Information processing device, information processing method, and program

Abstract

To provide technology that makes it possible to understand which doctors to target with which channels and how effective those promotions will be, and furthermore, to optimize the allocation of promotional resources. [Solution] The impact prediction unit 51 performs a process to predict the impact for a given physician for each of the 0 or more promotions for each of the 1 or more channels. The optimal promotion identification unit 52 performs a process to identify the number of promotions that will maximize the impact for each of the 1 or more channels for that given physician, based on the prediction results for the given physician by the impact prediction unit 51.

Description

Technical Field

[0001] The present invention relates to an information processing apparatus, an information processing method, and a program.

Background Art

[0002] Conventionally, as a system for supporting the business activities of MR (Medical Representatives) (medical information representatives) who belong to, for example, a pharmaceutical company and sell and convey information about the company's own pharmaceuticals to medical personnel such as doctors and pharmacists, the technology disclosed in the following Patent Document 1 is known.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] In pharmaceutical marketing, promotion to doctors is an important activity, and by promoting to doctors, it is expected to have an effect of changing the behavior such that doctors prescribe the drug. Examples of promotion include face-to-face explanations by MRs and allowing viewing of WEB media. However, it has not been grasped how much effect can be expected by promoting to which doctor through which channel, and furthermore, the optimization of promotion resource allocation has not been achieved.

[0005] The present invention has been made in view of such a situation, and an object thereof is to provide a technology that can grasp how much effect can be expected by promoting to which doctor through which channel, and furthermore, can optimize the allocation of promotion resources.

Means for Solving the Problems

[0006] To achieve the above objective, an information processing apparatus according to one aspect of the present invention is: An impact prediction means that predicts the impact for each promotional run of zero or more times for one or more types of channels for a designated physician, Based on the results of the prediction for the predetermined physician by the impact prediction means, an optimal promotion identification means identifies the number of promotions that maximize the impact for each of the one or more channels for the predetermined physician, It is equipped with.

[0007] An information processing method and program corresponding to the above-mentioned information processing apparatus according to one aspect of the present invention are also provided as an information processing method and program according to one aspect of the present invention. [Effects of the Invention]

[0008] According to the present invention, it is possible to determine which physicians to target with which channels and how effective the promotion should be, and furthermore, it is possible to provide a technology that enables the optimization of promotional resource allocation. [Brief explanation of the drawing]

[0009] [Figure 1] This figure shows an overview of the service that can be realized by an information processing system to which a server according to one embodiment of the information processing device of the present invention is applied. [Figure 2] This is a block diagram showing an example of the hardware configuration of a server according to one embodiment of the information processing apparatus of the present invention. [Figure 3] Figure 2 is a functional block diagram showing an example of the server's functional configuration. [Figure 4] This figure shows an example of an overview regarding the prediction of impact by physician and channel (expected impact is estimated using machine learning). [Figure 5]This figure shows an example of detailed training data for predicting the impact by physician and channel (estimating the expected impact using machine learning). [Figure 6] This figure shows an example of a method for breaking down facility-level sales data into individual physician-level data. [Figure 7] This figure shows an example of the assumptions used to predict the impact by physician and channel (estimating the expected impact based on subjective evaluations). [Figure 8] This figure shows an example of an overview regarding the prediction of impact by physician and channel (estimating expected impact based on subjective evaluations). [Figure 9] This figure shows an example of average sales by stage, regarding the prediction of impact by physician and channel (estimated expected impact based on subjective evaluation). [Figure 10] This figure shows an example of the probability of change and the impact at each stage, regarding the prediction of impact by physician and channel (estimated expected impact based on subjective evaluation). [Figure 11] This figure shows an example of the probability of occurrence by physician and stage, regarding the prediction of impact by physician and channel (estimating the expected impact based on subjective evaluation). [Figure 12] This figure shows an example of the impact on individual physicians, based on predictions of impact by physician and channel (estimated expected impact based on subjective evaluations). [Modes for carrying out the invention]

[0010] Embodiments of the present invention will be described below with reference to the drawings.

[0011] Figure 1 shows an overview of the service that can be realized by an information processing system to which a server according to one embodiment of the information processing device of the present invention is applied.

[0012] This service concerns the estimation of the expected impact of different promotional channels on individual physicians in pharmaceutical marketing, as well as the allocation of promotional resources to physicians. This service can first provide information on which doctor to promote through which channel and what kind of effects can be expected, and then how to allocate promotion resources. Since there is a limit to the number of MR visits possible, this service can meet the need to know the resource allocation with optimized effects.

[0013] As an approach of this service, for example, it can "predict the impact for each doctor and each channel" and then "optimize the effects considering resource constraints". Regarding the prediction (estimation) of the expected impact in the former "prediction of impact for each doctor and each channel", for example, the following two methods can be mentioned.

[0014] The first method is to predict (estimate) the expected impact through machine learning using sales data, promotion log data, and doctor attribute data. The second method is to predict (estimate) the expected impact based on the subjective evaluation of the promotion effect using questionnaire data. Details of these two methods will be described later.

[0015] In the latter "optimization of effects considering resource constraints", based on the predicted values of the impact for each doctor and each channel, optimization is carried out on how many promotions should be implemented for each doctor through which channel. Since the total number of promotion times for each channel may have an upper limit constraint, the number of times that meets the constraint is taken. Details of the latter "optimization of effects considering resource constraints" will also be described later.

[0016] In the service overview SG shown in Figure 1, this service pre-sets simulation patterns for the number of promotions, predicts the impact on each physician in those settings, and, through mathematical optimization, outputs the number of promotions for each physician that maximizes the impact, without exceeding the upper limit constraint on the total number of promotions per channel. In other words, this service allows for optimization of effectiveness while taking resource constraints into consideration.

[0017] As will be explained in more detail later, in this service, first, in step S1-1 shown in the service overview SG in Figure 1, a simulation pattern for the number of promotions is set in advance. For example, for a promotion involving visits by medical representatives (MRs) to all doctors, a simulation pattern for the number of visits is pre-set. This pattern would be one of several numbers, such as "0, 1, 2, ..." for a promotion involving web-based explanations, or one of several numbers, such as "0, 1, 2, ..." for a promotion involving visits by medical representatives (MRs). Next, predictions are made based on the set simulation patterns. Specifically, in step S1-2, predictions are made for the impact of each promotion and each type of physician.

[0018] Steps S1-1 and S1-2 correspond to the "prediction of impact by physician and channel" described above as the approach of this service.

[0019] In step S2-1, a constraint is set on the number of promotions. For example, in the case of a promotion involving visits to medical representatives (MRs), a constraint (upper limit) is set to limit the total number of visits to 1,000 or less (this number is just an example). In step S2-2, the impact by promotion frequency and by physician, as predicted in step S1-2, and the constraint on the number of promotions set in step S2-1 are input, and mathematical optimization is performed on server 1, which will be described later. In this case, mathematical optimization involves employing a mathematical optimization solver (this is just one example).

[0020] In step S2-3, the pattern for the number of promotions that maximizes impact is output. For example, for Doctor X1, the number of MR visits might be "0" and web-based explanations "2," while for Doctor X2, it might be "1" and "1" MR visit and web-based explanation. The system outputs the number of promotional activities per doctor that maximizes impact, without exceeding the upper limit constraint on the total number of activities.

[0021] Steps S2-1, S2-2, and S2-3 correspond to the aforementioned "optimization of effectiveness considering resource constraints" as the approach of this service.

[0022] This service allows you to understand which doctors to target and through which channels to expect the most effective results. Furthermore, this service allows for the optimization of promotional resource allocation.

[0023] Referring to Figure 2, the hardware configuration of server 1 according to one embodiment of the information processing device of the present invention will be described. Figure 2 is a block diagram showing an example of the hardware configuration of a server according to one embodiment of the information processing apparatus of the present invention.

[0024] Server 1 comprises a CPU (Central Processing Unit) 11, a ROM (Read Only Memory) 12, a RAM (Random Access Memory) 13, a bus 14, an input / output interface 15, an input unit 16, an output unit 17, a storage unit 18, a communication unit 19, and a drive 20.

[0025] The CPU 11 executes various processes according to the program recorded in the ROM 12 or the program loaded into the RAM 13 from the storage unit 18. RAM13 also stores data and other information necessary for the CPU11 to perform various processes.

[0026] The CPU 11, ROM 12, and RAM 13 are interconnected via a bus 14. An input / output interface 15 is also connected to this bus 14. An input / output interface 15 is connected to an input unit 16, an output unit 17, a storage unit 18, a communication unit 19, and a drive 20.

[0027] The input unit 16 consists of various hardware buttons and the like, and inputs various information according to the operator's instructions and operations. The output unit 17 is composed of a display such as an LCD and displays various images.

[0028] The memory unit 18 is composed of DRAM (Dynamic Random Access Memory) and stores various types of data. The communication unit 19 controls communication with other devices (for example, a physician information system, a pharmaceutical company system, an MR (Medical Representative) terminal, etc., and in this embodiment, an information processing device having a machine learning model KM, etc., described later) via a network including the Internet.

[0029] A drive 20 is provided as needed. A removable media 30, consisting of a magnetic disk, optical disk, magneto-optical disk, or semiconductor memory, is appropriately mounted on the drive 20. Programs read from the removable media 30 by the drive 20 are installed in the storage unit 18 as needed. The removable media 30 can also store various data stored in the storage unit 18, just like the storage unit 18.

[0030] Although not shown in the diagram, the physician information system, pharmaceutical company system, MR (Medical Representative) terminal, marketing staff terminal, and information processing device with the machine learning model KM (described later) can also have a configuration that is basically the same as the hardware configuration in Figure 2. Therefore, the explanation of the configuration of the physician information system, pharmaceutical company system, MR terminal, marketing staff terminal, and information processing device with the machine learning model KM will be omitted.

[0031] Through the collaboration of the various hardware and software components of Server 1 shown in Figure 2, Server 1 becomes capable of executing the various processes described later. As a result, the service provider of this service can provide the various services.

[0032] Next, we will explain the functional configuration of Server 1 in Figure 2, referring to Figure 3. Figure 3 is a functional block diagram showing an example of the functional configuration of the server in Figure 2.

[0033] As shown in Figure 3, the CPU 11 of server 1 functions as an impact prediction unit 51 and an optimal promotion identification unit 52. Furthermore, one area of ​​the storage unit 18 of server 1 is provided with sales information DB71, promotion information DB72, physician information DB73, impact information DB74, and optimal promotion identification information DB75.

[0034] The sales information DB71 stores, for example, sales data at the facility level, and sales data by physician based on this facility-level sales data. The promotion information DB72 stores information related to promotions, such as promotion log data for each physician. The physician information DB73 stores, for example, physician attribute data. In this embodiment, the physicians belonging to a facility and the number of patients each physician has are also stored in the physician information DB73.

[0035] The impact prediction unit 51 in the CPU 11 performs a process to predict the impact for each of the 0 or more promotional cycles for each of the 1 or more channels for a given physician. The results of the processing performed by the impact prediction unit 51 are stored in the impact information DB 74 by the impact prediction unit 51.

[0036] The optimal promotion identification unit 52 in the CPU 11 performs a process to identify the number of promotions that maximize the impact for a given physician, for each of one or more channels, based on the prediction results for that physician by the impact prediction unit 51. The results of the processing performed by the optimal promotion identification unit 52 are stored in the optimal promotion identification information DB 75 by the optimal promotion identification unit 52.

[0037] As described above, the impact prediction unit 51 and the optimal promotion identification unit 52 function to determine which physicians to target and through which channels to expect a certain level of effectiveness, as well as to optimize the allocation of promotional resources.

[0038] The impact prediction unit 51 performs a process to make a prediction for a given physician based on the results of a predetermined machine learning model KM, which is executed using past sales data for multiple physicians, promotion log data, and physician attribute data as training data. The prediction made by this impact prediction unit 51 for a given physician is related to the first method described with reference to Figure 1. While not strictly limited, the machine learning model KM and its learning unit are assumed to reside on an information processing device different from Server 1 (although they may, of course, reside on Server 1).

[0039] The impact prediction unit 51 performs a process to make predictions for a given physician based on a simulation pattern of the number of promotions set in advance for one or more types of channels. The prediction made by this impact prediction unit 51 for a given physician is related to the second method described with reference to Figure 1.

[0040] The optimal promotion identification unit 52 performs a process to identify the number of promotions that maximize the impact for a given physician, for each of the one or more channels, within the limits of not exceeding the upper limit constraint on the total number of promotions for each of the one or more channels.

[0041] Next, we will explain the overview of the predicted impact by physician and channel, referring to Figure 4. Figure 4 shows an example of an overview of the prediction of impact by physician and channel (expected impact is estimated using machine learning). The explanation in Figure 4 describes the first method for predicting (estimating) the expected impact in the "prediction of impact by physician and channel" approach described above as part of this service's strategy.

[0042] First, in the model building process MK, which shows the model construction phase, a machine learning model KM is constructed using sales data per physician, promotion log data, and physician attribute data. Next, in the model application MT, which indicates the time when the model is applied, the number of promotions and physician attribute data are applied to the machine learning model KM, and the impact prediction unit 51 of server 1 functions and outputs a sales forecast value. In other words, it is predicted how much impact will be achieved by conducting promotions for each physician.

[0043] Next, we will explain the details of the training data used to construct the machine learning model KM, referring to Figure 5. Figure 5 shows an example of detailed training data for predicting the impact by physician and channel (expected impact estimated using machine learning).

[0044] The training data GD shown in Figure 5 is created, for example, as tabular training data. In this embodiment, as shown in Figure 5, a machine learning model KM is constructed with sales as the dependent variable and physician attributes and the number of promotions as independent variables.

[0045] In the training data GD shown in Figure 5, physician attributes include, for example, physician ID, age, and medical specialty. The promotion also includes MR visits, e-detailing, and online seminars. The training data GD includes information such as, for example, that for physician ID AAA (age 61, internal medicine), one MR visit, one e-detailing session, and three web seminars were conducted in January, resulting in sales of 1 million yen. Furthermore, the training data GD also includes data for a physician with the same ID, AAA (age 61, internal medicine), showing that in February, for example, there were 0 MR visits, 0 e-detailing sessions, and 5 web seminars, resulting in sales of 900,000 yen. Furthermore, the training data GD includes information such as the following: for a physician with ID BBB (age 59, surgery), in January, there were 0 MR visits, 1 e-detailing session, and 0 web seminars, resulting in sales of 500,000 yen. Furthermore, the training data GD also includes data for a physician with the same ID, BBB (age 59, surgery), showing that in February, for example, one MR visit, two e-detailing sessions, and one web seminar resulted in sales of 650,000 yen.

[0046] While not particularly limited, let's assume that a process of taking a 3-month moving average is being performed. Furthermore, it is assumed that non-numerical data, such as medical departments, are flagged to indicate whether or not they apply. In this embodiment, the promotion channels can be rephrased as including MR visits, e-detailing, and web seminars.

[0047] Next, we will explain sales as the dependent variable in Figure 5, referring to Figure 6. Figure 6 shows an example of a method for breaking down facility-level sales data into individual physician-level data.

[0048] The sales in Figure 5, which serves as the dependent variable, is obtained by decomposing facility-level sales data into physician-level data, based on the data on the number of patients with specific diseases per physician, as shown in the sales decomposition UB in Figure 6. To illustrate with a simple example, Facility A's revenue for month B was 1.2 million yen. To break down this revenue by individual physician, the number of patients each physician (Physician X, Physician Y, Physician Z) belonging to Facility A is used. In Figure 6, Dr. X has 3 patients, Dr. Y has 2 patients, and Dr. Z has 1 patient. By allocating the 1.2 million yen in monthly sales for facility A to these patient numbers, Dr. X's monthly sales for month B are 600,000 yen, Dr. Y's for 400,000 yen, and Dr. Z's for 200,000 yen.

[0049] Next, we will explain the assumptions for predicting the impact by physician and channel, referring to Figure 7. Figure 7 shows an example of the assumptions used to predict the impact by physician and channel (estimating the expected impact based on subjective evaluations). Furthermore, the explanation in Figure 7 describes the second method for predicting (estimating) the expected impact in the "prediction of impact by physician and channel" approach described above as part of this service's strategy.

[0050] The second method is based on the AMTUL model, which is a physician's drug prescribing behavior process. Based on this, the physician's prescribing intention stage is obtained from questionnaire data, and the changes in AMTUL are converted into monetary amounts as shown in Figure 7, where KH represents the monetary conversion. Here, doctors are asked to indicate which stage each individual medication falls into. In AMTUL, "A" stands for "I didn't know / I know the name to some extent" (Aware: cognition). Furthermore, the "M" in AMTUL stands for "Memory," meaning "I understand the general product features, but I haven't prescribed it yet." Furthermore, in AMTUL, "T" stands for "Trial," meaning "prescribed only once." Furthermore, in AMTUL, "U" stands for "Prescribed in a few cases, but not on a regular basis" (Usage: Use). Furthermore, in AMTUL, "L" stands for "Regularly prescribed" (Loyal: Regular use).

[0051] The "impact of fluctuations by stage" in the monetary conversion KH is as follows: In other words, if the stage is "A" in AMTUL, and the stage moves up one level from "A" to "M", the expected impact will be 100 yen. Furthermore, if the stage is "M" in AMTUL, and the stage moves up one level from "M" to "T", the expected impact will be 1,000 yen. Furthermore, if the stage is "T" in AMTUL, and the stage moves up one level from "T" to "U", the expected impact would be 30,000 yen. Furthermore, if the stage is "U" in AMTUL, and the stage moves up one level from "U" to "L", the expected impact would be 100,000 yen.

[0052] On the other hand, if the stage is "L" in AMTUL, and the stage drops one level from "L" to "U", the expected impact would be 50,000 yen. Furthermore, if the stage is "U" in AMTUL, and the stage drops one level from "U" to "T", the expected impact would be 5,000 yen. Furthermore, if the stage is "T" in AMTUL, and the stage drops one level from "T" to "M", the expected impact would be 1,000 yen. Furthermore, if the stage is "M" in AMTUL, and the stage drops one level from "M" to "A", the expected impact would be 800 yen.

[0053] Next, we will explain the overview of the predicted impact by physician and channel, referring to Figure 8. Figure 8 shows an example of an overview of the prediction of impact by physician and channel (expected impact estimated based on subjective evaluation). The explanation in Figure 8 is also related to the second method described above. The impact prediction unit 51 of server 1 functions to predict the impact by physician and by channel.

[0054] The second method involves using survey data to output expected impact based on subjective evaluations of promotional effectiveness. In other words, in order to output the expected impact, as shown in the Impact Overview IG in Figure 8, the following processes are executed in order: "Calculation of average sales by AMTUL stage", "Calculation of AMTUL fluctuation probability and impact", "Calculation of AMTUL existence probability by physician", and "Calculation of impact by physician".

[0055] In the "Calculation of Average Sales by AMTUL Stage" described above, the percentage of doctors in each AMTUL stage who prescribe the drug is predetermined, the total number of prescribing doctors is calculated, and the total sales of the drug are divided by the total number of prescribing doctors to calculate the average sales of doctors in each AMTUL stage. Furthermore, in the "Calculation of AMTUL Fluctuation Probability and Impact," the probability of increase and decrease between each stage of AMTUL is calculated from the survey, and the difference in sales when AMTUL progresses is used as the impact when it increases. Furthermore, in the "Calculation of the Probability of AMTUL Presence by Physician," the probability of AMTUL presence for each physician is calculated from the AMTUL distribution of physicians who responded to the survey. Furthermore, in the "Calculation of Impact per Physician," the expected impact is calculated by adding up the probability of AMTUL presence and the impact when it increases for each physician.

[0056] Next, referring to Figure 9, we will explain the average sales for each stage in relation to the prediction of impact by physician and channel. Figure 9 shows an example of average sales by stage regarding the prediction of impact by physician and channel (expected impact estimated based on subjective evaluation). The explanation in Figure 9 is also related to the second method described above.

[0057] The average sales HU shown in Figure 9 is an example of average sales by stage. Figure 9 corresponds to "Calculation of average sales by AMTUL stage" in Figure 8. Average Sales HU indicates that, using survey data, the percentage of physicians in each stage of AMTUL prescribes the drug is predetermined, the total number of prescribing physicians is calculated, and the total sales of the drug are divided by the total number of prescribing physicians to calculate the average sales of physicians in each stage of AMTUL. In the example in Figure 9, the average revenue per physician at each stage of AMTUL is shown, assuming total sales of 12 billion yen and 100 physicians. In the example shown in Figure 9, the prescription rate is set manually, but it is not limited to this and can also be set automatically.

[0058] Next, referring to Figure 10, we will explain the prediction of impact by physician and channel, including the probability of fluctuation at each stage and the impact during fluctuation. Figure 10 shows an example of the probability of change and the impact at each stage regarding the prediction of impact by physician and channel (estimated expected impact based on subjective evaluation). The explanation for Figure 10 is also related to the second method described above.

[0059] The fluctuation impact HI shown in Figure 10 is an example of fluctuation probability and fluctuation impact for each stage. Figure 10 corresponds to "Calculation of AMTUL fluctuation probability and impact" in Figure 8. The Impact HI during fluctuations uses survey data to calculate the probability of increases and decreases between each stage of AMTUL, and the difference in sales when AMTUL progresses is defined as the impact during an increase. Regarding the survey data on AMTUL, examples of questions include, "Please select your prescribing intention before receiving the promotion," and "Please select your prescribing intention after receiving the promotion."

[0060] The "probability of fluctuation by stage" in the fluctuation impact HI is as follows: In other words, if the stage is "A" in AMTUL, and the stage moves up one level from "A" to "M", the probability of this change is 80%. Furthermore, if the stage is "M" in AMTUL, and the stage moves up one level from "M" to "T", the probability of this change is 60%. Furthermore, if the stage is "T" in AMTUL, and the stage moves up one level from "T" to "U", the probability of change is 10%. Furthermore, if the stage is "U" in AMTUL, and the stage moves up one level from "U" to "L", the probability of this change is 3%. Note that the "impact during fluctuations by stage" in the fluctuation impact HI is the same as in Figure 7.

[0061] Next, referring to Figure 11, we will explain the probability of occurrence by physician and stage in relation to the prediction of impact by physician and channel. Figure 11 shows an example of the probability of occurrence by physician and stage, regarding the prediction of impact by physician and channel (estimating the expected impact based on subjective evaluation). The explanation for Figure 11 is also related to the second method described above.

[0062] The probability of existence SK shown in Figure 11 is an example of how to calculate the probability of existence of AMTUL. Figure 11 corresponds to "Calculation of the probability of existence of AMTUL by physician" in Figure 8. The probability of existence SK indicates that the probability of AMTUL existence for each physician is calculated from the AMTUL distribution of physicians who responded to the survey. In the example shown in Figure 11, since a large number of questionnaires is necessary for accurate estimation, the probability of AMTUL being present in physicians who did not respond to the questionnaire is estimated, and the probability of AMTUL being present in each physician is calculated.

[0063] Next, referring to Figure 12, we will explain the impact by individual physicians in relation to the prediction of impact by individual physicians and channel. Figure 12 shows an example of the impact on individual physicians, based on predictions of impact by physician and channel (estimated impact based on subjective evaluation). The explanation in Figure 12 is also related to the second method described above. Figure 12 corresponds to the "Calculation of Impact by Physician" in Figure 8.

[0064] The physician-specific, channel-specific impact DI shown in Figure 12 indicates that the expected impact is calculated by adding the probability of AMTUL presence and the impact when it increases, respectively, for each physician. In the example in Figure 12, for example, in the case of a physician with physician code X, the expected impact is calculated to be 26,220 yen. A concrete example of the calculation is as follows: The expected impact is, Impact = (Probability of A existing) * (Increased impact of A) + (Probability of M existing) * (Increased impact of M) + (Probability of T existing) * (Increased impact of T) + (Probability of U existing) * (Increased impact of U) = 20% * 100 + 20% * 1,000 + 20% * 30,000 + 20% * 100,000 = 26,220 (yen).

[0065] As described above, Server 1 pre-sets simulation patterns for the number of promotions and predicts the impact on each physician in those settings. Then, through mathematical optimization, Server 1 outputs the number of promotions for each physician that maximizes the impact, within a limit that does not exceed the upper limit constraint on the total number of promotions per channel. For example, as shown in Figure 1, Server 1 (Optimal Promotion Identification Unit 52) ​​outputs the number of promotions for each physician that maximizes impact, such as "0" MR visits and "2" web explanations for Physician X1, and "1" MR visit and "1" web explanation for Physician X2, within a total limit constraint. Therefore, according to Server 1, it is possible to optimize the effect while taking resource constraints into consideration.

[0066] Although embodiments of the present invention have been described above, the present invention is not limited to the embodiments described above. Furthermore, the effects described in these embodiments are merely a list of the most preferred effects arising from the present invention, and the effects of the present invention are not limited to those described in these embodiments.

[0067] For example, the above series of processes can be executed by hardware or by software. In other words, the functional configuration shown in Figure 3 is merely an example and is not particularly limited to the functional configuration shown in Figure 3. In other words, it is sufficient to have a function that can execute the above series of processes as a whole, and the type of functional block used to realize this function is not particularly limited to the examples in Figures 1 to 12. Also, the location of the functional block is not particularly limited to Figures 1 to 12 and can be arbitrary. For example, the functional block and database of Server 1 may be transferred to a physician information system, a pharmaceutical company system, a terminal for medical representatives, a terminal for marketing personnel, an information processing device having a machine learning model KM, etc. Conversely, the functional block and database of a physician information system, a pharmaceutical company system, a terminal for medical representatives, a terminal for marketing personnel, etc., an information processing device having a machine learning model KM, etc. may be transferred to a server, etc. Furthermore, a single functional block and database may consist of hardware alone, software alone, or a combination of both.

[0068] When a series of processes are executed by software, the programs that make up that software are installed on a computer or other device from a network or storage medium. The computer may be a computer that is built into dedicated hardware. Furthermore, a computer can be any computer capable of performing various functions by installing various programs, such as a server, a general-purpose smartphone, or a personal computer.

[0069] Such recording media containing programs may consist not only of removable media (not shown) distributed separately from the main unit of the device to provide programs to users, but also of recording media provided to users, etc., that are pre-installed in the main unit of the device.

[0070] In this specification, the step of describing a program to be recorded on a recording medium includes not only processes that are performed chronologically in that order, but also processes that are not necessarily performed chronologically, but are executed in parallel or individually. Furthermore, in this specification, the term "system" refers to an overall system composed of multiple devices, means, etc.

[0071] In other words, the information processing device to which the present invention applies only needs to have the following configuration, and can take various forms. In other words, the information processing device to which the present invention is applied (for example, Server 1 in Figure 3) is: An impact prediction means (e.g., the impact (expected impact) of a physician indicated by physician codes X1 to X5 in Figure 12) is used to predict the impact for each of the zero or more promotional activities (e.g., the number of MR visits, e-detailing, and web seminars in the promotions shown in Figure 5) for a designated physician (e.g., physicians indicated by physician codes X1 to X5 in Figure 12), for each of the one or more types of channels (e.g., MR visits, e-detailing, and web seminars in the promotions shown in Figure 5), and for each of the zero or more promotional activities (e.g., the number of MR visits, e-detailing, and web seminars in the promotions shown in Figure 5), and the impact (e.g., the impact of a physician indicated by physician codes X1 to X5 in Figure 12 (expected impact)) is used for each designated physician (e.g., physicians indicated by physician codes X1 to X5 in Figure 12), and the impact prediction means (e.g., the impact prediction unit 51 in Figure 3, etc.) is used to predict the impact for each of the zero or more promotional activities (e.g., the number of MR visits, e-detailing, and web seminars in the promotions shown in Figure 5), and Based on the prediction results for the predetermined physician by the impact prediction means, an optimal promotion identification means (for example, the optimal promotion identification unit 52 in Figure 3) identifies the number of promotions that maximize impact for each of the one or more channels for the predetermined physician (for example, as shown in Figure 1, for physician X1, "0" MR visits and "2" web explanations, and for physician X2, "1" MR visit and "1" web explanation, etc., which maximizes the number of promotions that maximize impact), It is equipped with.

[0072] According to the present invention, it is possible to understand which doctors to target with which channels and how effective the promotion will be, and furthermore, to optimize the allocation of promotional resources.

[0073] In the above-mentioned information processing device (for example, Server 1 in Figure 3), The impact prediction means is Based on the results of a predetermined machine learning process using past sales data for multiple physicians (for example, sales data shown in the model construction MK in Figure 4, and sales shown in Figures 5 and 6), promotion log data (for example, promotion log data shown in the model construction MK in Figure 4, and the number of MR visits, e-detailing sessions, and web seminars in promotions shown in Figure 5), and physician attribute data (for example, physician attribute data shown in the model construction MK in Figure 4, and physician attributes (physician ID, age, medical specialty) shown in Figure 5) as training data (for example, training data GD in Figure 5), the aforementioned prediction for the predetermined physician is performed. It is possible.

[0074] In the above-mentioned information processing device (for example, Server 1 in Figure 3), The impact prediction means is Based on subjective evaluations of promotional effectiveness obtained using questionnaire data from multiple physicians (for example, questionnaire data on AMTUL shown in Figures 9 and 10, etc.) (for example, subjective evaluations shown in the Impact Overview IG in Figure 8, etc.), the above prediction is made for the specified physician. It is possible.

[0075] In the above-mentioned information processing device (for example, Server 1 in Figure 3), The aforementioned optimal promotion identification means identifies the number of promotions that maximizes the impact for each of the one or more channels for the designated physician, within the limits of the upper limit constraint on the total number of promotions for each of the one or more channels (for example, the constraint on the number of promotions shown in Figure 1 (a constraint of not exceeding 1,000 in total)). It is possible.

[0076] In the above-mentioned information processing device (for example, Server 1 in Figure 3), The impact prediction means is Based on a simulation pattern of the number of promotions set in advance for each of the one or more channels (for example, the promotion pattern shown in Figure 1, or the number of promotions shown in the model application MT in Figure 4), the prediction for the specified physician is performed. It is possible. [Explanation of Symbols]

[0077] 1...Server, 11...CPU, 18...Storage Unit, 51...Impact Prediction Unit, 52...Optimal Promotion Identification Unit, 71...Sales Information DB, 72...Promotion Information DB, 73...Physician Information DB, 74...Impact Information DB, 75...Optimal Promotion Identification Information DB, SG...Service Overview, MK...Model Construction, MT...Model Application, KM...Machine Learning Model, GD...Training Data, UB...Sales Decomposition, KH...Amount Conversion, IG...Impact Overview, HU...Average Sales, HI...Impact During Fluctuations, SK...Probability of Existence, DI...Impact by Physician and Channel

Claims

1. An impact prediction means that predicts the impact for each promotional run of zero or more times for one or more types of channels for a designated physician, Based on the results of the prediction for the predetermined physician by the impact prediction means, an optimal promotion identification means identifies the number of promotions that maximize the impact for each of the one or more channels for the predetermined physician, An information processing device equipped with the following features.

2. The impact prediction means is Based on the results of a predetermined machine learning process performed using past sales data, promotion log data, and physician attribute data for multiple physicians as training data, the prediction for the predetermined physician is performed. The information processing apparatus according to claim 1.

3. The impact prediction means is Based on subjective evaluations of promotional effectiveness obtained using questionnaire data from multiple physicians, the above prediction is made for the specified physician. The information processing apparatus according to claim 1.

4. The optimal promotion identification means identifies the number of promotions that maximizes the impact for each of the one or more channels for the designated physician, within the limits of not exceeding the upper limit constraint on the total number of promotions for each of the one or more channels. The information processing apparatus according to claim 1.

5. The impact prediction means is Based on a simulation pattern of the number of promotions set in advance for each of the one or more channels, the prediction for the designated physician is performed. The information processing apparatus according to claim 1.

6. In an information processing method executed by an information processing device, An impact prediction step is performed for a designated physician, where the impact is predicted for each promotional run of zero or more times for one or more types of channels. Based on the prediction results for the predetermined physician in the impact prediction step, the optimal promotion identification step identifies the number of promotions that maximize the impact for each of the one or more channels for the predetermined physician. Information processing methods including

7. On the computer, An impact prediction step is performed for a designated physician, where the impact is predicted for each promotional run of zero or more times for one or more types of channels. Based on the prediction results for the predetermined physician in the impact prediction step, the optimal promotion identification step identifies the number of promotions that maximize the impact for each of the one or more channels for the predetermined physician. A program that executes control processes, including those mentioned above.

Images