning that includes the history in the learning data, The management device according to claim 1.

3. The machine learning unit executes machine learning that includes the association of the prediction by the relationship prediction unit and the performance corresponding to the prediction in the learning data, the management device according to claim 1.

4. When the management unit does not include a correspondence relationship in which the correspondence relationship in the relationship prediction unit satisfies both the fulfillment of the requirements and the achievement of the goals, the management unit stops the advertisement related to the prediction, the management device according to claim 1.

5. A feature discrimination unit that discriminates the features of the advertisement, A feature prediction unit that predicts appropriate features, which are features of an advertisement suitable for achieving the target, based on machine learning that includes the features in learning data; An advertisement generation unit that generates an advertisement having the appropriate features; further comprising; The management device according to claim 1.

Images