JP7790779B2 - Marketing optimization device, system, method and program - Google Patents

Description

This disclosure relates to a marketing optimization device, a marketing optimization system, a marketing optimization method, and a marketing optimization program that optimize marketing strategies.

In order to increase the effectiveness of advertising in web marketing, it is necessary to use the appropriate media and the appropriate type of advertising to effectively target users.

For example, Patent Document 1 describes an advertising distribution device that distributes advertisements to increase brand recognition among more viewers. The advertising distribution device described in Patent Document 1 calculates the expected number of times multiple viewers will view an advertisement and the number of viewers whose expected value is equal to or greater than a threshold, and determines the number of advertisements to distribute to multiple programs within a specified budget so as to increase the number of viewers whose expected value is equal to or greater than the threshold.

Japanese Patent Application Laid-Open No. 2021-9558

The advertising distribution device described in Patent Document 1 aims to efficiently distribute advertisements by focusing on the number of views, but does not take into consideration the target audience for brand awareness. As a result, it is difficult to say that the advertising distribution device described in Patent Document 1 is necessarily able to market to the target audience.

Therefore, the present disclosure aims to provide a marketing optimization device, marketing optimization system, marketing optimization method, and marketing optimization program that can derive the optimal marketing measures to be taken for a target.

The marketing optimization device disclosed herein is characterized by comprising: an input means for accepting input of numerical information associating conditions, effects, and costs for marketing actions, the numerical information being generated based on performance data that associates the conditions for those actions with the effects achieved by those actions and the costs incurred for those actions; an effect suppression designation means for accepting designation of actions to suppress the influence of those effects; a coefficient calculation means for performing a reduction process to reduce the value indicating the effect corresponding to the designated action among the numerical information, and calculating a first coefficient, which is the coefficient of a cost function indicating the effect of the action, and a second coefficient, which is the coefficient of a cost function indicating the cost of the action, using both the numerical information that has been reduced and the numerical information that has not been reduced; an objective function definition means for using the first and second coefficients to define an energy function as an objective function that specifies a relationship in which the value decreases as the effect increases and the value increases as the cost increases, depending on the combination of conditions; an optimization process execution means for transmitting the defined objective function to a quantum computer and causing it to execute an optimization process to minimize the value of the objective function; and an output means for outputting the results of the optimization process.

The marketing optimization system disclosed herein comprises a quantum computer that performs optimization processing of a transmitted objective function, and a marketing optimization device connected to the quantum computer. The marketing optimization device is characterized by including: input means for accepting input of numerical information that associates conditions, effects, and costs for marketing actions, the numerical information being generated based on performance data that associates the conditions for those actions with the effects achieved by those actions and the costs incurred for those actions; effect suppression designation means for accepting designation of actions to suppress the impact of those effects; coefficient calculation means for performing a reduction process to reduce the value indicating the effect corresponding to the designated action among the numerical information, and calculating a first coefficient, which is the coefficient of a cost function indicating the effect of the action, and a second coefficient, which is the coefficient of a cost function indicating the cost of the action, using both the numerical information that has been reduced and the numerical information that has not been reduced; objective function definition means for using the first and second coefficients to define as the objective function an energy function that specifies a relationship in which the value decreases as the effect increases and the value increases as the cost increases, depending on the combination of conditions; optimization process execution means for transmitting the defined objective function to the quantum computer and causing it to perform optimization processing to minimize the value of the objective function; and output means for outputting the results of the optimization process.

The marketing optimization method disclosed herein accepts input of numerical information associating conditions, effects, and costs for marketing actions, which is generated based on performance data that associates the conditions for those actions with the effects achieved by those actions and the costs incurred for those actions; accepts the specification of an action to suppress the impact of the effects; performs a reduction process to reduce the value of the numerical information that indicates the effect corresponding to the specified action; calculates a first coefficient, which is the coefficient of a cost function indicating the effect of the action, and a second coefficient, which is the coefficient of a cost function indicating the cost of the action, using both the numerical information that has been reduced and the numerical information that has not been reduced; defines an energy function as an objective function using the first and second coefficients, which specifies a relationship in which the value decreases as the effect increases and the value increases as the cost increases, depending on the combination of conditions; transmits the defined objective function to a quantum computer, causing it to perform an optimization process that minimizes the value of the objective function; and outputs the results of the optimization process.

The marketing optimization program disclosed herein is characterized by having a computer execute an input process that accepts input of numerical information that associates conditions, effects, and costs for marketing actions, the numerical information being generated based on performance data that associates the conditions for those actions with the effects achieved by those actions and the costs incurred for those actions; an effect suppression designation process that accepts the designation of an action to suppress the impact of an effect; a coefficient calculation process that performs a reduction process to reduce the value of the numerical information that indicates the effect corresponding to the designated action, and calculates a first coefficient, which is the coefficient of a cost function that indicates the effect of the action, and a second coefficient, which is the coefficient of a cost function that indicates the cost of the action, using both the numerical information that has been reduced and the numerical information that has not been reduced; an objective function definition process that uses the first and second coefficients to define an energy function as an objective function that specifies a relationship in which the value decreases as the effect increases and the value increases as the cost increases, depending on the combination of conditions; an optimization process execution process that sends the defined objective function to a quantum computer and causes the quantum computer to execute an optimization process that minimizes the value of the objective function; and an output process that outputs the results of the optimization process.

This disclosure makes it possible to derive the optimal marketing strategies to be implemented for the target.

1 is a block diagram illustrating a configuration example of an embodiment of a marketing optimization system according to the present disclosure. FIG. 10 is an explanatory diagram showing an example of performance data. FIG. 10 is an explanatory diagram showing an example of defined conditions. FIG. 10 is an explanatory diagram showing an example of a screen for accepting designation of conditions for a target advertisement; FIG. 10 is an explanatory diagram showing an example of an energy function. 10 is a flowchart illustrating an example of the operation of the marketing optimization system. 1 is a block diagram illustrating an overview of a marketing optimization device according to the present disclosure. FIG. 1 is a block diagram illustrating an overview of a marketing optimization system according to the present disclosure.

Embodiments of the present disclosure are described below with reference to the drawings.

Figure 1 is a block diagram showing an example configuration of one embodiment of the marketing optimization system disclosed herein. The marketing optimization system 100 of this embodiment includes a marketing optimization device 10 and a quantum computer 20. The marketing optimization device 10 is interconnected with the quantum computer 20 via a communication line.

The marketing optimization system 100 is a system that derives the optimal action to be taken for a target (hereinafter sometimes referred to as a target). Here, an action is an activity carried out for a target, such as advertising or market research. In this embodiment, a specific example of an action is an advertisement to a target.

In other words, the marketing optimization system 100 of this embodiment aims to derive the optimal combination of targets to which advertising should be delivered (hereinafter sometimes referred to as advertising targets) and the type of advertising to be used when delivering advertising to those advertising targets (hereinafter sometimes referred to as advertising types). This is because deriving the optimal combination can prevent ineffective and wasteful advertising. Hereinafter, each element included in the combination may be referred to as a condition.

Advertising types include listing ads, display ads, and email ads (email newsletters). These ad types are further subdivided depending on the advertising media in which they are displayed. Ad targets are also classified according to criteria such as gender, age, hobbies/interests, family composition, and place of residence. Note that the above-mentioned classifications of ad types and ad targets are examples, and other types of ads and classifications may also be used.

In addition, the marketing optimization system 100 of this embodiment uses a quantum computer 20 to select the optimal combination of conditions with the highest cost-effectiveness from a vast number of conditions. This makes it possible to derive the optimal solution within a realistic time frame.

Quantum computer 20 is a computer that realizes parallel computation using the phenomena of quantum mechanics and is a device that performs optimization processing of an objective function transmitted from another device. For example, a quantum annealing quantum computer (hereinafter referred to as a quantum annealing machine) is a device dedicated to determining the ground state of the Hamiltonian of an Ising model and is a device that performs annealing based on the Ising model. More specifically, a quantum annealing machine is a device that probabilistically determines the value of a binary variable that minimizes or maximizes the objective function (i.e., the Hamiltonian) of an Ising model that takes binary variables as arguments.

The quantum computer 20 of this embodiment may take any form. The quantum computer 20 may be configured with any hardware that probabilistically determines the value of a binary variable that minimizes or maximizes an objective function that takes binary variables as arguments. The quantum computer 20 may be, for example, a non-von Neumann computer in which the objective function is implemented in hardware in the form of an Ising model. Furthermore, the quantum computer 20 may be a quantum annealing machine or a general annealing machine.

The marketing optimization device 10 includes a memory unit 11, a numerical information generation unit 12, an input unit 13, an effect suppression designation unit 14, a coefficient calculation unit 15, an objective function definition unit 16, an optimization process execution unit 17, and an output unit 18.

The memory unit 11 stores various information used for processing by the marketing optimization device 10. In this embodiment, the memory unit 11 stores data (hereinafter referred to as performance data) that associates, for each advertisement that has been made in the past, conditions including the advertisement type and the advertising target targeted by that advertisement, the cost incurred by that advertisement, and the effects achieved by that advertisement.

The memory unit 11 may also store information on measures that set incentives for target actions (advertisements) (hereinafter, sometimes referred to as incentive measures). The memory unit 11 may be realized, for example, by a magnetic disk or the like.

Figure 2 is an explanatory diagram showing an example of performance data and an incentive measure list. Figure 2 shows an example of performance data, which is a list of advertisements posted on the web (web advertising list).

The example of a web ad list shown in Figure 2 indicates that display ads and interest-based ads using multiple media were selected as the ad type, and that multiple ads have been run in the past. The example shown in Figure 2 also indicates that gender, interest range, etc. were used as ad targeting criteria. Furthermore, the example shown in Figure 2 indicates that the number of impressions, number of sessions, etc. are tallied as effects.

In this embodiment, we will explain the case where advertisements made for one project and advertising type are treated as one advertisement. In the example shown in Figure 2, a combination of each campaign and each advertising type is considered to be one advertisement. For example, in a web advertising list, three items with the campaign name "Light Mountain Climbing Stamp Rally" and the advertising type "Display Advertising" correspond to one advertisement. Note that one advertisement is sometimes referred to as one move. Therefore, the total value of the costs and effects of each move in each media is treated as the cost and effect of each advertisement.

Note that the performance data illustrated in Figure 2 is an example, and advertising conducted by other methods may also be associated with the costs and effectiveness of that advertising.

The numerical information generation unit 12 generates numerical information that associates the conditions of the action (specifically, the conditions under which the advertisement was placed) with the effectiveness and cost of the action (specifically, the advertisement) based on the performance data. Note that this embodiment illustrates an example in which the number of sessions is used as the effect. However, the aggregate value used as the effect is not limited to the number of sessions.

The method for calculating costs and effects is also arbitrary. For example, the numerical information generation unit 12 may calculate costs by adding the cost of the advertisement itself as well as the time required to advertise, converted into costs based on hourly labor costs. Furthermore, if an incentive is applied, the numerical information generation unit 12 may also calculate costs by adding the amount of the reward.

As shown in the above process, the numerical information generation unit 12 calculates the effectiveness and cost of each advertisement.

Next, the numerical information generation unit 12 quantifies the conditions of the move (specifically, the conditions satisfied by the advertisement). Specifically, the numerical information generation unit 12 generates numerical information that quantifies each condition of the move indicated by the performance data using one bit or a string of multiple bits. Note that the conditions in the performance data that are to be quantified may be determined in advance. Figure 3 is an explanatory diagram showing an example of defined conditions. The example shown in Figure 3 shows that the conditions satisfied by the advertisement are quantified using a 50-bit binary variable, and the conditions satisfied by the advertisement are represented by one bit or a string of multiple bits.

Figure 3 illustrates two cases: one where whether a predetermined condition is met is represented by a 1 (if met) or a 0 (if not met), and one where the content of the condition is represented by multiple bit strings. In the example shown in Figure 3, the bit strings from bits 1 to 3 represent the content of one of six ad types, and each bit from bits 4 to 8 represents the use of each media with a 1 or 0. Note that the ad types from bits 1 to 3 are controlled based on constraints to prevent undefined bit strings from being selected.

The following is a specific explanation of the digitization process for the first advertisement shown in Figure 2. The first advertisement shown in Figure 2 uses a display advertisement. Among the conditions shown in Figure 3, advertisement type {3} expresses the advertisement type condition with three variables (3 bits), _x1 to _x3 . The numerical values 0 to 5 assigned to each advertisement type are converted to binary notation, and the variables are assigned in the order of ( _x3 , _x2 , _x1 ).
0 (000): Listing ads 1 (001): Display ads 2 (010): Retargeting ads 3 (011): Interest-based ads 4 (100): Dynamic ads 5 (101): Email ads (email newsletters)

According to the above definitions, x ₁ =1, x ₂ =0, and x ₃ =0 are set.

Similarly, the first advertisement shown in Figure 2 uses media 1 to 3. Among the conditions shown in Figure 3, advertising media {5} expresses the media conditions with five variables _x4 to _x8 . Each variable _xn is set to 0 (not used) or 1 (used). As shown in Figure 3, each media variable corresponds to the following media:
0th bit (= x ₄ ): Media 1
1st bit (= x ₅ ): Media 2
2nd bit (= x ₆ ): Media 3
3rd bit (= x ₇ ): Media 4
4th bit (= x ₈ ): Media 5

According to the above definitions, x ₄ =1, x ₅ =1, x ₆ =1, x ₇ =0, and x ₈ =0 are set. The same applies to other conditions. As described above, in the advertisement type, the bit strings of 6 (110) and 7 (111) are undefined, and therefore, they are controlled based on constraints so that they are not selected. In the case of a condition that allows only one of a condition to be selected, for example, a one-hot constraint may be set.

The numerical information generation unit 12 stores the generated numerical information in the memory unit 11.

The input unit 13 accepts the input of the above-mentioned numerical information. The input unit 13 may accept the input by reading the numerical information from the memory unit 11, or may accept the input of the numerical information directly from the numerical information generation unit 12.

The effect suppression designation unit 14 accepts designations to suppress the effects of actions (specifically, advertisements) during optimization, which will be described later. Specifically, the effect suppression designation unit 14 accepts designations of actions (advertisements) whose influence is to be suppressed during optimization. Note that the degree to which the effect is to be suppressed may be predetermined, and the effect suppression designation unit 14 may accept a designation of the desired degree of suppression.

For example, suppose there is performance data for Advertisement A targeting people whose hobby is skiing. However, if you were to run an advertisement in August, Advertisement A targeting people whose hobby is skiing would be less effective. In other words, if you were to run an advertisement in August, it would be preferable to be able to suppress and optimize the impact of the effects obtained from Advertisement A.

In this embodiment, the effect suppression designation unit 14 accepts designation of advertisements for which the influence of the effect should be suppressed. This enables the optimization process execution unit 17, described below, to derive the optimal marketing strategy to be implemented on the target based on the advertisement content.

The effect suppression designation unit 14 may accept designation of advertisements directly from the user. Alternatively, the relationship between the content of the advertisements and the advertisements to be suppressed may be determined in advance according to the time period and target demographic, and the effect suppression designation unit 14 may automatically determine the advertisements to be suppressed based on the content of the advertisements.

Furthermore, to improve user interpretability, the effect suppression designation unit 14 may accept the designation of conditions indicating the target of the action (advertisement). In this case, the effect suppression designation unit 14 may determine that an action that includes conditions different from the specified conditions is an action that suppresses the impact of the effect. Here, different conditions mean contradictory conditions, and the correspondence between them is defined in advance, for example, by an administrator, etc. Furthermore, different conditions can also be said to be conditions that designate non-target (non-target) advertising targets.

Figure 4 is an explanatory diagram showing an example of a screen that accepts the specification of conditions for the target advertising. Since the conditions for the target advertising are specified on the screen shown in Figure 4, the effect suppression specification unit 14 determines that an advertisement that includes conditions different from the specified conditions is an advertisement for which the influence of the effect should be suppressed.

For example, if "August" is specified as the target month, advertisements that ran in "January" or advertisements targeting people whose hobby is skiing may be inappropriate. Therefore, advertisements for which the influence of the effect should be suppressed may be determined in advance based on the specified conditions, and the effect suppression designation unit 14 may determine the advertisements for which the influence of the effect should be suppressed based on the specified conditions.

Specific examples of suppression are explained below. A first specific example is when a target month is specified. In this case, the effect suppression designation unit 14 may determine that a move that is not suitable for the specified month has been specified as a designation to suppress the influence of the effect. This makes it easier to select a combination of conditions that is suitable for the specified month. Examples of an inappropriate move include a different month or a different season.

A second specific example is a method in which performance data is categorized in advance by content type and compared with the category of the specified move. In this case, the effect suppression designation unit 14 may determine the category of the specified move and determine that non-target content (content of a different category) has been designated as a designation for suppressing the influence of the effect. This makes it easier to select similar conditions for similar content.

For example, consider a situation in which the advertising content is about "running in the city" and you want to advertise it as one of the "activity" categories. In this case, the effect suppression designation unit 14 determines that a content category other than "activity" has been designated to suppress the impact of the effect.

A third specific example is when the attributes of the target user are specified. In this case, the effect suppression designation unit 14 may determine that a move that is not suitable for the attributes of the specified user has been specified to suppress the influence of the effect. This makes it easier to select a combination of conditions that is suitable for a user with the specified attributes.

For example, consider a situation where the advertising content is "about cosmetics products as souvenirs" and is particularly recommended for women. In this case, the effect suppression designation unit 14 determines that a strategy other than "women only" has been specified to suppress the impact of the effect.

A fourth specific example is a method in which performance data is categorized in advance by the product status of the content (e.g., new, updated, revived, recognized, etc.) and compared with the product status of the content of the specified player. In this case, the effect suppression designation unit 14 may determine the product status of the content of the specified player and determine that an out-of-target situation (a different situation) has been designated as a designation to suppress the influence of the effect. This makes it easier to select similar conditions for similar situations.

For example, imagine a situation where the advertising content is new and not yet known to the public, and you want people to know about it first. In this case, the effect suppression designation unit 14 can determine that a measure other than "awareness" has been specified as the content status to suppress the impact of the effect.

The specific examples of suppression shown above may be implemented individually or in combination with multiple examples. Furthermore, various suppression specifications may be made in addition to the specific examples of suppression shown above.

For example, as a combination of the second and fourth specific examples, consider a situation in which the advertising content is "Notifying people that applications for the agricultural experience campaign that was recently advertised" and the desire is for interested parties to apply. In this case, the effect suppression designation unit 14 determines that the designation to suppress the impact of the effect is such that the content category is "experience" and the content status is a measure other than "purchase."

Based on the above numerical information, the coefficient calculation unit 15 calculates the coefficient (hereinafter referred to as the first coefficient) of a cost function indicating the effect of a move (hereinafter referred to as the first cost function) and the coefficient (hereinafter referred to as the second coefficient) of a cost function indicating the cost (hereinafter referred to as the second cost function). In this embodiment, it is assumed that the effect and cost are determined by a combination of quantified conditions. Therefore, the coefficient calculation unit 15 defines a cost function in which the conditions of the move (more specifically, the conditions of the quantified performance data) are used as explanatory variables, and the effect and cost of the move are used as objective variables, respectively.

In this embodiment, the cost function indicating the effect of a move is a cost function that formulates the effect obtained depending on the conditions satisfied by the move, and the cost function indicating the cost of a move is a cost function that formulates the cost incurred depending on the conditions satisfied by the move. These cost functions are linear models formulated based on the numerical information (specifically, performance data) described above.

For example, the cost function of the d-th effect is y _sd , and the cost of the d-th expense is function y _cd . When an advertisement is quantified using a 50-bit variable as shown in Figure 3, the cost function y _s1 of the first effect and the cost function y _c1 of the first expense can be defined by the following equations 1 and 2, respectively. Note that x _d,n in equations 1 and 2 represents the n-th variable when the data index is d.

y _s1 =q _s0 +q _s1 x _1,1 +...+q _s49,50 x _1,49 x _1,50 (Formula 1)
y _c1 =q _c0 +q _c1 x _1,1 +...+q _c49,50 x _1,49 x _1,50 (Formula 2)

When Equations 1 and 2 are expressed as vectors, they are expressed as the following examples of Equations 3 and 4. Qs in Equation 3 corresponds to the first coefficient, and Qc in Equation 4 corresponds to the second coefficient.

The cost function defined in this way corresponds to a linear model with x ₁ (vector) as an explanatory variable and y _s1 and y _c1 as response variables.

The coefficient calculation unit 15 calculates the coefficient of the first cost function (i.e., the first coefficient) and the coefficient of the second cost function (i.e., the second coefficient) defined in this manner. Here, in this embodiment, the coefficient calculation unit 15 performs a process of reducing (hereinafter referred to as a reduction process) the value indicating the effectiveness of the advertisement specified as having its effectiveness suppressed, among the above numerical information. Specifically, the coefficient calculation unit 15 performs a process of multiplying the value indicating the effectiveness of the specified advertisement by a positive number less than 1. Note that the positive number to be multiplied may be predetermined, such as 0.01.

The coefficient calculation unit 15 calculates the coefficient Qs of the first cost function and the coefficient Qc of the second cost function using both the numerical information that has undergone the reduction process and the numerical information that has not undergone the reduction process. Note that any method can be used to calculate the coefficients of the cost functions. The coefficient calculation unit 15 may calculate the coefficients using, for example, the least squares method.

Below, we will explain a specific example of a method for calculating the coefficients of the first cost function (i.e., the cost function that indicates the effect). In this embodiment, the goal is to find Qs so that Equation 5, shown below, holds as closely as possible for each piece of performance data. Note that d is the index of the performance data.

In this case, it is sufficient to calculate Qs so that the difference between the right-hand side and the left-hand side of Equation 5 is as small as possible, so the coefficient calculation unit 15 can calculate Qs as shown in the following Equation 6. The method for calculating the coefficients of the second cost function is similar.

The objective function definition unit 16 defines the objective function (Hamiltonian) used by the quantum computer 20 in the optimization process using the calculated coefficient Qs of the first cost function and the coefficient Qc of the second cost function.

In this embodiment, each condition for a move is treated as a binary value x _i ∈ {0, 1}, so as a problem for obtaining a combination of x _i that is highly effective at low cost, we consider the problem of minimizing the energy function E(x) shown in the following equation 7.

On the other hand, in order to solve the above problem using the quantum computer 20, the problem needs to be put into the form of an energy function H(x) (objective function) (QUBO (Quadratic Unconstrained Binary Optimization)) as exemplified in the following equation 8.

In this embodiment, Q = {q _ij } is an interaction term for each condition (an upper triangular matrix with all zeros in the lower left corner), and interaction terms Qs and Qc for effect (number of sessions) and cost, respectively, are considered. If the energy function for effect is H _session (x, Qs) and the energy function for cost is H _cost (x, Qc), the problem becomes one of minimizing the energy function shown in the following equation 9 (λ is a weighting coefficient).

In other words, the energy function shown in Equation 9 is an objective function defined using the first coefficient and the second coefficient, and can be said to be an energy function that specifies a relationship in which the value decreases as the effect increases and increases as the cost increases, depending on the combination of conditions. Furthermore, depending on the necessary constraints, Equation 9 may further include terms representing constraints (such as a one-hot constraint or a constraint that suppresses the output of undefined bit strings).

Figure 5 is an explanatory diagram showing an example of an energy function. The energy function shown in Figure 5 is an example of an energy function (Hamiltonian) set when the conditions shown in Figure 3 are defined. By including necessary constraints in the energy function other than the term in Equation 9 shown above, it is possible to obtain a more appropriate optimal solution.

The optimization process execution unit 17 sends the defined objective function (Hamiltonian) to the quantum computer 20, causing it to execute an optimization process to derive the optimal combination of play conditions. Note that methods for having a quantum computer execute an optimization process based on an objective function (Hamiltonian) are widely known, so a detailed explanation will be omitted here.

The output unit 18 outputs the results of the optimization process performed by the quantum computer 20. The method for outputting the results is arbitrary. For example, the output unit 18 may output the combination of advertising conditions obtained by associating the name of each condition with the combination.

The numerical information generation unit 12, input unit 13, effect suppression designation unit 14, coefficient calculation unit 15, objective function definition unit 16, optimization process execution unit 17, and output unit 18 are realized by a computer processor (e.g., a CPU (Central Processing Unit)) that operates according to a program (marketing optimization program).

For example, the program may be stored in the memory unit 11 of the marketing optimization device 10, and the processor may read the program and operate as the numerical information generation unit 12, input unit 13, effect suppression designation unit 14, coefficient calculation unit 15, objective function definition unit 16, optimization process execution unit 17, and output unit 18 in accordance with the program. Furthermore, the functions of the marketing optimization device 10 may be provided in the form of SaaS (Software as a Service).

Furthermore, the numerical information generation unit 12, input unit 13, effect suppression specification unit 14, coefficient calculation unit 15, objective function definition unit 16, optimization process execution unit 17, and output unit 18 may each be realized by dedicated hardware. Furthermore, some or all of the components of each device may be realized by general-purpose or dedicated circuits, processors, etc., or combinations of these. These may be configured by a single chip, or by multiple chips connected via a bus. Some or all of the components of each device may be realized by a combination of the above-mentioned circuits, etc., and programs.

Furthermore, when some or all of the components of the marketing optimization device 10 are realized by multiple information processing devices, circuits, etc., the multiple information processing devices, circuits, etc. may be centrally or decentralized. For example, the information processing devices, circuits, etc. may be realized as a client-server system, cloud computing system, etc., in which each device is connected via a communication network.

Next, the operation of the marketing optimization system 100 of this embodiment will be described. Figure 6 is a flowchart showing an example of the operation of the marketing optimization system 100 of this embodiment.

The input unit 13 accepts input of numerical information generated based on performance data (step S11). The effect suppression designation unit 14 accepts designation of a player whose influence is to be suppressed (step S12). The coefficient calculation unit 15 performs a reduction process to reduce the value of the numerical information indicating the effect corresponding to the designated player (step S13). The coefficient calculation unit 15 then calculates a first coefficient and a second coefficient using both the numerical information that has been reduced and the numerical information that has not been reduced (step S14). The first coefficient is a coefficient of a cost function that indicates the effect of the player, and the second coefficient is a coefficient of a cost function that indicates the cost of the player.

The objective function definition unit 16 defines an energy function defined using the first and second coefficients as the objective function (step S15). The optimization process execution unit 17 transmits the defined objective function to the quantum computer 20, causing it to execute an optimization process that minimizes the value of the objective function (step S16). The output unit 18 then outputs the results of the optimization process (step S17).

As described above, in this embodiment, the input unit 13 accepts input of numerical information, and the effect suppression designation unit 14 accepts designation of a move to suppress the influence of an effect. Furthermore, the coefficient calculation unit 15 performs a reduction process to reduce the value of the numerical information indicating the effect corresponding to the designated move, and calculates a first coefficient and a second coefficient using this numerical information. The optimization process execution unit 17 then transmits an energy function defined using the first coefficient and second coefficient to the quantum computer 20 to execute an optimization process, and the output unit 18 outputs the results of the optimization process. With this configuration, it is possible to derive the optimal marketing move to be taken against a target.

Next, an overview of the present disclosure will be described. FIG. 7 is a block diagram showing an overview of a marketing optimization device according to the present disclosure. A marketing optimization device 80 (e.g., marketing optimization device 10) according to the present disclosure includes input means 81 (e.g., input unit 13) that accepts input of numerical information that associates conditions, effects, and costs, which is generated based on performance data (e.g., data exemplified in FIG. 2 ), which is data that associates the conditions of a marketing strategy (e.g., advertising), the effects (e.g., number of sessions) achieved by that strategy, and the costs incurred by that strategy; effect suppression designation means 82 (e.g., effect suppression designation unit 14) that accepts designation of a strategy that suppresses the impact of the effect; and a reduction process (e.g., multiplying by 0.01) that reduces the value indicating the effect corresponding to the designated strategy from the numerical information, and calculates a cost that indicates the effect of the strategy using both the numerical information that has been reduced and the numerical information that has not been reduced. The computer system includes coefficient calculation means 83 (e.g., coefficient calculation unit 15) that calculates a first coefficient (e.g., the above Qs) that is a coefficient of a cost function and a second coefficient (e.g., the above Qc) that is a coefficient of a cost function that indicates the cost of a move; objective function definition means 84 (e.g., objective function definition unit 16) that uses the first coefficient and the second coefficient to define, as an objective function, an energy function (e.g., the energy function shown in Equation 9 above) that specifies a relationship in which the value decreases as the effect increases and the value increases as the cost increases depending on a combination of conditions; optimization process execution means 85 (e.g., optimization process execution unit 17) that transmits the defined objective function to a quantum computer (e.g., quantum computer 20) and causes the quantum computer to execute an optimization process that minimizes the value of the objective function; and output means 86 (e.g., output unit 18) that outputs the results of the optimization process.

This configuration allows us to determine the optimal marketing strategies to be implemented for the target audience.

In addition, the input means 81 may accept input of numerical information in which each condition of the move is quantified using a binary variable, and the objective function definition means 84 may define the objective function using QUBO (for example, the above equation 8).

In addition, the effect suppression designation means 82 may accept the designation of conditions indicating the target of the move, and may determine that a move that includes conditions different from the specified conditions is a move that will suppress the influence of the effect.

Specifically, the effect suppression designation means 82 may determine conditions that are different from the specified conditions based on the relationship between predetermined conditions.

The marketing optimization device 80 may also include a numerical information generation means (e.g., the numerical information generation unit 12) that generates numerical information that quantifies each condition of the move indicated by the performance data using one bit or a string of multiple bits. The input means 81 may then accept input of this numerical information.

Also, the cost function showing the effectiveness of a move may be represented by a first linear model with the quantified conditions as explanatory variables and the effectiveness as the objective variable, and the cost function showing the cost of a move may be represented by a second linear model with the quantified conditions as explanatory variables and the cost as the objective variable.The coefficient calculation means 83 may then calculate the first coefficient, which is the coefficient of the first linear model, and the second coefficient, which is the coefficient of the second linear model, using the least squares method.

In addition, the coefficient calculation means 83 may perform a reduction process to reduce the value indicating the effect by multiplying the value by a positive number less than 1.

Figure 8 is a block diagram showing an overview of a marketing optimization system according to the present disclosure. A marketing optimization system 90 (e.g., marketing optimization system 100) according to the present disclosure includes a quantum computer 91 (e.g., quantum computer 20) that performs optimization processing of a transmitted objective function, and a marketing optimization device 80 (e.g., marketing optimization device 10) connected to the quantum computer 91. The configuration of the marketing optimization device 80 is similar to the configuration of the marketing optimization device 80 illustrated in Figure 7.

With such a configuration, it is possible to derive the optimal marketing measures to be taken for the target.

Some or all of the above embodiments may also be described as, but are not limited to, the following notes:

(Note 1) An input means for receiving input of numerical information that associates the conditions of a marketing strategy, the effects obtained by the strategy, and the costs required for the strategy, the numerical information being generated based on performance data that is data that associates the conditions, the effects, and the costs required for the strategy;
an effect suppression designation means for accepting designation of the player for suppressing the influence of the effect;
a coefficient calculation means for performing a reduction process to reduce a value indicating the effect of a designated play from among the numerical information, and calculating a first coefficient, which is a coefficient of a cost function indicating the effect of the play, and a second coefficient, which is a coefficient of a cost function indicating the cost of the play, using both the numerical information that has been reduced and the numerical information that has not been reduced;
an objective function definition means for defining, as an objective function, an energy function that specifies a relationship in which the value decreases as the effect increases and the value increases as the cost increases, depending on the combination of the conditions, using the first coefficient and the second coefficient;
an optimization process execution means for transmitting the defined objective function to a quantum computer and causing the quantum computer to execute an optimization process for minimizing the value of the objective function;
and an output means for outputting the results of the optimization process.

(Note 2) The input means receives input of numerical information in which each condition of the move is quantified by a binary variable;
2. The marketing optimization device according to claim 1, wherein the objective function definition means defines the objective function in QUBO.

(Appendix 3) The effect suppression designation means accepts the designation of conditions indicating the target of the action, and determines that an action that includes conditions different from the designated conditions is an action that suppresses the influence of the effect. A marketing optimization device described in Appendix 1 or Appendix 2.

(Supplementary Note 4) The marketing optimization device according to Supplementary Note 3, wherein the effect suppression designation means determines whether a condition different from the designated condition is satisfied based on a relationship between predetermined conditions.

(Note 5) A numerical information generating means is provided for generating numerical information that quantifies each condition of the play indicated by the performance data using one bit or a plurality of bit strings,
5. The marketing optimization device according to claim 1, wherein the input means accepts input of the numerical information.

(Appendix 6) The cost function showing the effect of the move is expressed by a first linear model with the quantified conditions as explanatory variables and the effect as the objective variable.
The cost function showing the cost of the move is expressed by a second linear model with the quantified conditions as explanatory variables and the cost as the objective variable.
The marketing optimization device according to any one of Supplementary Note 1 to Supplementary Note 5, wherein the coefficient calculation means calculates a first coefficient, which is a coefficient of the first linear model, and a second coefficient, which is a coefficient of the second linear model, using a least squares method.

(Supplementary Note 7) The marketing optimization device described in any one of Supplementary Note 1 to Supplementary Note 6, wherein the coefficient calculation means performs a reduction process to reduce the value indicating the effect by multiplying the value by a positive number less than 1.

(Supplementary Note 8) A quantum computer that executes optimization processing of the transmitted objective function;
a marketing optimization device connected to the quantum computer;
The marketing optimization device includes:
an input means for receiving input of numerical information that associates the conditions, effects, and costs of a marketing strategy, the numerical information being generated based on performance data that is data that associates the conditions, effects, and costs of a marketing strategy;
an effect suppression designation means for accepting designation of the player for suppressing the influence of the effect;
a coefficient calculation means for performing a reduction process to reduce a value indicating the effect of a designated play from among the numerical information, and calculating a first coefficient, which is a coefficient of a cost function indicating the effect of the play, and a second coefficient, which is a coefficient of a cost function indicating the cost of the play, using both the numerical information that has been reduced and the numerical information that has not been reduced;
an objective function definition means for defining, as an objective function, an energy function that specifies a relationship in which the value decreases as the effect increases and the value increases as the cost increases, depending on the combination of the conditions, using the first coefficient and the second coefficient;
an optimization process execution means for transmitting the defined objective function to the quantum computer and causing the quantum computer to execute an optimization process that minimizes the value of the objective function;
and an output means for outputting a result of the optimization process.

(Supplementary Note 9) The input means receives input of numerical information in which each condition of the move is quantified by a binary variable;
The marketing optimization system according to claim 8, wherein the objective function definition means defines the objective function in QUBO.

(Supplementary Note 10) Accepting input of numerical information that associates the conditions, effects, and costs of a marketing strategy, which is generated based on performance data that is data that associates the conditions, effects, and costs of the strategy;
Accepting designation of the player who will suppress the influence of the effect;
A reduction process is performed to reduce a value indicating the effect of a designated move among the numerical information, and a first coefficient is a coefficient of a cost function indicating the effect of the move, and a second coefficient is a coefficient of a cost function indicating the cost of the move, using both the numerical information on which the reduction process has been performed and the numerical information on which the reduction process has not been performed;
Using the first coefficient and the second coefficient, an energy function is defined as an objective function, which defines a relationship in which the value decreases as the effect increases and the value increases as the cost increases, depending on the combination of the conditions;
transmitting the defined objective function to a quantum computer and causing the quantum computer to execute an optimization process that minimizes the value of the objective function;
and outputting a result of the optimization process.

(Appendix 11) Each condition of the move is quantified as a binary variable and numerical information is input.
11. The marketing optimization method of claim 10, wherein the objective function is defined in QUBO.

(Appendix 12) To a computer,
an input process for receiving input of numerical information that associates the conditions, effects, and costs of a marketing strategy, the numerical information being generated based on performance data that is data that associates the conditions, effects, and costs of a marketing strategy;
an effect suppression designation process for accepting designation of the player who will suppress the influence of the effect;
a coefficient calculation process for performing a reduction process to reduce a value indicating the effect corresponding to a designated play among the numerical information, and calculating a first coefficient which is a coefficient of a cost function indicating the effect of the play and a second coefficient which is a coefficient of a cost function indicating the cost of the play using both the numerical information on which the reduction process has been performed and the numerical information on which the reduction process has not been performed;
an objective function definition process that defines, as an objective function, an energy function that specifies a relationship in which the value decreases as the effect increases and the value increases as the cost increases, using the first coefficient and the second coefficient, depending on the combination of the conditions;
an optimization process execution process for transmitting the defined objective function to a quantum computer and causing the quantum computer to execute an optimization process for minimizing the value of the objective function; and
A marketing optimization program for executing an output process that outputs the results of the optimization process.

(Appendix 13) To a computer,
In the input process, each condition of the move is quantified as a binary variable and numerical information is input.
The marketing optimization program according to claim 12, wherein the objective function is defined in QUBO during the objective function definition process.

The present invention has been described above with reference to the embodiments, but the present invention is not limited to the above embodiments. Various modifications that would be understood by a person skilled in the art can be made to the configuration and details of the present invention within the scope of the present invention.

This application claims priority based on Japanese Patent Application No. 2022-181737, filed November 14, 2022, the entire disclosure of which is incorporated herein by reference.

REFERENCE SIGNS LIST 10 Marketing optimization device 11 Storage unit 12 Numerical information generation unit 13 Input unit 14 Effect suppression designation unit 15 Coefficient calculation unit 16 Objective function definition unit 17 Optimization process execution unit 18 Output unit 20 Quantum computer 100 Marketing optimization system

Claims (10)

an input means for receiving input of numerical information that associates the conditions, effects, and costs of a marketing strategy, the numerical information being generated based on performance data that is data that associates the conditions, effects, and costs of a marketing strategy;
an effect suppression designation means for accepting designation of the player for suppressing the influence of the effect;
a coefficient calculation means for performing a reduction process to reduce a value indicating the effect of a designated play from among the numerical information, and calculating a first coefficient, which is a coefficient of a cost function indicating the effect of the play, and a second coefficient, which is a coefficient of a cost function indicating the cost of the play, using both the numerical information that has been reduced and the numerical information that has not been reduced;
an objective function definition means for defining, as an objective function, an energy function that specifies a relationship in which the value decreases as the effect increases and the value increases as the cost increases, depending on the combination of the conditions, using the first coefficient and the second coefficient;
an optimization process execution means for transmitting the defined objective function to a quantum computer and causing the quantum computer to execute an optimization process for minimizing the value of the objective function;
and an output means for outputting the results of the optimization process. The input means receives input of numerical information in which each condition of the move is quantified using a binary variable;
The marketing optimization device according to claim 1 , wherein the objective function definition means defines the objective function in QUBO. The marketing optimization device according to claim 1 or 2, wherein the effect suppression designation means accepts the designation of conditions indicating the target of the action, and determines that an action that includes conditions different from the designated conditions is an action that suppresses the influence of the effect. 4. The marketing optimization device according to claim 3, wherein the effect suppression designation means determines whether or not a condition different from the designated condition is satisfied based on a relationship between predetermined conditions. Numerical information generating means for generating numerical information that quantifies each condition of the move indicated by the performance data using one bit or a string of multiple bits;
3. The marketing optimization device according to claim 1, wherein the input means accepts input of the numerical information. The cost function showing the effect of the move is expressed by a first linear model with the quantified conditions as explanatory variables and the effect as the objective variable.
The cost function showing the cost of the move is expressed by a second linear model with the quantified conditions as explanatory variables and the cost as the objective variable.
3. The marketing optimization device according to claim 1, wherein the coefficient calculation means calculates the first coefficients, which are coefficients of the first linear model, and the second coefficients, which are coefficients of the second linear model, using the least squares method. 3. The marketing optimization device according to claim 1, wherein the coefficient calculation means performs a reduction process to reduce the value indicating the effect by multiplying the value by a positive number less than 1. a quantum computer that executes an optimization process for the transmitted objective function;
a marketing optimization device connected to the quantum computer;
The marketing optimization device includes:
an input means for receiving input of numerical information that associates the conditions, effects, and costs of a marketing strategy, the numerical information being generated based on performance data that is data that associates the conditions, effects, and costs of a marketing strategy;
an effect suppression designation means for accepting designation of the player for suppressing the influence of the effect;
a coefficient calculation means for performing a reduction process to reduce a value indicating the effect of a designated play from among the numerical information, and calculating a first coefficient, which is a coefficient of a cost function indicating the effect of the play, and a second coefficient, which is a coefficient of a cost function indicating the cost of the play, using both the numerical information that has been reduced and the numerical information that has not been reduced;
an objective function definition means for defining, as an objective function, an energy function that specifies a relationship in which the value decreases as the effect increases and the value increases as the cost increases, depending on the combination of the conditions, using the first coefficient and the second coefficient;
an optimization process execution means for transmitting the defined objective function to the quantum computer and causing the quantum computer to execute an optimization process that minimizes the value of the objective function;
and an output means for outputting a result of the optimization process. Accepting input of numerical information that associates the conditions, effects, and costs of a marketing strategy, the numerical information being generated based on performance data that is data that associates the conditions, effects, and costs of a marketing strategy;
Accepting designation of the player who will suppress the influence of the effect;
A reduction process is performed to reduce a value indicating the effect of a designated move among the numerical information, and a first coefficient is a coefficient of a cost function indicating the effect of the move, and a second coefficient is a coefficient of a cost function indicating the cost of the move, using both the numerical information on which the reduction process has been performed and the numerical information on which the reduction process has not been performed;
Using the first coefficient and the second coefficient, an energy function is defined as an objective function, which defines a relationship in which the value decreases as the effect increases and the value increases as the cost increases, depending on the combination of the conditions;
transmitting the defined objective function to a quantum computer and causing the quantum computer to execute an optimization process that minimizes the value of the objective function;
and outputting a result of the optimization process. On the computer,
an input process for receiving input of numerical information that associates the conditions, effects, and costs of a marketing strategy, the numerical information being generated based on performance data that is data that associates the conditions, effects, and costs of a marketing strategy;
an effect suppression designation process for accepting designation of the player who will suppress the influence of the effect;
a coefficient calculation process for performing a reduction process to reduce a value indicating the effect corresponding to a designated play among the numerical information, and calculating a first coefficient which is a coefficient of a cost function indicating the effect of the play and a second coefficient which is a coefficient of a cost function indicating the cost of the play using both the numerical information on which the reduction process has been performed and the numerical information on which the reduction process has not been performed;
an objective function definition process that defines, as an objective function, an energy function that specifies a relationship in which the value decreases as the effect increases and the value increases as the cost increases, using the first coefficient and the second coefficient, depending on the combination of the conditions;
an optimization process execution process for transmitting the defined objective function to a quantum computer and causing the quantum computer to execute an optimization process for minimizing the value of the objective function; and
A marketing optimization program for executing an output process that outputs the results of the optimization process.