Advertising management device, system, method, and program

Abstract

To provide an advertising management device, system, method, and program that can further optimize the delivery of advertisements to users. [Solution] In the advertising management system 1, the advertising management server includes: an activity history DB that stores activity history information, which is information about the activity history of users who have visited the e-commerce site; a delivery content setting unit that identifies the products viewed by users who have left the e-commerce site from the activity history DB and includes a retargeting message in the advertising information that suggests the purchase of the product; a delivery timing setting unit that sets the delivery timing of advertising information to the users who have left the site by using a mathematical model that searches for delivery timings with high open rates for advertising information previously delivered to users who have left the site based on the activity history information stored in the activity history DB and outputs the optimal delivery timing; and an advertising delivery unit that delivers the advertising information to the users who have left the site at the delivery timing.

Description

Technical Field

[0001] The present disclosure relates to an advertisement management device, system, method, and program.

Background Art

[0002] Advertisers and operators of e-commerce sites (EC sites) (hereinafter collectively referred to as advertisement managers) are considering the appropriate delivery timing of advertisements so that the delivery effect of the advertisements is maximized. For example, even when a viewer (user) of an EC site leaves the EC site without putting the viewed product in the EC cart, the advertisement manager sends a retargeting email (so-called browser abandonment email) that proposes to the user to purchase the viewed product in order to prompt the user to purchase the viewed product. As an advertisement manager, there is a desire to send such a retargeting email at a timing when the probability that the user opens the email (open rate) is maximized.

[0003] Patent Documents 1 to 3 describe technologies aimed at appropriately setting the delivery timing of advertisements. For example, Patent Document 1 describes a mail proposal system that determines the delivery timing of a mail magazine to a user by utilizing machine learning based on information such as the open rate of the mail magazine. Patent Document 2 describes a content delivery system that determines the delivery timing of electronic content to a user by utilizing machine learning based on the time zone when the past electronic content was clicked. Further, Patent Document 3 describes a delivery device that determines the delivery timing of notification information (for example, information related to a product) based on the purchase history of a product in a user's e-commerce service by utilizing machine learning based on the delivery timing when the past notification information was clicked.

Prior Art Documents

Patent Documents

[0004]

Patent Document 1

[0005] However, the technologies described in Patent Documents 1 to 3 all have a simple process for determining the timing of ad delivery, leaving room for improvement in optimizing ad delivery timing. Furthermore, ad managers also want the content of retargeting emails and other advertisements to be more tailored to the user.

[0006] Therefore, the purpose of this disclosure is to provide an advertising management device, system, method, and program that can further optimize the delivery of advertisements to users. [Means for solving the problem]

[0007] To achieve the above-mentioned objectives, the advertising distribution device relating to this disclosure is an advertising management device that distributes advertising information to users who have left a website after viewing product information on the website, and comprises: a behavior history storage unit that stores behavior history information, which is information regarding the behavior history of users who have visited the website; a distribution content setting unit that identifies the products viewed by the user who left the website from the behavior history storage unit and includes a retargeting message in the advertising information that proposes the purchase of the product; a distribution timing setting unit that sets the distribution timing of the advertising information to the user by using a mathematical model that searches for distribution timings with high open rates for advertising information distributed to past users who left the website based on the behavior history information stored in the behavior history storage unit and outputs an optimal distribution timing; and an advertising distribution unit that distributes the advertising information set by the distribution content setting unit to the user at the distribution timing set by the distribution timing setting unit.

[0008] Furthermore, in order to achieve the above-mentioned objectives, the advertising management system relating to this disclosure is an advertising management device that delivers advertising information to users who have left a website after viewing product information on the website, and comprises: a behavior history storage unit that stores behavior history information, which is information regarding the behavior history of users who have visited the website; a delivery content setting unit that identifies the products viewed by users who have left the website from the behavior history storage unit and includes a retargeting message in the advertising information that proposes the purchase of the product; a delivery timing setting unit that sets the delivery timing of the advertising information to the users who have left the website by using a mathematical model that searches for delivery timings with high open rates for advertising information delivered to past users who have left the website based on the behavior history information stored in the behavior history storage unit and outputs an optimal delivery timing; and an advertising delivery unit that delivers the advertising information set by the delivery content setting unit to the users who have left the website at the delivery timing set by the delivery timing setting unit.

[0009] Furthermore, in order to achieve the above-mentioned objectives, the advertising distribution method relating to this disclosure is an advertising management method in which a computer distributes advertising information to users who have left a website after viewing product information, and includes in the advertising information a retargeting message proposing the purchase of the product that the user who left the website viewed, based on behavioral history information which is information about the behavioral history of users who visited the website; a distribution content setting step in which the computer distributes the advertising information to the user who left the website, based on the behavioral history information stored in the behavioral history storage unit, identifies the products that the user who left the website viewed, and includes in the advertising information a retargeting message proposing the purchase of the product; a distribution timing setting step in which the computer distributes the advertising information to the user who left the website, based on the behavioral history information stored in the behavioral history storage unit, searches for distribution timings with high open rates for advertising information distributed to past users who left the website, and outputs an optimal distribution timing; and an advertising distribution step in which the computer distributes the advertising information set in the distribution content setting step to the user who left the website at the distribution timing set in the distribution timing setting step.

[0010] Furthermore, in order to achieve the above-mentioned objectives, the advertising distribution program relating to this disclosure is an advertising management program that causes a computer to distribute advertising information to users who have left a website after viewing product information on the website, and includes in the advertising information a retargeting message suggesting the purchase of the product that the user viewed, based on the behavior history information which is information about the behavior history of users who have visited the website, from the behavior history storage unit which stores behavior history information which is information about the behavior history of users who have visited the website; a distribution timing setting step which sets the distribution timing of the advertising information to the user by using a mathematical model that searches for distribution timings with high open rates for advertising information distributed to past users who have left the website based on the behavior history information stored in the behavior history storage unit and outputs the optimal distribution timing; and an advertising distribution step which distributes the advertising information set in the distribution content setting step to the user at the distribution timing set in the distribution timing setting step. [Effects of the Invention]

[0011] The advertising management devices, systems, methods, and programs described herein can be used to further optimize the delivery of advertisements to users. [Brief explanation of the drawing]

[0012] [Figure 1] This is an overall configuration diagram showing one embodiment of the advertising management system related to this disclosure. [Figure 2] This is a flowchart showing the advertising delivery procedure in the advertising management system related to this disclosure. [Modes for carrying out the invention]

[0013] Figure 1 shows an overall configuration diagram illustrating one embodiment of the advertising management system according to this disclosure. Based on this figure, the configuration of the advertising management system 1 according to this embodiment will be described.

[0014] The advertising management system 1 comprises an advertising management server (advertising management device) 2 and a plurality of user terminals 3 that can communicate with each other via a network N. The network N is a communication network capable of bidirectional information transmission using wired or wireless communication means, such as the internet, intranet, or VPN (Virtual Private Network).

[0015] The advertising management server 2 is a server that publishes an e-commerce site (hereinafter referred to as the EC site), which is a website, and is also a server that appropriately delivers advertising information related to the EC site to user terminals via email. The advertising management server 2 comprises an input unit 11, a display unit 12, a communication unit 13, a management control unit 14, and a storage unit 15. In the following, "administrator" means the administrator of the EC site in this embodiment and the operator of advertising information distribution (i.e., the administrator of the advertising management server 2), and "user" means a visitor (including users) of the EC site.

[0016] The input unit 11 is an operating means that can input instructions to the management control unit 14, and is, for example, a touch panel such as a liquid crystal display or an organic EL display, a keyboard, a mouse, or an audio input device.

[0017] The display unit 12 is, for example, a liquid crystal display or an organic EL display, and is configured to display various types of information. For example, the display unit 12 can view and confirm advertising information that an administrator will distribute to the user terminal 3.

[0018] The communication unit 13 is configured to communicate with the user terminal 3 via the network N. For example, the communication unit 120 is a wired LAN, wireless LAN, etc., that can connect to the internet.

[0019] The management control unit 14 is a computer having a central processing unit (CPU) of the advertising management server 2 and is capable of executing the advertising management program in the present embodiment. The management control unit 14 of the present embodiment includes an EC site management unit 21, a behavior history acquisition unit 22, a related product generation unit 23, a distribution content setting unit 24, a distribution timing setting unit 25, an advertisement distribution unit 26, a reaction information acquisition unit 27, and a learning unit 28. Each of these units is, for example, a program module for exerting each function and executes each process under the instruction from the management control unit 14.

[0020] The storage unit 140 is configured to store various kinds of information and is composed of, for example, a main storage device (RAM, ROM, etc.), an auxiliary storage device (HDD, SSD, optical disk, etc.), a cache memory, and registers. The storage unit 15 of the present embodiment includes databases for each type of information, specifically, a site database 31, a user database 32, a product database 33, a behavior history database 34, a related product database 35, and a learned model database 36 (hereinafter, the database is referred to as DB). The information stored in each DB is mutually associated by using, for example, a shared identification number (user ID, product ID, etc.).

[0021] Hereinafter, each unit of the management control unit 14 and each DB of the storage unit 15 will be described in detail.

[0022] The EC site management unit 21 of the management control unit 14 has a function of managing (including publishing and updating) an e-commerce site (EC site), which is a website for the purpose of selling products. Specifically, the EC site management unit 21 publicly discloses the EC site information stored in the site DB 31 via the network N so that it can be browsed on the user terminal 3, performs access management at the time of user access, and performs processes such as updating each page. In the present embodiment, an EC site that sells clothing items as products will be described as an example. Note that what is sold on the EC site is not limited to clothing items, and other items may be sold, and the product is not limited to an item and includes a service.

[0023] The behavior history acquisition unit 22 has the function of acquiring behavior history information, which is information about the behavior history of users who have visited the e-commerce site managed by the e-commerce site management unit 21. The behavior history acquisition unit 22 also has the function of detecting abandoned users who have left individual product pages from the acquired behavior history information. In this embodiment, abandoned users are users who view a product page but leave the site without adding items to their cart, in other words, users who abandon their browsers.

[0024] The behavioral history acquisition unit 22 can acquire behavioral history information by using website analysis tools, cookies, analyzing server logs, or a combination of these methods. The behavioral history acquisition unit 22 also has the function of linking the acquired behavioral history information with user information stored in the user database 32 and saving it in the behavioral history database 34. Furthermore, the behavioral history acquisition unit 22 has the function of extracting product information from the product database 33 corresponding to the product pages viewed by the user, generating viewed product information, and storing it in the user database 32 in association with user information.

[0025] The related product generation unit 23 has the function of generating related product information, which is information about products related to the viewed product among the products handled on the e-commerce site. Specifically, the related product generation unit 23 acquires the viewed product information generated by the behavior history acquisition unit 22 or stored in the user DB 32, generates related product information related to the viewed product information using a predetermined mathematical model, and stores it in the related product DB. The predetermined mathematical model here is, for example, a collaborative filtering algorithm. In other words, the related product generation unit 23 in this embodiment uses a collaborative filtering algorithm to set one or more products as related products that are preferred by other users with similar tastes to the user who viewed the viewed product. Note that the predetermined mathematical model for generating related product information is not limited to a collaborative filtering algorithm, and other recommendation algorithms may be used. Furthermore, the predetermined mathematical model is stored in the trained model DB 36, and the related product generation unit 23 may generate the viewed product information using the mathematical model stored in the trained model DB 36.

[0026] The content setting unit 24 has the function of obtaining behavioral history information from the behavioral history DB 34, identifying the products viewed by users who left individual product pages (hereinafter referred to as "left users"), and setting the content of advertising information, including a retargeting message suggesting the purchase of those viewed products. Specifically, the advertising information in this embodiment is information contained in an email (HTML email) or information attached to an email.

[0027] Furthermore, the content setting unit 24 retrieves related product information from the related product database 35 that is associated with the viewed product that is the target of the retargeting message, and includes this related product information in the advertising information.

[0028] Furthermore, the content setting unit 24 can set different advertising information depending on the product. For example, different retargeting messages are stored in the product DB 33 depending on the product category (see below for details on categories), whether or not it is a campaign item, etc. The unit extracts the retargeting message corresponding to the viewed product from the product DB 33 and includes it in the advertising information.

[0029] The delivery timing setting unit 25 has the function of setting the delivery timing of the advertisement information to the lapsed user by using a mathematical model that searches for delivery timings with high open rates for advertisement information delivered to past lapsed users based on behavioral history information stored in the behavioral history DB 34 and outputs the optimal delivery timing. The mathematical model used to calculate the delivery timing is a trained model that has been machine-learned in the learning unit 28 (described later) and stored in the trained model DB 36.

[0030] The ad delivery unit 26 has the function of delivering ad information set by the delivery content setting unit 24 to churned users at the delivery timing set by the delivery timing setting unit 25. Specifically, the ad delivery unit 26 delivers ad information to churned users by including it in an email.

[0031] The advertising distribution unit 26 has the function of distributing or not distributing advertising information according to predetermined distribution conditions. There may be more than one predetermined distribution condition, and the administrator can set them as they see fit. The predetermined distribution conditions may be conditions for the user or conditions for the product, and may be either permission conditions that allow distribution when the conditions are met, or prohibition conditions that prevent distribution when the conditions are met.

[0032] For example, the advertising distribution unit 26 in this embodiment limits the products to which advertising information is distributed as a predetermined distribution condition. In other words, the advertising distribution unit 26 distributes advertising information only when the product in the viewed product information is a product that meets a predetermined condition (for example, a campaign product).

[0033] Furthermore, the advertising distribution unit 26 of this embodiment, as a predetermined distribution condition, will not distribute advertising information related to the product to a user who has left the site if the user's behavioral history information includes a history of a return or exchange.

[0034] The response information acquisition unit 27 has the function of acquiring response information to advertising information delivered to lapsed users. Specifically, the response information acquisition unit 27 acquires open information, which is response information, by acquiring an open beacon that is sent back when the lapsed user's terminal 3, which received the email containing the advertising information, opens the email. More specifically, the email delivered by the advertising delivery unit 26 in this embodiment is an HTML email, and a transparent image link as an open beacon is embedded in the HTML email. When the image link is loaded by the mailer on the lapsed user's terminal, the response information acquisition unit, which is the destination of the link, can detect that the email has been opened (image loaded). In other words, opening means that the user has become able to view the retargeting message. The response information acquisition unit 27 then stores the acquired response information in the behavior history DB, linked to the user information.

[0035] The learning unit 28 has the function of generating a mathematical model that searches for delivery timings with high open rates for advertising information delivered to past lapsed users based on behavioral history information stored in the behavioral history DB 34, and outputs the optimal delivery timing. In other words, the mathematical model in this embodiment is a trained model generated by machine learning.

[0036] More specifically, the learning unit 28 uses the behavior history information stored in the behavior history DB 34 and the response information acquired by the response information acquisition unit to perform reinforcement learning and generate a trained model (an agent in reinforcement learning). The basic idea of ​​the reinforcement learning algorithm here is the "Top-two Thompson Sampling" algorithm, which is an existing algorithm for the "optimal arm determination bandit problem" with {0,1} rewards. The "Top-two Thompson Sampling" algorithm is one of the reinforcement learning algorithms that balances probabilistic search and utilization, and is used in the so-called "bandit problem," which is a problem in which limited resources are allocated to several choices (arms).

[0037] In this existing algorithm, (1) the number of successes and failures is recorded for each arm (both initial values ​​are 0). (2) A random number following a Beta distribution (number of successes + 1, number of failures + 1) is generated for each arm, and the arm that gives the maximum value is designated as the optimal arm. (3) With probability β (usually 1 / 2), the optimal arm is selected, and with probability 1-β, another arm is selected using the method in (2). (4) The selected arm is tested, and if a result is obtained, (1) is performed. This existing algorithm assumes that the estimated reward of each arm is completely unrelated.

[0038] The learning unit 28 of this embodiment improves upon existing algorithms that, when a result is obtained for a given arm, also give discounted rewards to its neighboring arms, thereby creating a correlation between the success / failure values ​​of neighboring arms. For example, if an arm succeeds, not only is its success count increased by +1, but the success counts of its two adjacent arms are also increased by +1 / 2 each. The weights given to adjacent or neighboring arms can be of any functional form that decreases with distance, for example, a Gaussian function (normal distribution) may be used. The variance of the Gaussian function is determined according to the steepness of the change in the reward function to be approximated. This method requires fewer trials to estimate the optimal arm compared to existing algorithms when the unknown reward function changes smoothly between adjacent arms, and is computationally simpler than Gaussian process-based algorithms such as GP-UCB, making it suitable as an online algorithm.

[0039] Specifically, the learning unit 28 performs reinforcement learning based on the improved algorithm described above, using the behavior history information and response information from the behavior history DB. In other words, the learning unit 28 divides the elapsed time from the time a user leaves the e-commerce site into predetermined time intervals (e.g., every 10 minutes) from 0 to 72 hours later, and assigns a reward of +1 (main reward) to the time period that includes the delivery time of the advertisement information corresponding to the open beacon (i.e., the opened advertisement information), and gives a reward of +1 / 2 (sub-reward) to the adjacent time periods. The trained model that has undergone such reinforcement learning outputs the time period with the highest cumulative reward value as the optimized delivery timing.

[0040] Furthermore, if response information cannot be obtained within a certain period of time after ad delivery, it is determined to be a failure, and the failure is counted as a reward for the arm during the time period in which the ad delivery time was included (and, as mentioned above, also discounted for nearby arms). The predetermined time from ad delivery to determine a failure (failure determination time) is set as a parameter different from the upper limit of the delivery time period (72 hours in this embodiment). Failures are also used in learning in the form of failure rewards, and are utilized in a way that prevents the selection of arms that have failed.

[0041] Furthermore, the learning unit 28 performs the above-mentioned reinforcement learning each time it acquires reaction information from the reaction information acquisition unit 27, when a predetermined number of reaction information has been accumulated, or at predetermined intervals, and updates the trained model DB 36.

[0042] Furthermore, the settings for the arms in reinforcement learning are not limited to those described above. The upper limit of the elapsed time from the time of departure from the e-commerce site, as well as the number and width of the time zones to be divided, can be set as appropriate. Also, the way sub-rewards are given in success and failure counts is not limited to the time zones adjacent to the time zone of the main reward. A first sub-reward may be given to the time zones adjacent to both of them, and a second sub-reward may be given to the time zone two steps away from the main reward time zone. In addition, a third sub-reward may be given to the arm adjacent to the arm that received the second sub-reward, and a fourth sub-reward to the arm next to that, and so on, with multiple sub-rewards being given to neighboring arms (including the arms adjacent to both) centered around the arm that received the main reward. Moreover, in this embodiment, the parameter being optimized is a one-dimensional parameter, the elapsed time from the time of departure. However, the method of giving discounted sub-rewards to neighboring arms used in the optimization method of this disclosure can also be applied to the simultaneous optimization of multiple parameters by considering an appropriate discount function that decays with distance or number from the arm that received the main reward. For example, in addition to optimizing the time of day for delivery, it is also possible to optimize the length of the retargeting message as a parameter.

[0043] Next, we will explain each DB in the memory unit 15.

[0044] The site database 31 in the memory unit 15 stores e-commerce site information managed by the e-commerce site management unit 21. This e-commerce site information includes multiple types of web pages, such as the top page, category-specific product introduction pages, individual product pages, purchase pages, and member pages, with multiple page entries for each type. Page information includes details such as page design and URL. The e-commerce site information also includes other necessary site information, such as content information including images and videos.

[0045] User DB32 stores user information about users. This user information includes the user's identification number (User ID), user attribute information (age, gender, address, contact information, clothing size, etc.), user preference information, purchase history information (purchase date and time, purchased items, purchase price, etc.), etc. User attribute information is obtained through input via the user's terminal. User preference information is obtained from responses to questionnaires and trend analysis based on purchase history information. In addition, user information is linked to behavioral history information, viewed product information, related product information, response information, etc.

[0046] Product DB33 stores product information related to products handled on the e-commerce site. Product information includes the product identification number (product ID), product name, product category, product description, price, inventory information, etc. Product categories include multiple categories, such as categories based on target audience (e.g., men's clothing, women's clothing, children's clothing, accessories), categories based on clothing type (e.g., shirts, sweaters, pants, underwear, etc.), categories based on size (e.g., S, M, L), categories based on clothing brand, and whether or not it is a promotional item.

[0047] The activity history DB34 includes activity history information such as the pages viewed by the user, page transition information indicating which pages the user moved from and to, the start time of viewing each page, the exit time indicating the time the user left that page, the time spent on the page from the start time to the exit time, and click history information indicating where the user clicked on the viewed pages.

[0048] The related products database 35 stores related product information generated by the related products generation unit 23. This related product information is linked to the browsing product information stored in the user database 32. Furthermore, information on products set as related products is linked to the product information stored in the product database 33.

[0049] The trained model DB36 stores the aforementioned trained distribution timing model generated by the learning unit 28, as well as mathematical models such as the collaborative filtering algorithm.

[0050] The ad management server 2, configured as described above, delivers ad information to users who leave the e-commerce site after viewing product information via the user terminal 3.

[0051] Specifically, Figure 2 shows a flowchart illustrating the advertising delivery procedure in the advertising management system related to this disclosure, and the advertising delivery procedure of this embodiment will be described below in accordance with this flowchart. Specifically, the management control unit 14 of the advertising management server executes a program that performs each process in the advertising delivery procedure.

[0052] First, in step S1, the behavior history acquisition unit 22 of the management control unit 14 acquires the user's behavior history information and determines whether or not it has detected a user who has left an individual product page. If the result of this determination is true (Yes), that is, if a user who has left has been detected, the management control unit 14 proceeds to the next step S2. On the other hand, if the result of this determination is false (No), that is, if no user who has left has been detected, the management control unit 14 returns to the flowchart.

[0053] In step S2, the ad delivery unit 26 of the management control unit 14 obtains the browsing information of the departing user and determines whether the product in the browsing information (browsing product) is a product that meets predetermined delivery conditions. If the determination result is true (Yes), that is, if the product that the departing user was browsing was a product that meets the predetermined delivery conditions, the management control unit 14 proceeds to the next step S3. On the other hand, if the determination result is false (No), that is, if the product that the departing user was browsing was a product that does not meet the predetermined delivery conditions, the management control unit 14 returns to the flowchart without delivering ad information including a retargeting message.

[0054] In step S3, the ad delivery unit 26 of the management control unit 14 determines from the browsing history information of the departing user whether the viewed product is subject to return or exchange (return / exchange product). If the determination result is true (Yes), that is, if the viewed product is not a return / exchange product, the management control unit 14 proceeds to the next step S4. On the other hand, if the determination result is false (No), that is, if the viewed product is a return / exchange product, the management control unit 14 returns to the flowchart without delivering ad information including a retargeting message.

[0055] In step S4, the related product generation unit 23 of the management control unit 14 generates related product information related to the viewed product using a predetermined mathematical model (e.g., a collaborative filtering algorithm). The related product generation unit 23 stores the generated related product information in the related product DB 35.

[0056] In step S5, the distribution content setting unit 24 of the management control unit 14 sets a retargeting message as advertising information corresponding to the viewed product. Here, the distribution content setting unit 24 is configured to also include related product information in the advertising information.

[0057] In step S6, the delivery timing setting unit 25 of the management control unit 14 sets the delivery timing of the advertising information using a trained model that calculates the optimized delivery timing.

[0058] In step S7, the advertising distribution unit 26 of the management control unit 14 delivers advertising information, as set by the distribution content setting unit 24 in step S5, to lapsed users at the distribution timing set by the distribution timing setting unit 25 in step S6, and returns the flowchart.

[0059] The management control unit 14 delivers advertising information, including retargeting messages, to lapsed users in this manner. When a lapsed user opens an email containing advertising information, the response information acquisition unit 27 acquires the opening information (response information) and stores it in the behavior history database. Furthermore, the learning unit 28 performs further reinforcement learning based on the opening information stored in the behavior history database and updates the learned model for delivery timing.

[0060] As described above, the advertising management system 1 in this embodiment uses a pre-trained delivery timing model to deliver advertising information, including retargeting messages to churned users, at an optimized delivery timing based on the user's behavior history information, thereby more reliably delivering (getting seen by) churned users and improving the final results (so-called conversions) on the e-commerce site. In other words, the advertising management system 1 (advertising management server 2, advertising management method, advertising management program) can further optimize the delivery of advertisements to users.

[0061] According to the advertising management system 1, by using a trained model generated through reinforcement learning of the time from ad delivery to opening for multiple past churned users, the timing of ad delivery can be optimized more efficiently and accurately than with conventional learning processes.

[0062] Furthermore, according to the advertising management system 1, the trained model is subjected to reinforcement learning, where each time period, which is divided into predetermined time intervals from the time a user leaves the website, is used as an arm. A main reward is given to the time period that includes the time the opened ad information was delivered, and a sub-reward smaller than the main reward is given to time periods adjacent to that time period. By determining the time period with the highest cumulative reward as the optimized delivery timing, the delivery timing can be optimized more reliably, efficiently, and with higher accuracy than conventional learning processes.

[0063] According to the advertising management system 1, by including related product information associated with the products viewed by users who have left the site in the advertising information, it is possible to promote the purchase of related products.

[0064] According to the advertising management system 1, advertising information is delivered only to users who have churned off for a specific product. In this way, instead of delivering advertising information indiscriminately, it is possible to deliver advertising information only to specific products, enabling more efficient advertising delivery that focuses on products that you particularly want to promote or products for which retargeting messages are particularly effective.

[0065] Furthermore, according to the advertising management system 1, if a user who has abandoned a website has a history of returning or exchanging a product they viewed, the advertising information will not be delivered to that user. Delivering advertising information for products that were the subject of a return or exchange may offend the user and potentially reduce the effectiveness of the advertisement. Therefore, by checking such behavioral history and prohibiting the delivery of advertising information, the reduction in advertising effectiveness can be mitigated.

[0066] According to the advertising management system 1, it is possible to set different advertising information depending on the product. Rather than delivering uniform advertising information to users who have left the site, advertising effectiveness can be improved by setting the content of advertising information according to the product category, such as delivering content for men's products and content for women's products.

[0067] This concludes the description of each embodiment of the present disclosure, but the embodiments of the present disclosure are not limited to these embodiments.

[0068] In the advertising management system 1 of the above embodiment, advertising information including retargeting messages is delivered via email, but the medium of delivery is not limited to email. For example, the medium of delivery may be other messaging services such as SNS (Social Networking Service) or SMS (Short Message Service). In the case of SNS and SMS as well, "opened" means that the user has become able to view the retargeting message, or in other words, that it has been read.

[0069] Furthermore, in the above embodiment of the advertising management system 1, the advertising management server 2 also manages the e-commerce site in the e-commerce site management unit 21 of the management control unit 14, but the configuration may also be such that the management of the e-commerce site is performed on a separate server.

[0070] Furthermore, the advertising management system 1 of the above embodiment delivers advertising information (so-called browser abandonment emails) including retargeting messages to users who view a product page but leave the site without adding it to their cart, which is known as browser abandonment. However, the types of retargeting messages are not limited to these. For example, this disclosure may be applied to (1) advertising information including a retargeting message suggesting the purchase of an item added to an e-commerce cart (so-called abandoned cart email) to users who have left the e-commerce site without purchasing the item they added to their cart (added item); (2) advertising information including a retargeting message notifying users when an added item is discounted (so-called discount notification email); (3) advertising information including a retargeting message notifying users that an added item has low stock (so-called low stock notification email); (4) advertising information including a retargeting message notifying users when an added item is out of stock and restocked (so-called restock notification email); (5) advertising information including a retargeting message requesting a review of a product purchased by the user (so-called review request email); and (6) advertising information including a retargeting message suggesting that user terminal 3 repurchase a product purchased by user terminal 3 (so-called repurchase recommendation email).

[0071] Finally, the configuration of ad management server 2 and ad management system 1 is illustrated below. [1] An advertising management device that delivers advertising information to users who leave a website after viewing product information, A behavioral history storage unit stores behavioral history information, which is information about the behavioral history of users who visited the aforementioned website. A distribution content setting unit identifies the products viewed by users who left the website from the behavior history storage unit and includes a retargeting message in the advertising information that suggests the purchase of those products. A delivery timing setting unit sets the delivery timing of the advertisement information to the lapsed user by using a mathematical model that searches for delivery timings with high open rates for advertisement information delivered to past lapsed users based on the behavioral history information stored in the behavioral history storage unit and outputs the optimal delivery timing. An advertising distribution unit delivers advertising information set by the distribution content setting unit to the churned user at the distribution timing set by the distribution timing setting unit. An advertising management device equipped with the following features. [2] The advertising management device described in [1] above, wherein the delivery timing setting unit sets the delivery timing of the advertising information using a trained model generated by reinforcement learning of the open rates for each time period from when a user leaves the website until the delivery of the advertising information for multiple past users who have left the website. [3] The trained model uses each time interval, determined by dividing the elapsed time since a user left the website into predetermined time intervals, as its arms. It performs reinforcement learning by giving a main reward during the time interval that includes the time the opened advertisement information was delivered, and a smaller sub-reward than the main reward during the time intervals adjacent to that interval. The time interval with the highest cumulative reward is then considered the optimized delivery timing. The advertising management device described in [2] above. [4] Furthermore, the website includes a related product storage unit that stores related product information, which is information about products related to the viewed product among the products handled on the website. The advertising management device according to any one of [1] to [3] above, wherein the distribution content setting unit obtains related product information related to the viewed product from the related product storage unit and includes the related product information as advertising information. [5] The advertising distribution unit distributes the advertising information to departing users who have viewed predetermined products as determined in advance from the behavior history storage unit, as described in any one of the above [1] to [4]. [6] The advertising distribution unit, if it has a history of returning or exchanging products viewed by the user who left the site, does not distribute the advertising information to the user who left the site, according to any one of the above [1] to [5]. [7] The aforementioned content distribution setting unit is an advertising management device according to any one of the above [1] to [6], which sets different advertising information depending on the product. [8] An advertising management device that delivers advertising information to users who leave a website after viewing product information, A behavioral history storage unit stores behavioral history information, which is information about the behavioral history of users who visited the aforementioned website. A distribution content setting unit identifies the products viewed by users who left the website from the behavior history storage unit and includes a retargeting message in the advertising information that suggests the purchase of those products. A delivery timing setting unit sets the delivery timing of the advertisement information to the lapsed user by using a mathematical model that searches for delivery timings with high open rates for advertisement information delivered to past lapsed users based on the behavioral history information stored in the behavioral history storage unit and outputs the optimal delivery timing. An advertising distribution unit delivers advertising information set by the distribution content setting unit to the churned user at the distribution timing set by the distribution timing setting unit. An advertising management system equipped with the following features. [9] An advertising management method in which a computer delivers advertising information to users who have viewed product information on a website and then left the site, A step to set delivery content includes: using a behavior history storage unit that stores behavior history information, which is information about the behavior history of users who visited the website, to identify the products viewed by users who left the website, and to include a retargeting message in the advertising information that suggests purchasing those products; A delivery timing setting step involves setting the delivery timing of the advertisement information to the lapsed user by using a mathematical model that searches for delivery timings with high open rates for advertisement information delivered to past lapsed users based on the behavioral history information stored in the behavioral history storage unit and outputs the optimal delivery timing, and An ad delivery step which delivers the ad information set in the delivery content setting step to the churned user at the delivery timing set in the delivery timing setting step, A computer-based method of managing advertisements.

[10] An advertising management program that causes a computer to deliver advertising information to users who have viewed product information on a website and then left the site, A step to set delivery content includes: using a behavior history storage unit that stores behavior history information, which is information about the behavior history of users who visited the website, to identify the products viewed by users who left the website, and to include a retargeting message in the advertising information that suggests purchasing those products; A delivery timing setting step involves setting the delivery timing of the advertisement information to the lapsed user by using a mathematical model that searches for delivery timings with high open rates for advertisement information delivered to past lapsed users based on the behavioral history information stored in the behavioral history storage unit and outputs the optimal delivery timing, and An ad delivery step which delivers the ad information set in the delivery content setting step to the churned user at the delivery timing set in the delivery timing setting step, An advertising management program that causes a computer to run ads. [Explanation of Symbols]

[0072] 1. Advertising Management System 2. Ad management server (ad management device) 3. User terminals 11 Input section 12 Display section 13 Communications Department 14 Management and Control 15 Storage section 21 EC Site Management Department 22. Activity History Acquisition Unit 23 Related Products Generation Department 24 Distribution Content Settings Section 25. Distribution Timing Setting Section 26 Advertising Distribution Department 27 Reaction Information Acquisition Unit 28 Learning Department 31 Site Database 32 User Database 33 Product Database 34. Behavioral History Database 35 Related Products Database 36 Pre-trained model databases N Network

Claims

1. An advertising management device that delivers advertising information to users who leave a website after viewing product information, A behavioral history storage unit stores behavioral history information, which is information about the behavioral history of users who visited the aforementioned website. A distribution content setting unit identifies the products viewed by users who left the website from the behavior history storage unit and includes a retargeting message in the advertising information that suggests the purchase of those products. A delivery timing setting unit sets the delivery timing of the advertisement information to the lapsed user by using a mathematical model that searches for delivery timings with high open rates for advertisement information delivered to past lapsed users based on the behavioral history information stored in the behavioral history storage unit and outputs the optimal delivery timing. An advertising distribution unit delivers advertising information set by the distribution content setting unit to the churned user at the distribution timing set by the distribution timing setting unit. An advertising management device equipped with the following features.

2. The delivery timing setting unit sets the delivery timing of the advertising information using a trained model, which is generated by reinforcement learning of the open rates for multiple past users from the time they leave the website until the delivery of the advertising information, as the mathematical model. The advertising management device according to claim 1.

3. The trained model uses each time interval, determined by dividing the elapsed time since a user left the website into predetermined time intervals, as its arms. It performs reinforcement learning by giving a main reward during the time interval that includes the time the opened advertisement information was delivered, and a smaller sub-reward than the main reward during the time intervals adjacent to that interval. The time interval with the highest cumulative reward is then considered the optimized delivery timing. The advertising management device according to claim 2.

4. Furthermore, the website includes a related product storage unit that stores related product information, which is information about products related to the viewed product among the products handled on the website. The distribution content setting unit obtains related product information related to the viewed product from the related product storage unit and includes the related product information as the advertising information. The advertising management device according to claim 1.

5. The advertising management device according to claim 1, wherein the advertising distribution unit distributes the advertising information to lapsed users who have viewed predetermined products from the behavior history storage unit.

6. The advertising management device according to claim 1, wherein the advertising distribution unit, if the behavioral history storage unit has behavioral history related to a return or exchange of a product viewed by the lapsed user, does not distribute the advertising information to the lapsed user.

7. The aforementioned distribution content setting unit sets different advertising information depending on the product. The advertising management device according to claim 1.

8. An advertising management device that delivers advertising information to users who leave a website after viewing product information, A behavioral history storage unit stores behavioral history information, which is information about the behavioral history of users who visited the aforementioned website. A distribution content setting unit identifies the products viewed by users who left the website from the behavior history storage unit and includes a retargeting message in the advertising information that suggests the purchase of those products. A delivery timing setting unit sets the delivery timing of the advertisement information to the lapsed user by using a mathematical model that searches for delivery timings with high open rates for advertisement information delivered to past lapsed users based on the behavioral history information stored in the behavioral history storage unit and outputs the optimal delivery timing. An advertising distribution unit delivers advertising information set by the distribution content setting unit to the churned user at the distribution timing set by the distribution timing setting unit. An advertising management system equipped with the following features.

9. An advertising management method in which a computer delivers advertising information to users who have viewed product information on a website and then left the site, A step to set delivery content includes: using a behavior history storage unit that stores behavior history information, which is information about the behavior history of users who visited the website, to identify the products viewed by users who left the website, and to include a retargeting message in the advertising information that suggests purchasing those products; A delivery timing setting step involves setting the delivery timing of the advertisement information to the lapsed user by using a mathematical model that searches for delivery timings with high open rates for advertisement information delivered to past lapsed users based on the behavioral history information stored in the behavioral history storage unit and outputs the optimal delivery timing, and An ad delivery step which delivers the ad information set in the delivery content setting step to the churned user at the delivery timing set in the delivery timing setting step, A computer-based method for managing advertisements.

10. An advertising management program that causes a computer to deliver advertising information to users who have viewed product information on a website and then left the site, A step to set delivery content includes: using a behavior history storage unit that stores behavior history information, which is information about the behavior history of users who visited the website, to identify the products viewed by users who left the website, and to include a retargeting message in the advertising information that suggests purchasing those products; A delivery timing setting step involves setting the delivery timing of the advertisement information to the lapsed user by using a mathematical model that searches for delivery timings with high open rates for advertisement information delivered to past lapsed users based on the behavioral history information stored in the behavioral history storage unit and outputs the optimal delivery timing, and An ad delivery step which delivers the ad information set in the delivery content setting step to the churned user at the delivery timing set in the delivery timing setting step, An advertising management program that causes a computer to run ads.

Images