Information processing device, information processing method, and program

Abstract

An information processing device, an information processing method, and a program are provided that are capable of outputting information that contributes to the consideration or formulation of effective measures. [Solution] The method using an information processing device is for calculating information to assist in the formulation of effective actions or behaviors or measures to appeal to the target of those actions or behaviors, taking into account a specified event or situation, and includes the steps of an input unit inputting the action or behavior or the target of those actions, an input unit inputting information about measures implemented when the measures are implemented at multiple points in time based on the time of the specified event or situation, an input unit inputting the content of feedback to the implemented measures, a feedback trend calculation unit calculating the response trend of the content of feedback to the measures, and an impact degree calculation unit extracting and / or setting features of the information about the event or situation and the measures using machine learning based on the response trend, and calculating the degree of influence that the features have on the content of feedback.

Description

[Technical Field]

[0001] The present invention relates to an information processing device, an information processing method, and a program that can be applied to examining or formulating measures for a subject. [Background technology]

[0002] A direct mail creation support device that proposes optimal conditions for direct mail has been proposed (see Patent Document 1). [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Patent Publication No. 2021-184212 Summary of the Invention [Problem to be solved by the invention]

[0004] An object of the present invention is to provide an information processing device, an information processing method, and a program that can output information that contributes to the consideration or formulation of effective measures. [Means for solving the problem]

[0005] 1. Information processing method 1. An information processing method for calculating information to support the formulation of an effective action or behavior for a predetermined person, or a measure to appeal to a target of the action or behavior, taking into account a predetermined event or situation, an input unit inputting an action or behavior, or a target of an action or behavior; an input unit inputting information about the implemented measures when the measures are implemented at a plurality of time points relative to the time point of a predetermined event or situation; an input unit inputting feedback on the implemented measures; a step in which a response transition calculation unit calculates a response transition of the response content of the measure; The method includes a step in which the impact degree calculation unit extracts and / or sets features of information about the event or situation and the measures using machine learning based on the response trend, and calculates the degree of influence of the features on the response content.

[0006] The information processing method of the present invention may include a step of estimating the impact rate when a measure is implemented at a predetermined time from the time of occurrence of the same or similar event, based on the degree of impact on the impact rate.

[0007] In the above invention, the information processing method can include a step of inputting or deriving the relationship between the feature quantities. The information processing device can include an input unit or a derivation unit that inputs the relationship between the feature quantities. When calculating the degree of influence of the feature quantities on the echo content, a more accurate degree of influence can be calculated.

[0008] The information processing method of the present invention may include a step of estimating the response rate when a measure is implemented at a specified time based on the feature, the degree of influence on the response rate, and a correspondence table that matches the time from the occurrence of an event that is the same or similar to the reference event with the amount of change in the response rate.

[0009] 2. Information processing equipment The information processing device of the present invention comprises: An information processing device for calculating information to support the formulation of an effective action or behavior for a predetermined person, or a measure to appeal to a target of the action or behavior, taking into consideration a predetermined event or situation, an input unit for inputting an action or behavior, or a target of an action or behavior; an input unit for inputting information about the implemented measures when the measures are implemented at a plurality of time points based on the time point of a predetermined event or situation; an input unit for inputting feedback on the implemented measures; a response transition calculation unit for calculating a response transition of the response content of the measure; The system includes an influence degree calculation unit that uses machine learning to extract and / or set features of information about the event or situation and the measures based on the response transition, and calculates the degree to which the features affect the response content.

[0010] 3. Program The program of the present invention is A program for causing a computer to execute an information processing method for calculating information to support the formulation of an effective action or behavior for a predetermined person, or a measure to appeal to the target of the action or behavior, taking into consideration a predetermined event or situation, an input unit inputting an action or behavior, or a target of an action or behavior; an input unit inputting information about the implemented measures when the measures are implemented at a plurality of time points relative to the time point of a predetermined event or situation; an input unit inputting feedback on the implemented measures; a step in which a response transition calculation unit calculates a response transition of the response content of the measure; The impact degree calculation unit extracts and / or sets features of information about the event or situation and the measure using machine learning based on the response trend, and calculates the degree of influence that the features have on the response content.

[0011] A plurality of points in time relative to the time of a given event or circumstance includes points in time before and after the time of the given event or circumstance.

[0012] Actions or behaviors are a concept that includes all actions or behaviors, such as participation actions, purchasing actions, actions of transferring or lending, actions of borrowing, actions of using, viewing actions, actions of providing services, actions of receiving services, and actions of traveling. The subject of an action or behavior includes, for example, an object to be transferred or received (including an object to be purchased), an object to be used, an object to be lent, an object to be paid for, or an object from which a specified person receives some kind of service, and more specifically, a concept that includes goods and services, real estate, events, activities, or any kind of gathering, objects used in the provision of a service, or objects used for a service.

[0013] The degree of influence of the reaction to the reference event or situation may also be calculated. As a specific example, the degree of influence can be expressed in the form of "Regarding legal amendment A, feature quantity X had this much influence."

[0014] The degree of influence of the response to the content of the implemented measures may also be calculated. As a specific example, the degree of influence can be expressed in the form of "for measure A, feature Y had this much influence."

[0015] In this specification, machine learning is a concept that also includes deep learning. [Effects of the Invention]

[0016] According to the present invention, it is possible to realize an information processing device, an information processing method, and a program that can output information that contributes to the consideration or formulation of effective measures. [Brief explanation of the drawings]

[0017] [Figure 1] 1 shows a conceptual diagram of information processing according to an embodiment. [Figure 2] 1 shows an information processing flow according to an embodiment. [Figure 3] An example of the relationship between time points for understanding the response rate in relation to the implementation of measures and obtaining the response transition is shown below. [Figure 4] 10 is a display example showing the contribution of a feature amount. [Figure 5] The following shows a screen displaying information on the implementation of measures, an example of the output of influencing factor estimation, and an example of the display of improvement proposals. [Figure 6] This is an illustration of the model interpretation technology. [Figure 7] 1 shows an example of an information processing device. [Figure 8] An example of an information management screen is shown. [Figure 9] An example of an analysis screen is shown. [Figure 10] This is an example of a diagram that organizes candidate features and the relationships between each feature in a tree format when deciding what features to include in machine learning. DETAILED DESCRIPTION OF THE INVENTION

[0018] Preferred embodiments of the present invention will now be described with reference to the drawings.

[0019] The basic idea of ​​the information processing method according to this embodiment is to support the realization of the following steps 1 to 4, for example. Step 1: Data preparation and input Step 2: Creating a model using machine learning (feature extraction, etc.) Step 3: Calculating the degree of influence of each feature through model interpretation Step 4: Recommending the optimal direct mail strategy

[0020] "At least one step of feature extraction and setting" is one of the tasks in Step 1. By breaking down information such as the project creative list (described below) into elements (i.e., features) and applying machine learning, a mathematical model using the features as variables can be created. By looking at the results of Steps 2 and 3, you can return to Step 1 and create a more accurate model. By looking at the results of Steps 2 and 3, you can repeat the process of returning to Step 1. The concept of feature contribution includes both coefficients (so-called mechanical contribution) and human-visible, easily understood contributions (so-called human-oriented contributions, such as SHAP values). The concept of model formulas includes a first model formula showing the features and their contributions, and a second model formula using the first model formula. The second model formula is calculated by returning to Step 1, reviewing the features, repeatedly inputting and processing, and improving the accuracy of the model formula. This involves performing machine learning for model interpretation to visualize and understand the model. Visualization allows for inference and deduction steps to be used to consider improvement measures.

[0021] The modeling method for machine learning can be selected appropriately depending on the data and field. There are three ways to use machine learning. The first is to extract features and set which features to include. The second is to use machine learning to apply it to input data, calculate contributions, and use a modeling method to create a mathematical model with a model formula (3a+4d+···) that can be understood by machines. The third is machine learning (visualization) to make the data interpretable by humans. When selecting which features to include, the candidate features shown in Figure 10 and the relationships between each feature can be organized in a tree format to select the features.

[0022] The "DM" described in this specification is described as an example of a "strategy" and an "appeal strategy." The use of "DM" is not limited to "DM," but can be interpreted as being broadly applicable to "strategies" and "appeal strategies." Therefore, the use of "DM" does not mean that appeal measures that are not DM are excluded from the concepts of "measures" and "appeal measures," but rather that "DM" is used as an example, and descriptions replacing it with "measures" and "appeal measures" are also within the scope of disclosure of this specification.

[0023] As shown in FIG. 1, the information processing method according to this embodiment inputs information about an event or situation, an appeal strategy, and the response to the appeal strategy, extracts and / or sets features using machine learning, calculates the degree of influence of the features, and can calculate a prediction model or an evaluation model based on the degree of influence.

[0024] One specific example of the concept of the information processing method is shown in Figure 2. This information processing method is an example of a series of processing flows, and can include step S1 of inputting information about an event or situation, step S2 of inputting information about each appeal measure, step S3 of inputting each response rate for each appeal measure, step S4 of calculating the trend in the response rate, step S5 of extracting or setting feature amounts by machine learning, step S6 of calculating a model formula with the feature amounts as variables, step S7 of calculating the contribution of each feature amount to the response rate, and step S8 of displaying the degree of contribution of each feature amount. Machine learning can be performed in each of the feature extraction and / or setting step S5 and the model formula calculation step S6. Machine learning can also be performed in the feature contribution calculation step S7 and the feature contribution display step S8.

[0025] Figure 3 shows an example of the relationship between time points for understanding the response rate in relation to the implementation of measures and obtaining the response transition. Using the time point at which the event or situation was understood as the base, when appeal measures A, B, and C are implemented, the corresponding response rates a, b, and c are obtained. The time point at which the response rates are understood from the implementation date of the measures can be determined arbitrarily. Multiple response rates may be understood for one appeal measure at different times.

[0026] An information processing method according to an embodiment can be used to calculate information to support the formulation of an effective action or behavior for a specific person, or a measure to appeal to a target of the action or behavior, taking into account a specific event or situation. The information processing method can include a step of inputting the action or behavior or the target of the action or behavior by an input unit; a step of inputting information about an implemented measure when the measure was implemented at multiple points in time relative to the time of the specific event or situation by the input unit; a step of inputting a response content of the implemented measure by the input unit; a step of calculating a response trend of the response content of the measure by a response transition calculation unit; and a step of extracting and / or setting feature values ​​of the information about the event or situation and the measure by machine learning based on the response trend, and calculating the degree of influence of the feature values ​​on the response content. The information processing method can also include a step of inputting or deriving a relationship between feature values. The relationship between feature values ​​can be, for example, a constraint between feature values, an increase / decrease, a denominator effect, a numerator effect, or the like, as shown in FIG. 10 . It is possible to derive or set the following: the budget and the number of shipments are in a constraint relationship, the shipping method and the shipping cost are in an increase / decrease relationship, the number of shipments and the CVR are in a denominator effect relationship, and the cost and the CPA (Cost per Acquisition) are in a numerator effect relationship. By including a step of inputting or deriving the relationship between features, it is possible to calculate a more accurate degree of influence when calculating the degree of influence that the feature has on the response content, which is also useful when calculating a model formula.

[0027] (1) Information about the event or situation to be used as reference Examples of information related to events or situations that can be used as reference include timing, clients (DM issuers), customer information (clients' customers), competitive information, economic trends, social trends, political trends, and technological trends. Examples of timing include seasons, industry events, and industry peak periods.

[0028] Examples of client (DM issuer) information include industry, market share, target, price, number of employees, sales volume, sales growth rate, and capital relationships. Examples of customer information include industry, number of employees, sales volume, sales growth rate, and capital relationships. Examples of competitive information include market share, target, price, DM frequency, and DM content and design. Examples of economic trends include industry growth rate, corporate investment appetite, economic fluctuations, inflation / deflation status, and exchange rates. Examples of social trends include changes in consumer purchasing behavior and differences in values ​​between generations. Examples of political trends include changes in laws and regulations, tax reforms, changes in trade relations, government support for the industry, and government subsidies. Examples of technological trends include advances in digital transformation, data analysis technology, and AI / automation technology.

[0029] Events or circumstances can be concepts that include external factors that affect products, services, and direct mail measures, such as legal changes and industry events. The implementation of the promotional measures themselves can also be reference events or circumstances.

[0030] The information on a company's investment intentions can be taken into consideration not only with numerical data such as "the target company's total annual investment amount" but also with non-numerical data such as "references to the field in financial statements."

[0031] (2) Information on measures As a measure, it refers to taking some kind of action to make the target person aware of the product, and is a concept that includes appeal strategies, with a more specific example being direct mail (DM).

[0032] The information about the measures is at least one type selected from the group consisting of the appeal content, conversion (results), conversion path, benefits, costs incurred from creating the appeal to sending it, the date the appeal was implemented (date of sending), the number of appeals implemented (number of shipments), the specifications of the measures, the content of the measures (contents published), information about the design of the measures, and a mailing list. An appeal strategy refers to a method of communicating information that can be perceived by the target audience, and is a concept that includes direct mail, sales promotion materials that involve printing, and email.

[0033] Examples of projects, creatives, and lists are provided as information about strategies.

[0034] The definition of the project creative list managed and analyzed in "Management and Analysis of Direct Mail (DM) Data and Performance Data" is as follows:

[0035] A project is defined as information about measures other than creative lists. Examples of projects include appeal content (information about management seminars, introductions to new services, etc.), conversions (inquiries, service applications, etc.), conversion paths (website, email, fax, etc.), offers or benefits (3 months free, sample gifts, etc.), costs (costs incurred from direct mail production to mailing, etc.), mailing date (date the direct mail was mailed, etc.), and number of mailings (number of direct mails sent, etc.).

[0036] Creative refers to information about the specifications, content, and design of the mailing. Creative specifications include the type of material (envelope, postcard, OPP bag (clear envelope), etc.), number of items (3 items enclosed (in the envelope), 1 item enclosed, etc.), size / shape (A4, B5, square, cut-out, etc.). Content includes the structure, content, and amount of text. Design includes color, font, and motif.

[0037] The list may include information about the shipping destination (area, industry, addressee, etc.).

[0038] Relative priorities may be set between each piece of information. Information before and after the implementation of an appeal strategy may also be entered. Specifically, it may be possible to compare the response rate of measures taken before the implementation of an appeal strategy with the response rate of measures taken after the implementation of an appeal strategy. Qualitative information can be indexed using keywords.

[0039] Fluctuations in consumer purchasing behavior can be investigated, for example, based on data on the number of searches for a keyword over a predetermined period (for example, by month).

[0040] (3)Reaction content An example of the response content is the response rate (CVR). The response rate (CVR) can be defined as the number of conversions (CVs) per number of copies sent for each project (direct mail delivery). The response rate (CVR) is a percentage (%) calculated based on Formula 1.

[0041] (Formula 1) Number of conversions ÷ number of copies sent = CVR

[0042] (4) Features The features can be derived using various analytical techniques and can be extracted using known feature extraction methods. When extracting the features, a step of quantifying the features is carried out as necessary so that they can be incorporated into machine learning. The quantification step may be set or calculated using software.

[0043] (5) Degree of contribution The SHAP value is an example of the degree of contribution. The SHAP value indicates the degree of contribution of each feature to the predicted value. An example of the display is shown in Figure 4. The vertical axis lists the features, and the horizontal axis shows the distribution of the influence value for each feature in the input data, which in this example is the SHAP value. The content of each measure can be displayed using different colors. The SHAP value is, for example, When calculating a machine learning model based on data from 10 direct mail campaigns, In DM strategy_1, the SHAP value is 100 In DM strategy_2, the SHAP value is -90 ... In DM strategy 10, the SHAP value is 30 The values ​​for each direct mail campaign can be plotted with red and blue circles to create a graph that visualizes the magnitude of their influence.

[0044] 3. Hardware configuration FIG. 7 shows an example of the hardware configuration of an information processing device.

[0045] The information processing device 10 according to this embodiment can be realized by one or more computers. The input device 20 can be a known input means such as a keyboard, mouse, or touch-type input device, and is a concept that includes a data acquisition unit that acquires information from a hardware resource (such as a server) connected for communication (including an internet connection). The input device 20 can include an event or situation input unit 20a, a measure input unit 20b, a feedback input unit 20c that inputs feedback content, a relevance degree input unit 20d that sets the relevance degree of information, and a feature setting unit 20e that sets feature amounts. The input device 20 can be configured by means of one or more input functions.

[0046] The processing device 30 may functionally include a relevance calculation unit 30a that calculates the relevance of information, a response transition calculation unit 30b that calculates response transitions from multiple response details, a feature extraction unit 30c that extracts features, a contribution calculation unit 30d that calculates the contribution of the features, a model formula calculation unit 30e that calculates a model formula from the contribution of the features, a response result prediction unit 30f that predicts response results from measures and various information, and an effect verification unit 30g that verifies the effectiveness of measures based on their implementation details and response results and verifies the model formula. These processes may be implemented by a single computing device or by multiple computing devices. The processes may also be implemented by multiple computers. The processing device 30 may be implemented by a computing device such as a CPU. The information processing device 10 may have a storage device 50 that stores various input data and calculation data. The storage device 50 may be stored in a known storage device such as a ROM, a hard disk, or an external storage device (CD, DVD, etc.). The storage device 50 may be configured as a separate unit independent of the computing device or may be configured within the computing device. A program that causes the information processing device 10 to execute information processing can be stored in a storage device (for example, a ROM or a hard disk) included in the information processing device 10. As an example of the processing of the response result prediction unit 30f, the response result prediction unit 30f calculates an expected CVR of 0.5%, and it is found that a feature with a high degree of contribution is "benefit." Therefore, when the content of "benefit" is changed and the response result prediction unit 30f predicts again, the expected CVR becomes 0.8%. Such processing and prediction can be performed. The information processing device can include an input unit or a derivation unit that inputs the relationship between features.

[0047] The information processing device 10 may include a display device 40 that displays various processing results such as the contribution degree of the feature amount, etc. The display device 40 may be a known display such as a liquid crystal display or an organic EL display.

[0048] The information processing device 10 may be configured such that the input device 20, the processing device 30, the display device 40, and the storage device 50 are configured as a single piece of hardware, or may be configured as multiple pieces of hardware that are communicable with each other. The information processing device 10 may be configured as a single electronic computer, or may be configured as multiple electronic computers. 4. Effects and usage examples The effects and application examples of the information processing method according to this embodiment will be described below.

[0049] (1) According to this embodiment, for example, it is possible to estimate the strength of influence of the information on the DM effect, and to analyze and predict the response results of the DM.

[0050] This embodiment contributes to the following objectives, for example.

[0051] 1) Management and analysis of DM data and outcome data 2) Proposal for improving direct mail campaigns using machine learning models This embodiment has the following effects.

[0052] When implementing direct mail campaigns, it is difficult to know which target, when, what content, and how to send direct mail to maximize effectiveness.This information processing method can detect the characteristics of direct mail that will be "effective" by managing and analyzing information on the projects, creative lists, and other information from past direct mail campaigns.

[0053] After implementing a direct mail campaign, it is difficult to accurately analyze the reasons for its effectiveness when verifying its effectiveness, making it difficult to formulate appropriate improvement plans.By managing and analyzing information on the projects, creative lists, and other information from past direct mail campaigns, the reasons for its effectiveness can be detected.

[0054] (2) As shown in Figure 5, it is possible to propose improvements to direct mail campaigns using machine learning models. When implementing direct mail campaigns, it is difficult to determine which target, when, what content, and how to send direct mail to maximize effectiveness. By implementing machine learning based on the following considerations and using model interpretation technology, it is possible to identify features (≒ elements) that have a large / small impact on direct mail campaigns.

[0055] (Consideration information) Project Information Creative information List information Other internal information: sales / number of employees of the company sending the direct mail, etc. Other external information: stock prices, legal changes, weather, etc.

[0056] (3) By constructing an inference model using the extracted or set features, it is possible to do the following, for example: 1) Before implementing a campaign, you can calculate the "predicted number of conversions," "predicted CVR," and "predicted CPA" by entering campaign information. 2) By inputting information about the measures before they are implemented, improvement proposals and the expected improvement ranges resulting from the proposals can be presented. 3) After the campaign is implemented, by entering campaign information and CVR, you can analyze the cause of discrepancies between the "predicted conversions / CVR / CPA" and the actual results, and improve the machine learning model. 4) It is useful in recommending the optimal direct mail strategy. It contributes to the recommendation. 5) Because it is possible to calculate the contribution of features, it can serve as a method for calculating model formulas, a method for estimating effects, and a method for calculating support data for formulating measures to improve effects. 6) It can be used to manage and analyze data on appeals (for example, direct mail) and results data. It can propose improvements to appeals using machine learning models. By subjecting information to machine learning, it is possible to estimate the strength of the information's influence on the effectiveness of appeals. It can also be used to analyze and predict the response results of appeals.

[0057] (4) As shown in Figure 6, by inputting data into a machine learning model, it is possible to learn and output complex relationships that humans cannot grasp. This is based on the following general flow: input data into the machine learning model, use model interpretation technology to identify the basis for the model's predictions and judgments, update the machine learning model, and describe the relationships, for example, improve the specific level of contribution. It is possible to propose improvements to DM measures using machine learning models. That is, as shown in Figure 5, 1) the machine learning model trained based on the DM measure information managed by this system makes it easy to estimate factors that affect the effectiveness of DM, and 2) based on the estimated influencing factors and their impact, it becomes easy to automatically propose areas for improving the effectiveness of new DM measures.

[0058] 5. Specific Examples The number of available internet searches or information on a company's investment intentions are examples of events or situations. It is also possible to estimate the relationship between the number of available internet searches and the response rate, or the relationship between a company's investment intentions and the response rate.

[0059] (1) First Specific Example The first specific example is given below.

[0060] (Premise) This is an example of reviewing and analyzing direct mail promotions for a human resources management tool that were carried out monthly from April to June 2024. (Results of direct mail promotion) The details and effects of direct mail promotion promoting product "X" are assumed to be as follows: <DM sent in April 2024> Number of copies shipped: 10,000 Number of CVs (number of responses): 10 CVR (response rate): 0.1% <DM sent in May 2024> Number of copies shipped: 10,000 Number of CVs (number of responses): 40 CVR (response rate): 0.4% <DM sent in June 2024> Number of copies shipped: 10,000 Number of CVs (number of responses): 120 CVR (response rate): 1.2% (Factor analysis) When analyzing the factors that led to the results of the above direct mail promotion, if we investigate the "number of internet searches for related keywords," which is thought to be one of the factors, the analysis results can be shown as follows.

[0061] Assume that the number of internet searches for the related keyword "human resources management" was checked on an internet keyword search information acquisition server, and the number of internet searches for the current period was found to be as follows: ·April 2024: 10,000 times / month ·May 2024: 30,000 times / month ·June 2024: 60,000 times / month Based on the above results, and taking into account the degree of influence of other factors, a machine learning analysis was performed and the degree of influence of "Internet search count" on CVR (response rate) was calculated to be ●●●. The extent of the influence varies depending on the program used to calculate the influence. In this way, the influence can be calculated.

[0062] (2) Second specific example The information processing may be as follows: Direct mail promoting "Product X" is highly relevant to Law Revision A (implementation date: mid-July), and the number of internet searches for related keywords is ·April: 10,000 times ·May: 30,000 times ·June: 60,000 times Assume that it was known to be so. 10,000 copies of the direct mail sent out in April received 10 conversions (response rate 0.1%). 10,000 copies of the direct mail sent out in May received 40 conversions (response rate 0.4%). 10,000 copies of the direct mail were sent out in June, resulting in 120 conversions (response rate 1.2%). Let's say the result is as follows. From the above results, the output we can get is that legal reform A has an impact of XX% on the direct mail response rate for product X, and that the response rate will fluctuate by △△% as the implementation date approaches. On the other hand, the number of internet searches has an impact of □□□.

[0063] The expression "degree of impact" can have several different meanings depending on the "event" entered. For example, in the case of a legal amendment, it would calculate the degree of impact in a causal relationship, such as "The implementation of this legal amendment will increase the need for the introduction of this service, leading to further adoption." In the case of the number of internet searches, it would calculate the degree of impact in a correlation, such as "If direct mail is sent during a period when the number of searches for the related keyword 'human resources management' is high, the CVR (response rate) tends to be high."

[0064] From the first and second specific examples, it is possible to calculate some degree of influence using the number of internet searches and CVR as input information. It is also possible to calculate the degree of influence that "if direct mail is sent out during a period when the number of searches is high, the CVR (response rate) of that direct mail also tends to be high."

[0065] (3) Third Specific Example It can also be used as a way to identify the extent to which the response rate of each direct mail project has been affected by various legal amendments and the dates of those amendments.

[0066] An example of the specification when the event is a legal amendment is given below. Step 1: Set the relevance of each project's "products and services promoted in direct mail" to the "legal amendment" (e.g., very high / high / medium / low / very low). Step 2: Categorize the "products and services to be promoted in direct mail" and various "legal amendments" for each project (e.g., [Products and Services: Medical / Construction / IT...], [Legal Amendments: Laws / Cabinet Orders / Ministerial and Ministerial Ordinances / Regulations / Written Laws]). Step 3: Group the items based on their classification and relevance, and calculate the response trends by tallying the "interval between the shipping date and the revision date" and the "response rate for each item" for each group. Step 4: Using machine learning, the influence (degree of contribution) of other features is taken into account in the calculated response trends to improve the accuracy of the prediction model. As a grouping, "Legal amendments related to the Labor Standards Act" can be grouped into five categories: "1. Laws, 2. Cabinet Orders, 3. Prefectural and Ministerial Ordinances, 4. Regulations, and 5. Written Laws and Regulations related to the Labor Standards Act," and the degree of relevance can be set. For example, in relation to a product or service, "1. Laws related to the Labor Standards Act" can be set to have a "very high" degree of relevance, while "4. Regulations related to the Labor Standards Act" can be set to have a "low" degree of relevance.

[0067] In the case of a legal amendment, it is possible to calculate the degree of influence in a causal relationship, such as "The implementation of this legal amendment will increase the need for the introduction of this service, leading to further introduction." In the case of the number of internet searches, it is also possible to calculate the degree of influence in a correlation, such as "If direct mail is sent during a period when there are many searches for the related keyword "human resources management," the CVR (response rate) tends to be higher."

[0068] Legal reform A has an impact of XX, and the response rate fluctuates by □□% as the implementation date approaches. On the other hand, it would also be possible to output something like "The number of internet searches has an impact of △△." It would also be possible to output something like "In a project (DM campaign) with an impact rate of 0.3%, legal reform A has an impact of ●●●." Changes in laws and regulations can be found by referring to the government's legal amendments and regulatory information database, and the response rate can be estimated by compiling past data on the changes in response before and after the legal amendments.

[0069] The following are examples of "aggregating the response trends before and after the legal amendment." By compiling data such as the response rates for direct mail sent one year before, six months before, three months before, one month before, and one month after legal change A, the response rates for direct mail sent one year before, six months before, three months before, one month before, one month after, and three months after legal change B, and the response rates for direct mail sent one year before, six months before, three months before, and three months after legal change C, it is possible to quantitatively present predictions such as: CVR increases by about +10% compared to the previous month up until three months before the legal change, but from three months before to the day of the change it increases sharply by +80% compared to the previous month, and then drops sharply after the change date. This can be used as a method to identify "the extent to which the response rate of each direct mail project has been affected by various legal amendments and the amendment dates." A model of response trends can be constructed.

[0070] 6.Examples of management and analysis screens DM data and performance data can be managed and analyzed on the information management screen (see Figure 8) and analysis screen (see Figure 9). Information on past DM initiatives and their effects (results) can also be registered and managed. Based on the managed data, it is also possible to aggregate and analyze initiative attributes (projects, creative lists) and chronologically. Future DM initiatives can be refined based on the aggregation and analysis.

[0071] 7. Regarding the period related to response trends The following are examples of the period assumed for deriving the response transition. (1) The period leading up to the deadline (e.g., the effective date) (2) The period that has elapsed since the start date (e.g., product release date) (3) The period before and after a specified period (e.g., an industry peak period) (4) The period from the first implementation date of the measure to the most recent implementation date

[0072] The above periods (1) to (3) are set on the assumption that a specific event will occur. Specifically, period (1) is the period that is applied when considering the response trends for measures implemented before the deadline. Period (2) has a set start date, and is the period that is applied when considering the response trends for measures implemented after that start date. Period (3) is the period during which an event will occur, and is the period that is applied when considering the response trends for measures implemented before and after that event. Period (4) is the period during which no specific event will occur, and is the period that is set when, for example, you want to look purely at a measure that will last for six months or a year. In the case of the time period settings (1) to (3), the degree of impact can be expressed in the form of, for example, "Feature X had this much impact on legal reform A," or "Feature Y had this much impact on policy A." In the case of the period setting in (4), the degree of influence can be indicated in the form of, for example, "for measure A, feature Y had this degree of influence."

[0073] The period (1) will be explained using a legal amendment as an example. As a premise, since companies will need to take certain measures in response to this amendment, a preparation period of approximately two years will be set between the date of announcement and the date of enforcement. When promoting software services to support the response to this amendment through direct mail, the response is expected to change as follows depending on the implementation date of the measures. Early period: Within one month of the announcement date (approximately two years until the enforcement date): Many responses were received from companies that responded quickly, such as those with a large business impact. Mid-term: Approximately one year from the date of announcement (approximately one year from the date of enforcement): Companies that respond early have already started, and companies that respond late still have about one year until the deadline, so there is not much response. Late period: About two years from the date of announcement (only a short time until the implementation date): A large response was received due to "last-minute demand" from companies that had been slow to respond.

[0074] By inputting and calculating the response trends for each implementation date of the measures as described above, it is possible to estimate the appropriate implementation date, as well as to calculate the appropriate creative and list conditions according to the implementation date, and the degree of influence of each feature. By calculating the degree of influence of the feature amount in this way, for example, the following evaluation can be obtained. Example 1: In the previous period, increase CVR by narrowing the list of recipients to listed companies that have a high business impact and require a quick response. Example 2: In the mid-term, the number of responses will decrease, so by reducing the cost per measure and increasing the number of measures, the strategy will be to "miss out less within the same budget." Example 3: In the latter half of the year, it is predicted that there will be a rush of demand, and many competing services will also be promoting themselves, so bonuses will be offered to increase CVR.

[0075] The period (2) will be explained using the example of a new product release. The premise is that the product is novel and meets market needs, but the barriers to entry are not high and competitors will enter the market one after another. If this service is promoted using direct mail, the response is expected to change as follows depending on the implementation date of the measure. Early stage: Within one year of product release date (no competitors): Receives a lot of feedback from innovators and early adopters Mid-term: 1-3 years after product release (several competitors enter the market): As companies compete for early majority market share, the growth in response gradually slows. Late stage: More than three years have passed since the product was released (many competitors have entered the market): Many competitors are working to acquire a late majority and switch brands from other companies, and the response is sluggish.

[0076] As described above, by inputting and calculating the response trends for each implementation date of a measure, it is possible to estimate the appropriate implementation date and cost allocation, as well as to calculate the appropriate creative and list conditions according to the implementation date, and the degree of impact of each feature. By calculating the degree of influence of the feature amount in this way, for example, the following evaluation can be obtained. Example 1: In the previous period, increase CVR by narrowing the shipping destination (list) to companies that are highly interested in highly novel services. Example 2: In the mid-term, the product will be improved based on customer feedback while also being differentiated from competitors, so CVR will be increased by creating creative that highlights the differences in functionality. Example 3: In the later stages, the market matures and the products of each company become more uniform, so the strategy is to highlight the company's track record while increasing the number of touchpoints with customers through the number of measures taken, thereby increasing the number of responses.

[0077] The period (3) will be explained using the peak period of a certain industry as an example. As a premise, a certain industry has a peak season in December, and demand for a certain product aimed at that industry tends to increase in December, which is the peak season. If this service is promoted using direct mail, the response is expected to change as follows depending on the implementation date of the measure. Early period: 3 months before the peak period: This is the preparation period for the peak period and we receive a lot of feedback. Mid-term: Peak period: There is some response to replenishing products that were in short supply during the peak period. Late period: 3 months after the peak period: As the off-season approaches, the response calms down.

[0078] As described above, by inputting and calculating the response trends for each implementation date of a measure, it is possible to estimate the appropriate implementation date and cost allocation, as well as to calculate the appropriate creative and list conditions according to the implementation date, and the degree of impact of each feature. By calculating the degree of influence of the feature amount in this way, for example, the following evaluation can be obtained. Example 1: In the first half of the year, increase CVR by differentiating from competitors through special offers, etc. Example 2: In the mid-term, it is expected that there will be an urgent need to replenish products, so the creative will emphasize that short delivery times are possible. Example 3: In the later stage, allocate segments and costs according to the number and total purchase amounts expected from each company, and take different measures for each segment.

[0079] The period (4) will be explained using the example of a new measure for an existing product. The premise is that two types of measures (Measure A and Measure B) will be implemented with new appeal to a market where the product already has a certain level of recognition. When implementing measures A and B using direct mail, if each is implemented three times with a period in between, the response is expected to change as follows depending on the implementation date of the measures. First half of the year, first implementation day: Strategy A received a large number of responses, while Strategy B received a moderate response. Mid-term: Second implementation date (6 months after the first implementation date): Measure A had a moderate response, and Measure B also had a moderate response. Late period: Third implementation date (one year after the first implementation date): Strategy A received a small response, while Strategy B received a moderate response.

[0080] As described above, by inputting and calculating the change in response for each implementation date of each measure, it is possible to estimate the appropriate number of implementations for each measure, as well as to calculate conditions such as creative and lists to prevent a decrease in the number of responses for each measure, and the degree of influence of each feature. Example 1: Measure A significantly reduces the number of responses as the number of times it is implemented increases, so it is implemented only twice, and the second shipping destination (list) is narrowed down based on the characteristics of customers who responded the first time, thereby increasing CVR. Example 2: Since strategy B only slightly reduces the number of responses as the number of times it is implemented increases, the number of times it is implemented is limited to 7, and the number of responses is increased by changing only the content of the benefits.

[0081] The above-described embodiment can be modified in various ways within the scope of the present invention. [Explanation of symbols]

[0082] 10. Information processing equipment 20 Input Devices 30 Processing equipment 40 Display device 50 Storage device

Claims

1. 1. An information processing method for calculating information to support the formulation of an effective action or behavior for a predetermined person, or a measure to appeal to a target of the action or behavior, taking into account a predetermined event or situation, an input unit inputting an action or behavior, or a target of an action or behavior; an input unit inputting information about the implemented measures when the measures are implemented at a plurality of time points relative to the time point of a predetermined event or situation; an input unit inputting feedback on the implemented measures; a step of calculating a response transition of each corresponding response content when each of the measures is implemented by a response transition calculation unit; an influence degree calculation unit calculates the degree of influence of the feature quantity on the content of the response based on the response transition and taking into account information about the measure, using a mathematical model in which the feature quantity is a variable that is extracted and / or set from information about the event or situation and the measure created by machine learning.

2. An information processing device for calculating information to support the formulation of an effective action or behavior for a predetermined person, or a measure to appeal to a target of the action or behavior, taking into consideration a predetermined event or situation, an input unit for inputting an action or behavior, or a target of an action or behavior; an input unit for inputting information about the implemented measures when the measures are implemented at a plurality of time points based on the time point of a predetermined event or situation; an input unit for inputting feedback on the implemented measures; a response transition calculation unit that calculates a response transition of each corresponding response content when each of the measures is implemented; and an influence degree calculation unit that calculates the degree of influence that the feature quantity has on the content of the response based on the response transition and taking into account information about the measure, using a mathematical model in which the feature quantity is a variable that is extracted and / or set from information about the event or situation and the measure created by machine learning.

3. A program for causing a computer to execute an information processing method for calculating information to support the formulation of an effective action or behavior for a predetermined person, or a measure to appeal to the target of the action or behavior, taking into consideration a predetermined event or situation, an input unit inputting an action or behavior, or a target of an action or behavior; an input unit inputting information about the implemented measures when the measures are implemented at a plurality of time points relative to the time point of a predetermined event or situation; an input unit inputting feedback on the implemented measures; a step of calculating a response transition of each corresponding response content when each of the measures is implemented by a response transition calculation unit; a step in which an influence degree calculation unit calculates the degree of influence of the feature quantity on the content of the response based on the response transition and taking into account information about the measure, using a mathematical model in which the feature quantity is a variable that is extracted and / or set from information about the event or situation and the measure created by machine learning.

4. An information processing method for calculating information to assist in the formulation of an effective action or behavior for a predetermined person, or a strategy for appealing to a target of the action or behavior, taking into account a predetermined event or situation, comprising: an input unit inputting an action or behavior, or a target of an action or behavior; an input unit inputting information about the implemented measures when the measures are implemented at a plurality of time points relative to the time point of a predetermined event or situation; an input unit inputting feedback on the implemented measures; a step of calculating a response transition of each corresponding response content when each of the measures is implemented by a response transition calculation unit; an influence degree calculation unit calculates the degree of influence of the feature on the content of the response based on the response transition, taking into account information about the measure, using a mathematical model in which the feature is extracted or set from information about the event or situation and the measure created by machine learning as a variable.

5. An information processing device for calculating information to assist in the formulation of an effective action or behavior for a predetermined person, or a measure to appeal to a target of the action or behavior, taking into consideration a predetermined event or situation, an input unit for inputting an action or behavior, or a target of an action or behavior; an input unit for inputting information about the implemented measures when the measures are implemented at a plurality of time points based on the time point of a predetermined event or situation; an input unit for inputting feedback on the implemented measures; a response transition calculation unit that calculates a response transition of each corresponding response content when each of the measures is implemented; and an influence degree calculation unit that calculates the degree of influence of the feature quantity on the content of the response based on the response transition and taking into account information about the measure, using a mathematical model in which the feature quantity extracted or set from information about the event or situation and the measure created by machine learning is used as a variable.

6. A program for causing a computer to execute an information processing method for calculating information to assist in the formulation of an effective action or behavior for a predetermined person, or a measure to appeal to the target of the action or behavior, taking into account a predetermined event or situation, an input unit inputting an action or behavior, or a target of an action or behavior; an input unit inputting information about the implemented measures when the measures are implemented at a plurality of time points relative to the time point of a predetermined event or situation; an input unit inputting feedback on the implemented measures; a step of calculating a response transition of each corresponding response content when each of the measures is implemented by a response transition calculation unit; a step in which an influence degree calculation unit calculates the degree of influence that the feature quantity has on the content of the response based on the response transition and taking into account information about the measure, using a mathematical model in which the feature quantity extracted or set from information about the event or situation and the measure created by machine learning is used as a variable.

Images