JP2016206736A - Customer service data processing device and customer service data processing method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide useful information to improve a customer service method.SOLUTION: A customer service data processing device includes: an estimation part which calculates the index data of an impression estimated that a customer receives from a salesclerk according to data showing the salesclerk at customer service ; a data acquisition part which acquires data showing the customer's purchase behavior; and a generation part which generates information containing the calculated index data of the impression and the data showing the customer's purchase behavior under such a state as they may be associated with each other or as they are associated with each other.SELECTED DRAWING: Figure 14

Description

  The present invention relates to a technique for processing data indicating customer service.

  In customer service, improving customer service quality is one of the important management issues. For example, in an electronic store, a customer who receives a good customer service is likely to purchase many products at the electronic store. On the other hand, a customer who is uncomfortable with the customer service may never visit the store again. Therefore, there is a great need for understanding and improving customer service quality. For example, in Patent Document 1 below, a non-contact IC (Integrated Circuit) tag is attached to a shopping cart or a basket, and IC tag readers are arranged on the floor in the store at equal intervals, so that the movement direction and movement speed of the customer can be determined. A method for grasping detailed purchase behavior including the proposal has been proposed.

  Patent Document 2 below proposes a technique for analyzing the relationship between customer satisfaction, salesclerk satisfaction, and sales in customer service at a store. This method recognizes customer emotion from customer voice, calculates customer satisfaction from recognized customer emotion, recognizes store clerk emotion from store clerk's voice, and calculates clerk satisfaction from recognized store clerk emotion Then, customer service data in which the calculated customer satisfaction and store clerk satisfaction are associated with sales results for the customer is recorded. Patent Document 3 below proposes a method of recording customer service data for grasping the relationship between the conversation ratio and customer satisfaction. This method calculates the ratio of the length of a store clerk's or customer's talk section to the sum of the length of the store clerk's talk section and the customer's talk section as the conversation ratio, and based on the result of customer emotion recognition from voice. To calculate customer satisfaction.

JP 2006-309280 A Japanese Patent No. 5533219 JP 2011-238028 A

  According to the data generated by the proposed method, it is possible to grasp the relationship between the customer satisfaction and the store clerk satisfaction and the sales record and the relationship between the conversation ratio and the customer satisfaction. However, with these data, it is not possible to find a way to improve the customer service method for increasing sales. Although it is possible to grasp that it is necessary to raise customer satisfaction according to these data, it is not possible to know what kind of customer service will increase the sales.

  The present invention has been made in view of such circumstances, and provides information useful for improving the customer service method. In the present specification, “customer service” means to contact a customer according to the wording, and the present invention does not limit the person who performs customer service, the place where customer service is performed, the purpose of customer service, and the like.

  Each aspect of the present invention employs the following configurations in order to solve the above-described problems.

  The first aspect relates to a service data processing apparatus. The customer service data processing device according to the first aspect shows, based on data indicating a store clerk at the time of customer service, estimation means for calculating index data of an impression estimated to be received by the customer from the store clerk, and customer purchase behavior Data acquisition means for acquiring data, and generation means for generating information including the calculated index data of the impression and the acquired data indicating the purchase behavior of the customer in a state that can be associated or in an associated state.

  The second aspect relates to a service data processing method executed by at least one computer. The customer service data processing method according to the second aspect calculates index data of impressions estimated to be received by a customer from the store clerk based on data indicating the store clerk at the time of customer service, and obtains data indicating customer purchase behavior And generating information including the calculated index data of the impression and the acquired data indicating the purchase behavior of the customer in a state where they can be associated or in an associated state.

  Another aspect of the present invention may be a program that causes at least one computer to execute the method of the second aspect, or a computer-readable recording medium that records such a program. May be. This recording medium includes a non-transitory tangible medium.

  According to each aspect described above, it is possible to provide information useful for improving the customer service method.

It is a figure which shows notionally the hardware constitutions of the customer service data processing apparatus in 1st embodiment. It is a figure which shows notionally the process structural example of the customer service data processing apparatus in 1st embodiment. It is a figure which shows the example of acquisition of the sales amount by a data acquisition part. It is a figure which shows the acquisition example of the customer's store stay time by a data acquisition part. It is a figure which shows the example of acquisition of the store stay time per customer of the customer by a data acquisition part. It is a figure which shows the example of detection of the customer service action (specific utterance) by a detection part. It is a figure which shows the example of calculation of the index data of the impression by an impression estimation part. It is a figure which shows the example of the information produced | generated by the production | generation part. It is a flowchart which shows the operation example of the customer service data processing apparatus in 1st embodiment. It is a figure which shows the example of the correlation information produced | generated by the production | generation part in 2nd embodiment. It is a figure which shows the example of the correlation information produced | generated by the production | generation part in 2nd embodiment. It is a figure which shows the example of the correlation information produced | generated by the production | generation part in 2nd embodiment. It is a figure which shows the example of an output by an output part. It is a figure which shows notionally the process structural example of the customer service data processing apparatus in 3rd embodiment. It is a flowchart which shows the operation example of the customer service data processing apparatus in 3rd embodiment. It is a figure which shows the example of correlation with the parameter | index data of the impression regarding a specific facial expression, and time information.

  Embodiments of the present invention will be described below. In addition, each embodiment given below is an illustration, respectively, and this invention is not limited to the structure of each following embodiment.

  The customer service data processing apparatus and the customer service data processing method in each embodiment exemplified below process data related to customer service. In each of the following embodiments, there are no restrictions on the person who receives a customer, the person who receives the customer, the place where the customer is served, the purpose of the customer service, and the like. However, in order to make the explanation easy to understand, in each of the following embodiments, a person who serves a customer is referred to as a “clerk”, and a person who serves the customer is referred to as a “customer”. Moreover, a retail store is illustrated as a place where customer service is performed. However, each embodiment is not limited to application to retail stores. For example, it is applied to places where door-to-door sales, face-to-face sales (banks, insurance, housing, cars, etc.), sales activities to customers, etc. are performed. May be.

[First embodiment]
The customer service data processing apparatus and customer service data processing method in the first embodiment will be described below with reference to the drawings.

[Hardware configuration]
FIG. 1 is a diagram conceptually illustrating the hardware configuration of the customer service data processing apparatus according to the first embodiment. A customer service data processing device (hereinafter abbreviated as a processing device) 10 is a so-called computer. As shown in FIG. 1, a CPU (Central Processing Unit) 11, a memory 12, an input / output interface (I / F) 13, The communication unit 14 and the like are included.

In addition to a general CPU, the CPU 11 includes an application specific integrated circuit (ASIC), a DSP (Digital Signal Processor), a GPU (Graphics Processing Unit), and the like.
The memory 12 is a RAM (Random Access Memory), a ROM (Read Only Memory), or an auxiliary storage device (such as a hard disk).

  The input / output I / F 13 can be connected to user interface devices such as the display device 15 and the input device 16. The display device 15 is a device that displays a screen corresponding to drawing data processed by the CPU 11 or the like, such as an LCD (Liquid Crystal Display) or a CRT (Cathode Ray Tube) display. The input device 16 is a device that receives an input of a user operation such as a keyboard and a mouse. The display device 15 and the input device 16 may be integrated and realized as a touch panel.

The communication unit 14 communicates with other computers via a communication network (not shown), exchanges signals with other devices, and the like. A portable recording medium or the like can be connected to the communication unit 14.
Each hardware configuration shown in FIG. 1 is an example, and the hardware configuration of the processing apparatus 10 is not limited to the example shown in FIG. The processing device 10 may include other hardware elements not shown. Also, the number of hardware elements is not limited to the example of FIG. For example, the processing apparatus 10 may have a plurality of CPUs 11.

[Processing configuration]
FIG. 2 is a diagram conceptually illustrating a processing configuration example of the customer service data processing device 10 in the first embodiment. The processing apparatus 10 in the first embodiment includes a first acquisition unit 21, a data acquisition unit 22, a second acquisition unit 23, a detection unit 24, an impression estimation unit 25, a generation unit 26, an analysis unit 27, an advice generation unit 28, and an output. Section 29, storage section 30 and the like. Each of these processing modules is realized by the CPU 11 executing a program stored in the memory 12. Further, the program may be installed from a portable recording medium such as a CD (Compact Disc) or a memory card or another computer on the network via the communication unit 14 and stored in the memory 12.

  The first acquisition unit 21 acquires data indicating customer behavior in the store (hereinafter referred to as customer behavior data). Examples of customer behavior data include accounting data, customer location data in a store, customer entry and exit times, customer utterance voice data, customer identification information for identifying a customer who visited the store, and the like. The first acquisition unit 21 can acquire customer settlement information as accounting data from a POS (Point Of Sales) system (not shown) or the like.

  Moreover, the 1st acquisition part 21 can acquire the position data of the customer in a store using the existing various methods. For example, the first acquisition unit 21 performs customer flow analysis based on videos obtained from a plurality of monitoring cameras (not shown) installed at a plurality of locations in the store, thereby obtaining customer position data. History can be acquired. This flow line analysis method will be described later. In addition, the first acquisition unit 21 uses information acquired from an IC tag or an RFID (Radio Frequency IDentifier) tag attached to a customer's portable object by authentication sensors (not shown) installed at a plurality of locations in the store. Based on this, a history of customer location data can also be obtained. Moreover, the 1st acquisition part 21 acquires the log | history of a customer's position data based on the information obtained from the customer's portable terminal which received the radio signal transmitted from the beacon apparatus installed in the multiple places in a shop You can also.

  Moreover, the 1st acquisition part 21 can acquire a customer's entrance time and exit time using various existing methods. For example, the 1st acquisition part 21 can acquire a customer's entrance time and exit time by detecting a customer's entrance and exit at the entrance of a store. Specifically, the 1st acquisition part 21 can acquire a customer's store entry time and store exit time by identifying a customer by the image | video obtained from the monitoring camera provided in the entrance / exit of the store. Further, the customer's store entry time and store exit time may be acquired by using information acquired from an IC tag or the like attached to the customer's portable object by an authentication sensor provided at the store entrance / exit. Further, the customer's store entry time and store exit time may be acquired based on information obtained from the customer's mobile terminal that has received a wireless signal transmitted from a beacon device provided at the store entrance.

  Moreover, the 1st acquisition part 21 can acquire a customer's speech audio | voice data from the pin microphone with which the shop clerk was mounted | worn, the microphone installed in the cash register, the microphone of the surveillance camera installed in the ceiling, etc. Moreover, the 1st acquisition part 21 can acquire the customer identification information acquired by a cash register terminal from the point card shown by a cash register, an electronic money card, a portable terminal, etc. from the cash register terminal or a POS system. Further, customer identification information may be acquired based on information obtained from a mobile terminal of a customer who has received a wireless signal transmitted from a beacon device in a store, or a feature of a customer's face from a video from a monitoring camera The quantity may be acquired as customer identification information. The acquisition method of the customer behavior data by the first acquisition unit 21 is not limited to such an example.

  The data acquisition unit 22 acquires data indicating customer purchase behavior from the customer behavior data acquired by the first acquisition unit 21. The “customer purchase behavior” indicated by the data acquired by the data acquisition unit 22 includes not only the customer's actual purchase behavior but also the customer from when the customer enters the store until the customer leaves the store. This includes various actions that may lead to customer service purposes (sales improvement). For example, “customer purchase behavior” includes application behavior such as insurance and bank account opening, and preliminary behavior for purchase such as window shopping and consultation at a bank or insurance window.

  The data acquisition unit 22 is at least one of the product purchased by the customer, the amount purchased by the customer, the customer's store stay time, the customer's in-store movement route (sales floor movement order), the customer's store visit frequency, and the customer's store visit frequency. Is acquired as data indicating the purchase behavior of the customer. However, the data indicating the purchase behavior of the customer acquired by the data acquisition unit 22 is not limited to such an example. The data acquisition unit 22 may acquire the stay time at a specific sales floor as data indicating the purchase behavior of the customer. The staying time at a specific sales floor can be calculated from the history of customer position data acquired by the first acquisition unit 21.

  The data acquisition unit 22 can obtain data of the product purchased by the customer and the amount of money (sales amount) purchased by the customer from the settlement information acquired by the first acquisition unit 21. The data acquisition unit 22 can acquire data indicating the customer's in-store movement route according to the locus of the customer's position data acquired by the first acquisition unit 21. In addition, the data acquisition unit 22 can acquire data indicating the store stay time for each customer based on the locus of the customer's position data acquired by the first acquisition unit 21 or the customer's store entry time and store exit time. . As another example, the data acquisition unit 22 may calculate the average store stay time of a customer in a predetermined time interval instead of the store stay time for each customer. For example, when a customer enters a store at times X1, X2, and X3 and a customer exits at times Y1, Y2, and Y3 in a certain time section, the total store stay time of these three customers is L = (Y1 + Y2 + Y3)-(X1 + X2 + X3). Therefore, the average store stay time per customer of this time section can be calculated by L / 3. The average store stay time per person can be calculated without determining whether the store entry and the store exit are the same person.

  The data acquisition unit 22 can also acquire the customer visit frequency and the number of customer visits as data indicating the purchase behavior of the customer based on the history of the customer identification information acquired by the first acquisition unit 21.

  As a specific example, the data acquisition unit 22 acquires the customer's payment amount (sales amount) and the customer's store stay time as data indicating the purchase behavior of the customer. The sales amount is used for customer service as it is, but the customer's staying time at the store is also linked to the purpose of customer service to improve sales. This is because it is generally said that the longer the customer stays in the store, the higher the sales.

  FIG. 3 is a diagram illustrating an example of sales amount acquisition by the data acquisition unit 22. The data acquisition unit 22 acquires the sales amount (total amount) and the settlement time from the settlement information. The acquired sales amount and settlement time are stored in the storage unit 30 in association with each settlement as shown in FIG. According to the example of FIG. 3, it is shown that sales were 1500 yen at 7:24:20.

  FIG. 4 is a diagram illustrating an example of acquisition of customer store stay time by the data acquisition unit 22. The data acquisition unit 22 calculates the store stay time for each customer from the store entry time and the store exit time for each customer acquired by the first acquisition unit 21 as described above. The calculated store stay time is associated with identification information (customer number), store entry time, and store exit time for each customer by the generation unit 26 described later, as shown in FIG. Stored. According to the example of FIG. 4, it is shown that the customer of customer number 1 entered the store at 7:05:00, stayed at the store for 43 minutes and 15 seconds, and exited at 7:48:15. .

  FIG. 5 is a diagram illustrating an example of acquisition of store stay time per customer by the data acquisition unit 22. As described above, the data acquisition unit 22 calculates the average store stay time of the customer in a predetermined time interval. As shown in FIG. 5, the calculated average store stay time is stored in the storage unit 30 in association with the identification information of the time section and the number of customers visiting the store for each time section, as shown in FIG. 5. . According to the example of FIG. 5, it is shown that four customers visited the store between 7:00 and 8:00, and the average stay time per customer was 40 minutes and 15 seconds.

  The second acquisition unit 23 acquires data indicating the behavior of the clerk (hereinafter referred to as clerk behavior data). Examples of the clerk behavior data include clerk utterance voice data, image data representing the clerk's face and motion, position data of the clerk in the store, operation history (login history) of each clerk at the cash register terminal, and the like.

  The second acquisition unit 23 applies a known voice segment detection technique or speaker verification technique to a voice signal obtained from a pin microphone worn by a store clerk, a microphone installed in a cash register, a microphone of a surveillance camera installed on a ceiling, or the like. Thus, the utterance voice data of the clerk can be acquired. For example, the second acquisition unit 23 acquires the voice data of the section where the sound volume equal to or higher than a predetermined value continues as the clerk's utterance voice data, and acquires the start time and the end time of the utterance voice data together. it can. Further, the second acquisition unit 23 applies the speaker verification technology using the voice or voice feature information of each clerk registered in advance to the obtained voice data, so that the utterance voice data of the clerk Can be obtained. At this time, the second acquisition unit 23 may also acquire customer utterance voice data obtained from the voice data as supplementary information of the clerk utterance voice data. For example, the 2nd acquisition part 23 acquires the audio | voice data of the area where the sound volume below a predetermined value is continuing as a customer's utterance audio | voice data. Moreover, the 2nd acquisition part 23 acquires the speech area different from the salesclerk's voice registered beforehand as customer's speech voice data.

In addition, the second acquisition unit 23 can acquire image data representing a store clerk's face and actions from a camera installed at a cash register, a monitoring camera installed on a ceiling, and the like.
The second acquisition unit 23 can obtain information on a store clerk's use history (operation history) for the cash register terminal managed there from a POS system or the like.

  Moreover, the 2nd acquisition part 23 can acquire the position data of a salesclerk by various existing methods. For example, the second acquisition unit 23 performs the flow analysis of the store clerk based on the images obtained from a plurality of monitoring cameras (not shown) installed at a plurality of locations in the store, thereby obtaining the location data of the store clerk. History can be acquired. In addition, the second acquisition unit 23 is based on information acquired by the authentication sensors (not shown) installed at a plurality of locations in the store from the IC tag or RFID tag attached to the store clerk's portable items. A history of position data can also be acquired. Moreover, the 2nd acquisition part 23 acquires the log | history of the position data of a salesclerk based on the information obtained from the mobile terminal of the salesclerk who received the radio signal transmitted from the beacon apparatus installed in the multiple places in a shop You can also.

  The detection unit 24 detects the customer service behavior of the salesclerk for the customer from the salesclerk behavior data acquired by the second acquisition unit 23. The customer service behavior detected by the detection unit 24 includes operations, actions, behaviors, remarks, attitudes, and the like performed by the store clerk for the customer. For example, customer interaction includes greetings to customers, sales talks, conversations with customers (including utterances to customers, response utterances to customer questions, etc.), provision of information to customers, provision of products or services to customers, stores Examples include guidance, facial expressions to be shown to customers, and accounting work. However, the customer service behavior detected by the detection unit 24 is not limited to such an example.

  The detection unit 24 can acquire speech text data indicating the clerk's utterance content by applying a speech recognition process to the clerk's speech data acquired by the second acquisition unit 23. Various existing voice recognition methods can be used for the voice recognition process. Furthermore, the detection unit 24 can use the start time and end time of the utterance acquired by the second acquisition unit 23 together with the utterance text data.

  The detection unit 24 detects a specific expression from the result of speech recognition (utterance text data) as described above, thereby performing customer service actions such as greetings, sales talks, providing information to customers, store guidance, and accounting work. Can be detected. For example, detection of “providing information to a customer” can be realized by detecting a specific expression representing a product or service name handled in a store. In addition, the store information can be detected by detecting a specific expression such as “Please come here” or “Here is”. The detection of the accounting work can be realized by detecting specific expressions such as “next person please”, “... will be a circle”, “do you have a point card”. At this time, the detection unit 24 may detect only an expression that completely matches the specific expression, or in addition, may detect an expression (such as a difference in endings) similar to the specific expression.

  As a specific example, the detection unit 24 detects a salesclerk greeting and a sales talk in a specific utterance of the salesclerk as a customer service action. Greetings and sales talks are widely regarded as basic elements in customer service. For example, at convenience stores and department stores, a method for producing good greetings and a method for sales talks has been made into a manual, which is used in the education and evaluation of shop assistants. As a specific representation of the greeting, "May I help you?", "Good morning", "Hello", "Good evening", "Thank you", and "also not please come" is used. In addition, as a specific expression representing sales talk, a product or service name handled in a store is used. For example, in the case of a convenience store, “coffee”, “melon bread”, and the like can be set as a specific expression representing sales talk.

  FIG. 6 is a diagram illustrating a detection example of customer service behavior (specific utterance) by the detection unit 24. The detection unit 24 detects the utterance corresponding to the specific expression by using different specific expressions for each specific utterance type (greeting and sales talk). Then, as shown in FIG. 6, the generation unit 26 described later, for each detected utterance, the start time and end time of the utterance, the type of the utterance (greeting or sales talk), and utterance content text data Are associated with each other and stored in the storage unit 30. According to the example of FIG. 6, it is shown that the greeting of the salesclerk “I welcome you” was detected from 7:05:10 to 7: 5: 12.

Moreover, the detection part 24 may detect all the salesclerk's utterances with respect to a customer as a customer service action based on the result (utterance text data) of the above speech recognition.
Moreover, the detection part 24 may detect the response speech with respect to a customer's inquiry as a customer service action. For example, the detection unit 24 uses the customer's utterance voice data acquired as supplementary information by the second acquisition unit 23 together with the clerk's utterance voice data, so that "the clerk's utterance to the customer" or "response utterance to the customer's inquiry" Can be detected. The detection unit 24 detects a clerk's utterance uttered within a certain time from the customer's utterance as a “response utterance”. In addition, when the customer utterance exists within a predetermined time before and after the clerk's utterance, the detection unit 24 detects the utterance of the clerk to the customer. However, the detection method of “the clerk's utterance to the customer” and “response utterance” by the detection unit 24 is not limited to such an example.

  Moreover, the detection part 24 can also detect a shop staff's specific facial expression from the image data acquired by the 2nd acquisition part 23 as a customer service action. For example, the detection unit 24 specifies an image area representing a store clerk's face from the image data using an image recognition technique, and applies a facial expression recognition technique to the image area, thereby specifying a smile, grimacing, and the like. A facial expression can be detected.

  Further, the detection unit 24 can detect the clerk's accounting work as a customer service action based on the clerk's operation history information for the cash register terminal acquired by the second acquisition unit 23. For example, the detection unit 24 extracts the start time and the end time of using the cashier terminal from the operation history information, and detects a period between the start time and the end time as the clerk's accounting work.

  The detection unit 24 uses the store clerk's store guidance as a customer service action based on the store location data history acquired by the second acquisition unit 23 and the customer location data history acquired by the first acquisition unit 21. Can be detected. For example, the detection unit 24 detects, as store guidance, that a store clerk and a customer are present within a certain distance for a certain period of time and both are moving.

  Various existing methods can be used for the flow line analysis method of the customer and the store clerk. For example, the 1st acquisition part 21 and the 2nd acquisition part 23 detect a customer or a salesclerk by the background subtraction method from the image obtained from a surveillance camera. In the background difference method, a model representing background information is constructed from a plurality of images input along a time series, a difference between the target image and the background model is calculated, and an image area having a large difference is used as a customer or a store clerk. Detected. Moreover, the 1st acquisition part 21 and the 2nd acquisition part 23 use the face detector and head detector which modeled the face and head of a shop assistant or the customer, and were learned by the model, without using a background model. Thus, a human face or head may be detected from the image. Such a human face detection method is also used for the specific facial expression detection method described above by the detection unit 24. The first acquisition unit 21 and the second acquisition unit 23 track (track) the customer or the store clerk detected in this way among a plurality of images by temporal and spatial condition matching, etc. The position data history can be obtained. However, not only image data but also position data obtained by using an authentication sensor or a beacon device may be used for the flow line analysis of the customer and the store clerk. Tracking accuracy can be improved by tracking a customer or a store clerk between a plurality of images using position data obtained from other than this image.

  The impression estimation unit 25 uses the data indicating the store clerk when the customer service action is performed based on the customer service behavior detected by the detection unit 24, and the impression that the customer is estimated to receive from the store clerk. The index data of is calculated. The impression index data indicates the type of the impression (bright, fine, etc.), the presence or absence of the impression, or the degree of the intensity of the impression. The impression estimation unit 25 may calculate index data of a plurality of impressions indicating the presence or absence or strength of each impression. The number and types of impressions estimated are not limited, and various impressions that the customer can feel from the customer service behavior may be the estimation target.

The impression estimation method by the impression estimation unit 25 can be classified into the following two methods.
One is a method of calculating index data of an impression that the customer is estimated to receive from the customer service behavior of the clerk detected by the detection unit 24 (hereinafter referred to as a first estimation method),
The other is the impression of the customer estimated to be received from the customer service behavior of the same store clerk, which is different from the customer service behavior detected at the same time as the customer service behavior detected by the detection unit 24. This is a method for calculating index data (hereinafter referred to as a second estimation method).

<First estimation method>
First, the first estimation method will be described.
When the clerk's utterance to the customer is detected by the detection unit 24 as a customer service behavior, the impression estimation unit 25 uses the detected clerk's utterance voice data to determine brightness, spirit, strength, coolness, and politeness. Index data indicating the presence or strength of at least one impression of calmness, cuteness, honesty, coldness, fluency, youth, elegance, and kindness is calculated. When the detection unit 24 detects a greeting or a specific utterance of sales talk as a customer service behavior, the impression estimation unit 25 determines brightness, spirit, strength, and politeness from the voice data of the detected clerk's speech. It is desirable to calculate index data indicating the presence or absence or strength of at least one impression of coldness. This is because these impressions are widely regarded as important in customer service quality control. For example, at convenience stores and department stores, “bright” and “energetic” greetings are thought to help increase sales and increase the customer's repeat rate. Also, “cold” customer service utterances are thought to lead to customer churn.

  When the detection unit 24 detects a greeting or a specific utterance of sales talk as a customer service behavior, the impression estimation unit 25 determines brightness, spirit, strength, and politeness from the voice data of the detected clerk's speech. And index data indicating the strength of the cold impression. In other words, the impression estimation unit 25 allows the customer who heard the salesclerk's greeting and the sales talk to “brightness”, “energy”, “power”, “poorness”, and “cold” Estimate how much “feel” is felt. Specifically, the impression estimation unit 25 extracts a feature amount from voice data corresponding to a salesclerk greeting or sales talk detected by the detection unit 24. The impression estimation unit 25 performs power (volume), fundamental frequency (voice pitch), Mel-Frequency Cepstial Coefficients (MFCC, frequency spectrum envelope shape) for each frame obtained by dividing the audio data by a window width of 10 msec. ), The phoneme duration, and the time difference (delta amount) between them. Further, the impression estimation unit 25 calculates statistics such as the average value, variance, minimum value, maximum value, and the like of the feature amount of each frame unit for the speech section representing the greeting or the sales talk, and uses these for the impression estimation. The feature amount to be used. Note that the feature amount used for impression estimation is not limited to these.

  The impression estimation unit 25 calculates the index data of the impression received by the customer from the salesclerk greeting and the sales talk by applying the voice impression estimation model to the calculated feature amount. For example, as a voice impression estimation model, for each of “brightness”, “energy”, “strength”, “poorness”, and “coldness” of the input voice data, A model calculated as a numerical value is used. As the voice impression estimation model, a model that estimates the presence / absence of these impressions may be used, or a model that classifies voice data into any of these impressions may be used. In order to realize the voice impression estimation model, a well-known pattern recognition model such as Support Vector Machine (SVM), neural network, logistic regression model, or decision tree can be used. The voice impression estimation model is learned as follows and may be held in advance by the impression estimation unit 25 or may be held by another device. Further, the impression estimation unit 25 may learn a voice impression estimation model.

  The voice impression estimation model is learned as follows using the speech data for learning. First, impression information obtained by subjective evaluation of how the sound data is heard is assigned to each of the learning sound data. Next, a voice impression estimation model is learned using the feature amount calculated by the above-described method from the speech data for learning and the given impression information. For example, when learning a model that estimates the “brightness” that is felt when listening to audio data, first, the impression information assigner listens to the learning audio data and feels the audio data “bright”. Impression information indicating whether or not is given. Next, a pattern recognition model for estimating the given impression information is learned. Note that when learning a model for estimating the impression of greetings, it is desirable to limit the speech data for learning to greetings. This makes it possible to obtain a voice impression estimation model in which features unique to greetings are learned, thereby improving impression estimation accuracy. Further, it is desirable that impression information is given by a plurality of workers, and impression information to be given is determined by majority vote. The impression received from customer service is largely subjectivity, but by integrating the results of multiple workers, it is possible to learn a model that can correctly estimate the impression received by the majority of customers.

  Such a method of calculating the index data of impression is applicable not only to specific utterances of greetings or sales talks, but also to all utterances of the store clerk to customers such as response utterances.

  FIG. 7 is a diagram illustrating a calculation example of impression index data by the impression estimation unit 25. The impression estimation unit 25 estimates “brightness”, “energy”, and “strength” estimated from the speech data for each greeting and sales talk detected by the detection unit 24 as customer service behavior. , Numerical values (index data) indicating the strength of each impression of “politeness” and “coldness” are calculated in the range of 0 to 100, respectively. Then, as shown in FIG. 7, the generation unit 26 described later, for each utterance detected by the detection unit 24, the start time and end time of the utterance, the type of the utterance (greeting or sales talk), and the utterance content The text data and the impression index data are associated with each other and stored in the storage unit 30. According to the example of FIG. 7, the impression that the customer receives from the greeting utterance “I welcome you” issued at 7:05:10 is “brightness = 92”, “energy = 80”, “strength” = 55 "," Politeness = 77 ", and" Coolness = 1 "are calculated. Also in this embodiment, in addition to the above-mentioned impressions, the impression estimation unit 25 is “cool”, “calm”, “cute”, “honesty”, “fluency”, “youth”, “elegant” "S", "kindness", etc. may be the estimation target.

  When the specific facial expression of the store clerk is detected as a customer service behavior by the detection unit 24, the impression estimation unit 25 uses the image data including the facial expression of the store clerk detected as the specific facial expression to estimate the impression that the customer receives from the facial expression. Calculate index data. In this case, the impression estimation unit 25 calculates a facial feature amount from the image data, and applies a facial impression estimation model to the calculated facial feature amount to estimate that the customer receives from the facial expression. It is possible to calculate the index data of the impression to be made. Various existing methods can be used as a method for calculating the facial feature amount. For example, the facial feature amount is represented using a vector describing the shape and positional relationship of eyes, nose, mouth, etc., a vector representing the gray value of each pixel, and the like. The face impression estimation model can be learned by the same method as the voice impression estimation model learning using the learning face image data. That is, the face impression estimation model is learned using learning face image data to which impression information obtained by subjective evaluation is added. As for the impression estimated to be received by the customer from the facial expression of the clerk, the same impression as that estimated from the clerk's utterance may be the estimation target.

《Second estimation method》
Next, the second estimation method will be described. In the second estimation method, when the first customer service behavior is detected by the detection unit 24, the impression estimation unit 25 is different from the first customer service behavior when the detected first customer service behavior is performed. Then, from the data indicating the second customer service behavior of the same store clerk, the index data of the impression estimated to be received by the customer from the second customer service behavior is calculated.

  For example, when the clerk's utterance (including a specific utterance and a response utterance) is detected by the detection unit 24 as a customer service behavior, the impression estimation unit 25 calculates index data of an impression received by the customer from the facial expression of the clerk during the utterance. To do. In this case, the impression estimation unit 25 uses the start time and end time of the clerk's utterance detected by the detection unit 24 to start the image data including the clerk's face acquired by the second acquisition unit 23. Image data between the time and the end time is specified. The impression estimation unit 25 identifies an image area indicating the face of the store clerk from the identified image data as described above, and calculates a feature amount of the image area. The impression estimation unit 25 can calculate the index data of the impression received by the customer from the facial expression of the clerk during the utterance by applying the above-described face impression estimation model to the calculated feature amount.

Further, when the clerk's accounting work is detected as a customer service behavior by the detection unit 24, the impression estimation unit 25 calculates the index data of the impression received by the customer from the utterance or facial expression of the clerk during the accounting work. In this case, the impression estimation unit 25 uses the disclosure time and the end time of the clerk's accounting work detected by the detection unit 24, and among the utterance voice data or image data of the clerk acquired by the second acquisition unit 23, Speech audio data or image data between the start time and end time is specified. The impression estimation unit 25 calculates index data of an impression received by the customer from the utterance or facial expression of the store clerk from the identified speech voice data or image data as described above.

  Further, when the store guide of the store clerk is detected as a customer service behavior by the detection unit 24, the impression estimation unit 25 calculates index data of an impression received by the customer from the utterance or facial expression of the store clerk in the store guide. In this case, the impression estimation unit 25 uses the start time and end time of the store guidance act of the store clerk detected by the detection unit 24, among the clerk speech data or image data acquired by the second acquisition unit 23. The speech audio data or the image data between the start time and the end time is specified. The impression estimation unit 25 calculates index data of an impression received by the customer from the utterance or facial expression of the store clerk from the identified speech voice data or image data as described above.

  The impression estimation unit 25 may execute only one of the two impression estimation methods described above, or both. When both are executed, the impression estimation unit 25 uses the data indicating the first customer service behavior and the data indicating the second customer service behavior of the same store clerk when the first customer service behavior detected by the detection unit 24 is performed. The index data of the impression is calculated.

  The generation unit 26 generates information including the index data of the impression calculated by the impression estimation unit 25 and the data indicating the purchase behavior of the customer acquired by the data acquisition unit 22 in an associated state or an associated state. The generation unit 26 stores the generated information in the storage unit 30. Each table illustrated in FIG. 3 to FIG. 7 described above is an example of information generated by the generation unit 26 and stored in the storage unit 30.

  The generation unit 26 associates data indicating the purchase behavior of the customer as illustrated in FIGS. 3, 4, and 5 with time information corresponding to the data and stores the data in the storage unit 30. In the example of FIG. 3, the generation unit 26 stores the sales amount and the settlement time acquired as data indicating purchase behavior in the storage unit 30 in association with each settlement. In the example of FIG. 4, the generation unit 26 associates the store stay time calculated as data indicating purchase behavior with the store entry time and store exit time used for the calculation, and the customer identification number, 30. In the example of FIG. 5, the generation unit 26 stores the average store stay time calculated as data indicating purchase behavior in the storage unit 30 in association with the identification information of the time interval and the number of customers visiting the time interval.

  Further, the generation unit 26 associates the impression index data as shown in FIG. 7 and the time information corresponding to the index data, and stores them in the storage unit 30. Here, the time information associated with the impression index data indicates the time at which a capacity action estimated to give the impression to the customer was performed. In other words, the generation unit 26 associates the impression index data with information on the time when the customer service behavior detected by the detection unit 24 is performed. In the example of FIG. 7, for each utterance detected by the detection unit 24 as a customer service behavior, the generation unit 26 starts and ends the utterance, type of the utterance (greeting or sales talk), utterance content text data, And impression index data are associated with each other and stored in the storage unit 30. As shown in FIG. 6, the generation unit 26, for each utterance detected by the detection unit 24 as customer service behavior, the start time and end time of the utterance, the type of the utterance (greeting or sales talk), and the content of the utterance The text data may be associated with each other and stored in the storage unit 30.

  In the examples of FIGS. 3, 4, 5, and 7, each piece of information (each table information) generated by the generation unit 26 is directly associated with impression index data and customer purchase behavior data. Absent. However, each table information has a common item of time information (payment time, store entry time and store exit time, time section, start time and end time), and can be associated with each other.

  The generation unit 26 can also generate information including the impression index data illustrated in FIG. 7 and the data indicating the purchase behavior of the customer illustrated in FIG. 3 in an associated state. For example, the generation unit 26 calculates an average sales amount for each predetermined time interval based on the sales amount and the settlement time shown in FIG. 3, associates the identification information of the time interval with the average sales amount, and stores the storage unit 30. To store. Further, the generation unit 26 calculates average index data for each predetermined time interval of the impression from the relationship between the start time and end time shown in FIG. 7 and the impression index data, and identifies identification information and average index data of the time interval. Are stored in the storage unit 30. Thereby, the impression index data and the data indicating the purchase behavior of the customer are associated with each other in the common time interval item.

  FIG. 8 is a diagram illustrating an example of information generated by the generation unit 26. As illustrated in FIG. 8, the generation unit 26 can also generate information including impression index data and data indicating customer purchase behavior in an associated state. In the example of FIG. 8, the generation unit 26 generates graph information in which impression index data and data indicating customer purchase behavior are directly associated with each other. In the example of FIG. 8, the impression index data (F20, F21) shown in FIG. 7, the sales amount (F22) shown in FIG. 3, the store stay time (F23) for each customer shown in FIG. Is generated. Each graph is related by a time axis on the horizontal axis. F20 represents the index data of the impression of “brightness” received by the customer and the time of occurrence of the salesclerk greeting (displayed with a circle mark) and sales talk (displayed with a star mark). F21 represents index data of the customer's greeting and sales talk occurrence time and the impression of “strength” received by the customer. F22 represents the occurrence time and the amount of sales. F23 represents a time section in which each customer stayed in the store with a horizontal bar. When the stay time is 30 minutes or more, the side line is shown thick. The numbers appended to the left of each side line indicate the customer number in FIG.

  In the above-described example, the generation unit 26 associates the time information with the impression index data and the data indicating the purchase behavior of the customer, so that both data can be associated with each other or can be associated with each other. In this way, the impression index data and the data indicating the customer's purchase behavior are associated with the time information, respectively, and the impression obtained from the customer service behavior of the store clerk who is close in time is associated with the customer's purchase behavior and analyzed. It becomes possible to do. In this case, when the data acquisition part 22 acquires the data which show a customer's purchase behavior, it acquires together with the time information corresponding to this data. When acquiring the product purchased by the customer or the amount purchased by the customer as the data, the data acquisition unit 22 can acquire the settlement time from the settlement information. When acquiring the customer's store stay time and the customer's in-store movement route as the data, the data acquisition unit 22 can acquire the start time and end time of the information. In addition, when calculating the impression index data, the impression estimation unit 25 acquires information on the time of the original data of the calculation, that is, the time when the store clerk's customer service behavior detected by the detection unit 24 is performed. Can do.

  Further, the generation unit 26 may include at least one of content identification data for identifying presence / absence of detection of customer service behavior of the store clerk, type of customer service behavior of the detected store clerk, and content of detected customer service behavior of the store clerk as an impression index. Information associated with data or data indicative of customer purchasing behavior may be generated. The type of customer service action indicates an action type such as utterance, facial expression, accounting work, store guidance. The content identification data may be identification information for further classifying speech behaviors such as greetings, sales talks, response speeches, etc., or data directly representing content such as the above-mentioned speech content text data. Also good.

  The analysis unit 27 calculates a correlation coefficient between the impression index data calculated by the impression estimation unit 25 and the data indicating the purchase behavior of the customer acquired by the data acquisition unit 22. When the index data relating to a plurality of impressions is calculated by the impression estimation unit 25, the analysis unit 27 calculates a correlation coefficient between the index data of each impression and data indicating the purchase behavior of the customer. The correlation coefficient is an index value indicating the degree of correlation between two data, and its calculation method is well known.

  When the storage unit 30 stores information including the impression index data and the data indicating the purchase behavior of the customer in association with each other, the analysis unit 27 displays the association information of both data stored in the storage unit 30. To calculate the correlation coefficient. As illustrated in FIG. 3 and FIG. 7, when both data are not associated in the storage unit 30, the analysis unit 27 instructs the generation unit 26 to associate.

  The advice generation unit 28 generates advice information about customer service based on the correlation coefficient calculated by the analysis unit 27. For example, the advice generation unit 28 determines whether the calculated correlation coefficient is higher than a predetermined first threshold value. The advice generation unit 28 includes information (such as bright) that indicates the type of impression that is highly correlated with the data indicating the purchase behavior of the customer, and that indicates that the impression type is important for improving sales. Generate. Furthermore, the advice generation unit 28 determines whether or not the calculated correlation coefficient is lower than a predetermined second threshold value (a threshold value that is the same as or different from the first threshold value). As a result of this determination, the advice generation unit 28 includes, in the advice information, information indicating a type of impression with low correlation (powerful, etc.) and an expression indicating that the impression type need not be particularly taken care of. Also good. Furthermore, the advice generation unit 28 may include time zone information to be improved in the advice information for the type of impression having a high correlation with the data indicating the purchase behavior based on the information as illustrated in FIG. 8. For example, in the example of FIG. 8, since the index value of the impression of “brightness” between 9 o'clock and 11 o'clock is low and the sales in that time zone are low, the advice generation unit 28 sets the “brightness” in that time zone. Include advice in the advice that should improve the impression.

  The output unit 29 outputs information stored in the storage unit 30. The specific form of the output by the output unit 29 is not limited. The output unit 29 may cause the display of the information to be output to the display device 15 via the input / output I / F 13 or cause the printing device (not shown) to print the information via the input / output I / F 13. May be. The output unit 29 may output the information to another device via the communication unit 14, or may output the information as an electronic file to a portable recording medium or the like via the communication unit 14. .

  Specifically, based on the information stored in the storage unit 30, the output unit 29 outputs the impression index data and the data indicating the purchase behavior of the customer in an associated state or an associated state. For example, as illustrated in FIGS. 3, 4, and 5, the output unit 29 outputs information (tables, graphs, and the like) in which data indicating customer purchase behavior is associated with time information. As shown in FIG. 7, the output unit 29 outputs information (table, graph, etc.) in which impression index data and time information are associated with each other. Thus, the output unit 29 may separately output the table shown in FIG. 3, FIG. 4 or FIG. 5 and the table shown in FIG. Further, as shown in FIG. 8, the output unit 29 outputs information (tables, graphs, etc.) associated with impression index data and data indicating customer purchase behavior.

  For example, when the table image of the index data of impression with time information shown in FIG. 7 and the table image of sales amount data with time information shown in FIG. 3 are output, The following can be grasped. The “brightness” of clerk greetings and sales talks from 7:00 to 8:00 is generally as high as 80 or more, and sales are as large as 1,000 yen or more. On the other hand, the “brightness” of salesclerk greetings and sales talks from 10:00 to 11:00 is generally as small as 30 or less, and sales are as small as 1,000 yen or less. From this, it can be seen that the greater the “brightness” of greetings and sales talks, the greater the sales. Furthermore, according to the table image of the customer's store stay time with time information shown in FIG. That is, when the “brightness” of the salesclerk greetings and sales talks from 7 am to 8 am is large, the customer stays at the store for about 40 minutes. On the other hand, when the “brightness” of the salesclerk greeting or sales talk from 10:00 to 11:00 is small, the customer's stay in the store is as short as about 10 minutes. From this, it can be seen that the longer the “brightness” of the greeting or sales talk, the longer the customer stays at the store.

  Further, as shown in FIG. 8, by outputting a graph in which impression index data and data indicating customer purchase behavior are associated with each other, it is possible to intuitively grasp the relationship between them. For example, it is easy to understand that the brighter the greeting or sales talk, the larger the sales and the longer the customer's stay in the store. It can also be seen that the “strength” of greetings and sales talks has little to do with sales and customer stay time, so there is no need to aim for a strong greeting.

  Further, the output unit 29 may output at least one of the advice information generated by the advice generation unit 28 and the correlation coefficient calculated regarding the impression.

[Operation example / Service data processing method]
Hereinafter, the service data processing method in the first embodiment will be described with reference to FIG. FIG. 9 is a flowchart showing an operation example of the customer service data processing apparatus 10 in the first embodiment. For example, each process shown in the drawing is executed by each of the above-described processing modules included in the processing apparatus 10. Since each process is the same as the processing content of each processing module described above that the processing apparatus 10 has, details of each process are omitted as appropriate.

  The processing apparatus 10 acquires store clerk behavior data (S90). At this time, the processing apparatus 10 also acquires time information corresponding to the data along with the clerk behavior data. The clerk behavior data is as described above. For example, the processing device 10 acquires the clerk's voice data, face image data, position data, and the like as clerk action data. The processing device 10 is connected to at least one of a camera and microphone installed in the store, a camera and microphone carried by a store clerk, a POS system, a cash register terminal, a server device, and the like via the input / output I / F 13 or the communication unit 14. , Get clerk behavior data. The processing device 10 can also acquire video recording data obtained from a camera and audio recording data obtained from a microphone from a portable recording medium or another device as clerk behavior data. The method for acquiring the clerk behavior data is not limited.

  The processing device 10 detects the customer service behavior of the clerk from the clerk behavior data acquired in (S90) (S91). The customer service behavior of the store clerk and its detection method are as described above. The processing device 10 detects the customer service behavior and also acquires information on the time when the customer service behavior was performed. This time information may be acquired together with the clerk behavior data in (S90). For example, the processing device 10 detects a salesclerk greeting or sales talk as a customer service action, and also acquires the start time and end time of the utterance.

  Based on the customer service behavior detected in (S91), the processing device 10 uses the data indicating the customer when the customer service behavior is performed, and the impression that the customer is estimated to receive from the customer Index data is calculated (S92). The impression estimation method in (S92) can be classified into two as described above (first estimation method and second estimation method). In the first estimation method, the processing device 10 calculates index data of an impression that is estimated to be received by the customer from the customer service behavior of the clerk detected in (S91) (S92). In the second estimation method, the processing device 10 determines that the customer has received the customer service from the customer service behavior of the same store clerk, which is different from the customer service behavior detected at the same time as the customer service behavior detected in (S91). The index data of the impression estimated to be received is calculated (S92). The calculation method of impression index data is as described above.

  For example, in the first estimation method, the processing device 10 uses the voice data of the greeting and sales talk detected in (S91), and the strength of the impression of brightness, energy, strength, politeness, and coldness. The index data indicating is calculated. In the second estimation method, for example, the processing device 10 receives an impression (brightness) that the customer receives from the facial expression of the clerk from image data including the clerk's face at the same time as the greeting and sales talk detected in (S91). , Energy, coldness, etc.) is calculated.

  The processing device 10 acquires customer behavior data (S93). At this time, the processing apparatus 10 also acquires time information corresponding to the data together with the customer behavior data. The customer behavior data is as described above. For example, the processing apparatus 10 acquires transaction data (accounting information), position data, store entry and exit times, audio data, and the like as customer behavior data. The processing device 10 receives customer behavior from at least one of a camera and a microphone installed in the store, a customer's mobile terminal, a POS system, a cash register terminal, a server device, etc. via the input / output I / F 13 or the communication unit 14. Get the data. The processing device 10 can also acquire video recording data obtained from a camera and audio recording data obtained from a microphone as customer behavior data from a portable recording medium or another device. The method for acquiring customer behavior data is not limited.

  The processing device 10 acquires data indicating the purchase behavior of the customer from the customer behavior data acquired in (S93) (S94). At this time, the processing apparatus 10 acquires time information corresponding to the data together with the data indicating the purchase behavior. The data indicating the purchase behavior of the customer and the method for acquiring the data are as described above. For example, the processing apparatus 10 acquires the sales amount and the settlement time from the settlement information acquired as the customer behavior data in (S93). Further, the processing device 10 calculates the store stay time for each customer or the average store stay time of the customer in a predetermined time interval from the store entry time and the store exit time acquired as customer behavior data in (S93).

  The processing apparatus 10 generates information including the index data of the impression calculated in (S92) and the data indicating the purchase behavior of the customer acquired in (S94) in an associated state or an associated state (S95). . Specifically, the processing device 10 separately associates the association information between the data indicating the purchase behavior of the customer and the time information corresponding to the data and the association information between the impression index data and the time information corresponding to the index data. Generate. In addition, the processing device 10 generates association information between data indicating customer purchase behavior and index data of impressions. The processing device 10 stores the generated information in the storage unit 30.

  The processing device 10 calculates a correlation coefficient between the index data of impression and data indicating the purchase behavior of the customer (S96). At this time, the processing apparatus 10 can use association information between the index data of impressions and data indicating the purchase behavior of the customer.

  The processing device 10 generates advice information based on the correlation coefficient generated in (S96) (S97). The contents of the advice information and the generation method thereof are as described above.

  The processing device 10 outputs at least one of the information generated in (S95), the correlation coefficient calculated in (S96), and the advice information generated in (S97) (S98). The specific form of this output is not limited. For example, the processing apparatus 10 outputs at least one of the tables shown in FIGS. 3, 4, and 5 and the table shown in FIG. 7 separately. Further, the processing apparatus 10 outputs the graph shown in FIG.

  The execution order of each process in the customer service data processing method in the first embodiment is not limited to the example shown in FIG. The execution order of each process can be changed within a range that does not hinder the contents. For example, (S90) to (S92), (S93), and (S94) may be executed in parallel. Further, (S96) and (S97) may not be executed sequentially from the previous step. For example, the processing device 10 executes (S96) based on at least one of the number of pieces of information generated in (S95), elapse of a predetermined time, detection of a user instruction operation using the input device 16, and the like. Determine whether to do it. Similarly, the processing apparatus 10 may determine whether or not to execute (S97). Thereby, in (S98), when only the information generated in (S95) and stored in the storage unit 30 is output, or when only the correlation coefficient calculated in (S96) is output, in (S97) When only the generated advice information is output, two or more of the information may be output.

[Operation and effect of the first embodiment]
As described above, in the first embodiment, data indicating the purchase behavior of the customer is acquired, the customer service behavior of the store clerk is detected, and it is estimated that the customer receives from the store clerk when the detected customer service behavior is performed. The impression index data is calculated. Then, information including impression index data and data indicating the purchase behavior of the customer in an associated state or an associated state is generated and output. Thus, according to the first embodiment, it is possible to provide information useful for improving the customer service method. That is, the store-side person in charge who sees the output information can find out specific measures for improving the store staff's customer service method, which are necessary for realizing improvement in store performance. For example, the person in charge can grasp the relationship between the impression received by the customer from the greeting of the store clerk and the sales talk, and the sales and the customer's store stay time. Then, the person in charge can find an improvement measure such as educating the store clerk to brighten greetings and sales talks in order to increase sales or increase the customer's stay time in the store. In addition, it is also possible to obtain knowledge that strong greetings and sales talks are meaningless for improving sales.

  Further, in the first embodiment, based on data indicating a store clerk's behavior (speech, facial expression, etc.), index data for an impression estimated to be received by the customer from the behavior is calculated. Therefore, the index data of the impression is different from the customer satisfaction calculated from the customer's own voice proposed in Patent Documents 2 and 3. According to the first embodiment, it is possible to calculate the index data of the impression received by the customer without the customer data. That is, according to the first embodiment, it is possible to reliably acquire information useful for improving the customer service method. This is because it may be difficult to take a voice or image of a customer in a place where customer service is performed due to personal information protection and privacy issues.

  In addition, according to the first embodiment, “brightness”, “energy”, “strength”, “poorness”, “coldness” from the impression that is regarded as important in customer service quality management in customer service. Thus, an impression to be estimated can be selected. In addition, by calculating the index data of a plurality of impressions in this way, it is possible to obtain knowledge of the relevance to the customer purchase behavior for each impression type. Depending on the type of business and the place where customer service is performed, the type of impression that affects the customer's purchasing behavior may differ. According to the first embodiment, it is possible to provide appropriate information that contributes to the improvement of the customer service method for each business type and place where customer service is performed. For example, the impression of “brightness” has a strong impact on sales when serving customers at a retail store such as a convenience store, and the impression of “politeness” has a greater impact on sales than “brightness” when serving customers at a hotel. It is possible. In addition, the impression of “strength” has little effect on sales when serving customers at retail stores and hotels, but the impression of “strength” has a strong effect on sales when serving customers at greengrocers and pubs. Can be considered.

[Second Embodiment]
In the second embodiment described above, the time information is used for associating the impression index data with the data indicating the purchase behavior of the customer. That is, the purchase behavior of the customer performed at a time close to the time when the store clerk's behavior that is the basis of the estimated impression was associated. However, in the association based only on the closeness of the time, it is not certain whether the behavior of the store clerk that is the basis of the estimated impression is related to the purchase behavior associated with the impression. In other words, in the association based only on the proximity of time, the customer estimated to have received the impression represented by the index data does not necessarily match the customer who performed the purchase behavior associated with the index data. .

  Therefore, the second embodiment appropriately associates the behavior of the clerk who is the basis of the estimated impression with the purchase behavior of the customer. Thereby, 2nd embodiment can analyze the relationship between the impression which a customer receives from the behavior of a store clerk, and purchase behavior of a customer with higher accuracy than the first embodiment. Hereinafter, the customer service data processing device and the customer service data processing method in the second embodiment will be described focusing on the contents different from the first embodiment. In the following description, the same contents as those in the first embodiment are omitted as appropriate.

[Processing configuration]
The customer service data processing apparatus 10 in the second embodiment is the same as the processing configuration of the first embodiment shown in FIG. The processing content of each processing module shown below is different from the first embodiment.

  The generation unit 26 associates both data so that the impression index data and the data indicating the purchase behavior belong to the same customer. In other words, the generation unit 26 associates both data so that the customer estimated to have received the impression by the impression estimation unit 25 matches the customer who has performed the purchase behavior associated with the index data of the impression. Thereby, for example, when the customer A enters the store and the store clerk greets the customer with “I welcome you”, the greeting is related to the sales by the customer A and the customer A's store stay time. In addition, when the sales clerk makes a sales talk to the customer B that the melon bread is baked, the sales talk is associated with the sales by the customer B and the customer B's store stay time.

  There are a plurality of association methods by the generation unit 26 as follows. Hereinafter, each association method will be described. However, the association method by the generation unit 26 is not limited to the example given below.

<First association method>
The generation unit 26 records the customer position data acquired by the first acquisition unit 21, the association between the impression index data and the time information described in the first embodiment, the data indicating the purchase behavior, and the time information. Is used to associate impression index data with data indicating purchase behavior. The history of customer location data includes a plurality of pairs of location data and time information, and further includes customer identification information (customer ID). The time information associated with the impression index data is substantially the same as the time when the customer service behavior detected by the detection unit 24 is performed.

  The generation unit 26 associates the customer ID with the impression index data based on the association between the impression index data and the time information and the history of the customer position data. At this time, the generation unit 26 specifies the ID of the customer estimated to have received the impression based on the position of the customer at the time associated with the impression index data. For example, the generation unit 26 associates the ID of the customer who was near the entrance of the store at 10 o'clock with respect to the index data of the impression corresponding to “welcome” issued by the store clerk at 10 o'clock.

  FIG. 10 is a diagram illustrating an example of association information generated by the generation unit 26 in the second embodiment. In the example of FIG. 10, the generation unit 26 associates impression index data and a customer ID for each utterance detected by the detection unit 24. According to the example of FIG. 10, the impression index data (“brightness = 92”, “energeticity = 80”, etc.) estimated from the greeting utterance “I welcome you” uttered at 7:05:10 , “Customer ID = 5”.

  Further, the generation unit 26 associates the customer ID with the data indicating the purchase behavior based on the association between the data indicating the purchase behavior and the time information and the history of the position data of the customer. At this time, the generation unit 26 specifies the ID of the customer who is the subject of the purchase behavior based on the position of the customer at the time associated with the data indicating the purchase behavior. For example, when the customer's store stay time is used as data indicating purchase behavior, the generation unit 26 enters the store at 10 o'clock and leaves the store at 10:10 with respect to the store stay time (10 min) of the customer. Associate customer IDs near the store entrance at 10:00 and 10:10.

  FIG. 11 is a diagram illustrating an example of association information generated by the generation unit 26 in the second embodiment. In the example of FIG. 11, the generation unit 26 associates a customer ID with a sales amount for each settlement, which is data indicating purchase behavior. According to the example of FIG. 11, the customer's sales amount “1200 yen” that has been settled at 7:47:56 is associated with the customer ID (8) of the customer who was at the cash register at about 7:47. Yes. The customer ID can also be acquired from the customer information of the point card presented at the cash register.

《Second association method》
In addition to the information used in the first association method described above, the generation unit 26 further uses the history of the position data of the clerk acquired by the second acquisition unit 23 to generate the index data and impression behavior data. Associate with. The history of the clerk position data includes a plurality of pairs of position data and time information, and further includes clerk identification information (clerk ID). In this method, the generation unit 26 associates a customer ID with the impression index data based on the association between the impression index data and the time information, the customer position data history, and the store clerk position data history. . The generation unit 26 specifies the ID of the customer who is estimated to have received the impression based on the positional relationship between the customer and the store clerk at the time associated with the impression index data. For example, the generation unit 26 associates the ID of the customer who was within a predetermined distance from the clerk at 11:00 with respect to the index data of the impression of the clerk's facial expression at 11:00.

  In this case, the impression index data needs to be associated with the time information and the ID of the clerk who took the action that is the source (input) of the impression estimation. Therefore, the impression estimation unit 25 specifies the clerk's ID indicated by data (sound, image, etc.) to which the impression estimation model is applied. For example, when acquiring the clerk behavior data, the second acquisition unit 23 specifies the clerk ID indicated by the clerk behavior data based on the data acquisition source. The second acquisition unit 23 can identify the clerk ID corresponding to the acquired clerk behavior data by referring to the correspondence information between the clerk ID and the ID of the microphone carried by the clerk. Moreover, the 2nd acquisition part 23 can also acquire ID of the salesclerk obtained from an IC tag with the position data of a salesclerk by making a salesclerk carry an IC tag. In such a case, the impression estimation unit 25 may use the clerk ID specified by the second acquisition unit 23.

  As another example, the impression estimation unit 25 applies the speaker verification technique or the face identification technique to the data (sound, image, etc.) to which the impression estimation model is applied, so that the clerk's ID indicated by the data is applied. Is identified. When the speaker verification technique is used, the impression estimation unit 25 holds an utterance model (voice feature amount) for each clerk in advance. Further, when using face identification information, the impression estimation unit 25 holds a face image model (image feature amount) for each clerk in advance.

  FIG. 12 is a diagram illustrating an example of association information generated by the generation unit 26 in the second embodiment. In the example of FIG. 12, the generation unit 26 associates the impression index data, the clerk ID, and the customer ID for each utterance detected by the detection unit 24. According to the example of FIG. 12, impression index data (“brightness = 92”, “brightness = 92”, “ And “customer ID = 5” are associated with each other.

  The association between the data indicating the purchase behavior and the customer ID may be the same as in the first association method. In this manner, the generation unit 26 associates the impression index data and the data indicating the purchase behavior with the customer ID. The generation unit 26 stores information associated with the customer ID in the storage unit 30. Here, the generating unit 26 stores the associated impression index data and data indicating purchase behavior, and associates the impression index data with the customer ID, and the association information between the data indicating the purchase behavior and the customer ID. It is not necessary to store.

  Further, when the second association method described above is taken, the generation unit 26 generates association information between the clerk ID and the customer ID estimated to have received the impression by the behavior of the clerk in the impression index data. can do. Furthermore, based on the association information, the generation unit 26 can associate the store clerk ID with the impression index data and the purchase behavior of the customer. According to this association information, an analysis of the relationship between the impression estimated to be received by the customer and the purchase behavior of the customer can be performed for each clerk. For example, salesclerk A can analyze the sales of target customers who have received a large “brightness” in greetings and large sales for salesclerk B, but salesclerk B can analyze the sales of the target customers in whom the “brightness” of greetings is small.

  FIG. 13 is a diagram illustrating an output by the output unit 29. In FIG. 13, a graph indicating the correlation between the index data of the impression of the salesclerk greeting and the data indicating the purchase behavior of the customer and the correlation coefficient thereof are output. The graph F50 shows the correlation between the index data of the “brightness” impression of the greeting and the sales by the customer who is the subject of the greeting. According to this graph F50, it can be seen that there is a relationship that sales from customers who have received a bright greeting are large. The graph F51 shows the correlation between the index data of the impression of “strength” of the greeting and the sales by the customer who is the subject of the greeting. According to the graph F51, it can be seen that there is almost no relationship between the “strength” of the greeting and the sales. The graph F52 shows the correlation between the index data of the “brightness” impression of the greeting and the customer's stay time in the store that is the subject of the greeting. According to the graph F52, it can be seen that a customer who receives a bright greeting has a longer staying time.

[Operation example / Service data processing method]
The order of each process of the customer service data analysis method in the second embodiment is the same as that in the first embodiment shown in FIG. In the second embodiment, the associating method in (S95) is different from the first embodiment. This association method is as described above.

[Operation and effect of the second embodiment]
As described above, in the second embodiment, impression index data and purchase behavior are shown based on impression index data and time information associated with customer purchase behavior, customer location data, store clerk location data, and the like. The impression index data and the data indicating the purchase behavior are associated with each other so that the data belongs to the same customer. Thereby, according to 2nd embodiment, the relationship between the impression which a customer receives from a store clerk's action, and the purchase action of a customer can be analyzed with high precision. Specifically, according to the second embodiment, both data are associated with each other so that the impression index data and the data indicating the purchase behavior belong to the same customer, and therefore more accurate than the first embodiment. A correlation coefficient can be obtained.

  Furthermore, according to the second association method described above, it is possible to generate association information between a store clerk ID and a customer ID estimated to have received the impression by the behavior of the store clerk in the impression index data. In addition, the clerk ID can be associated with the impression index data and the purchase behavior of the customer. According to this, the analysis of the relationship between the impression estimated to be received by the customer and the purchase behavior of the customer can be performed for each clerk.

[Modification]
The processing device 10 in each of the above embodiments may not include the analysis unit 27 and the advice generation unit 28. In this case, based on the information stored in the storage unit 30, the processing device 10 (the output unit 29) outputs the impression index data and the data indicating the purchase behavior of the customer in an associated state or an associated state. , Do not output correlation coefficient or advice information. In this case, the service data processing method shown in FIG. 9 does not include (S96) and (S97).

[Third embodiment]
Hereinafter, the service data processing apparatus and the service data processing method according to the third embodiment will be described with reference to FIGS. 14 and 15. The third embodiment may be a program that causes at least one computer (CPU) to execute the service data processing method, or a recording medium that can be read by the at least one computer that records such a program. It may be.

  FIG. 14 is a diagram conceptually illustrating a processing configuration example of the customer service data processing device 100 according to the third embodiment. As shown in FIG. 14, the customer service data processing apparatus 100 includes an estimation unit 101, a data acquisition unit 102, and a generation unit 103. The customer service data processing apparatus 100 shown in FIG. 14 has, for example, the same hardware configuration as the above-described processing apparatus 10 shown in FIG. 1, and the program is processed by the CPU 11 in the same manner as the processing apparatus 10. Each processing module described above is realized. However, the hardware configuration of the customer service data processing device 100 is not limited, and the display device 15 and the input device 16 may not be connected to the customer service data processing device 100.

  Based on the data indicating the store clerk at the time of customer service, the estimation unit 101 calculates the index data of the impression that the customer receives from the store clerk. A specific example of data indicating a clerk at the time of customer service is the clerk behavior data described above. The format and content of this data are not limited as long as the data indicates a store clerk at the time of customer service. The type of impression that is estimated (bright, fine, etc.) is not limited. However, it is desirable that the impression type is set to an impression having a great influence on the customer service purpose. Further, as described above, the calculated index data of the impression indicates the type of the impression (bright, energy, etc.), the presence / absence of the impression, or the degree of the intensity of the impression. The calculation method of the impression index data by the estimation unit 101 is the same as the impression estimation unit 25 described above.

  The data acquisition unit 102 acquires data indicating customer purchase behavior. The format and content of the acquired data are not limited as long as it indicates the purchase behavior of the customer. As described above, the customer's purchasing behavior is related not only to the customer actually purchasing something but also to the customer service in the customer's behavior from when the customer enters the store until when the customer leaves the store. Includes a variety of potential actions. The data acquisition method by the data acquisition unit 102 is the same as that of the data acquisition unit 22 described above.

  The generation unit 103 generates information that includes the impression index data calculated by the estimation unit 101 and the data indicating the purchase behavior of the customer acquired by the data acquisition unit 102 in an associated state or an associated state. The generation unit 103 may generate information in which impression index data and customer purchase behavior data are directly associated with each other. Further, the generation unit 103 may generate information including impression index data and data indicating the purchase behavior of the customer so that the viewer can make an association or can make an association by information processing. Furthermore, the generation unit 103 may generate both the former information and the latter information. In each of the above-described embodiments, the time information and the customer ID are exemplified as the information that associates or enables the association between the impression index data and the purchase behavior of the customer. The information for association is not limited to such an example. The information generation method by the generation unit 103 is the same as that of the generation unit 26 described above.

  FIG. 15 is a flowchart illustrating an operation example of the customer service data processing apparatus 100 according to the third embodiment. As shown in FIG. 15, the service data processing method in the third embodiment is executed by at least one computer such as the service data processing apparatus 100. For example, each illustrated process is executed by each processing module included in the customer service data processing apparatus 100. Since each process is the same as the processing content of each processing module described above that the customer service data processing apparatus 100 has, details of each process are omitted as appropriate.

The customer service data processing method in the present embodiment includes (S151), (S152), and (S153).
In (S151), the customer service data processing device 100 calculates the index data of the impression that the customer receives from the store clerk based on the data indicating the store clerk at the time of customer service.
In (S152), the customer service data processing apparatus 100 acquires data indicating the purchase behavior of the customer.
In (S153), the customer service data processing device 100 includes the impression index data calculated in (S151) and the data indicating the purchase behavior of the customer acquired in (S152) in an associated state or an associated state. Generate information. The generated information may be stored in the customer service data processing apparatus 100, may be sent to another apparatus, or may be stored in a portable recording medium.

  According to the third embodiment, since the information including the index data of the impression and the data indicating the purchase behavior of the customer in a state that can be associated or in an associated state is generated, the customer service method is similar to the above-described embodiments. Can provide useful information for improvement.

In the above-described embodiments, the following examples are mainly used.
(1) Detected customer service behavior = specific utterance such as greeting or sales talk (2) input data for impression estimation = voice data indicating the specific utterance (3) time information corresponding to impression index data = specific utterance Voice time (4) Data indicating purchase behavior = sales amount and store stay time (5) Time information corresponding to data indicating purchase behavior = Checkout time, store entry time, store exit time (6) Information for association = Time information (first embodiment), customer ID (second embodiment)
Hereinafter, other examples not specified in the above-described embodiments will be described as examples. However, the above-mentioned content is not limited to the content of the following Examples.

[First embodiment]
(1) Detected customer service behavior = specific facial expression of a store clerk (smile, grimacing, etc.)
(2) Impression estimation input data = image data representing the detected specific facial expression (3) Time information corresponding to impression index data = specific facial expression occurrence time (4) to (6) = Same processing device 10 (first The second obtaining unit 23) obtains image data including the clerk's face as clerk action data. The processing device 10 (detection unit 24) applies a face detection and facial expression recognition technique to the image data, and detects a specific facial expression of the clerk as a customer service action. Processing device 10 (impression estimation unit 25 calculates index data of an impression estimated to be received by the customer from the specific facial expression by applying a face impression estimation model to the image data including the detected specific facial expression. (First estimation method) The processing device 10 (impression estimation unit 25) acquires time information at which the specific facial expression appears as time information corresponding to the calculated index data of the impression. The generation unit 26) associates the impression index data and the time information and stores them in the storage unit 30.

  FIG. 16 is a diagram illustrating an example of association between impression index data regarding a specific facial expression and time information. In the example of FIG. 16, the processing device 10 (the generation unit 26) stores the generation time, the expression type, and the impression index data in association with each detected specific facial expression in the storage unit 30.

[Second Example]
(1) Detected customer service behavior = clerk's utterance (including specific utterance and response utterance)
(2) Impression estimation input data = image data representing detected clerk's facial expression during utterance (3) time information corresponding to impression index data = occurrence time of the clerk's facial expression (4) to (6) = The processing device 10 (second acquisition unit 23) acquires the clerk's speech data and image data including the clerk's face as clerk action data. The processing device 10 (detection unit 24) detects a clerk's utterance such as a specific utterance or a response utterance as a customer service action from the acquired utterance voice data. The processing device 10 (impression estimation unit 25) applies the face impression estimation model to the image data representing between the start time and the end time of the detected utterance, so that the customer receives from the facial expression of the clerk during the utterance Is calculated (second estimation method). The processing device 10 (impression estimation unit 25) acquires time information of image data to which the face impression estimation model is applied as time information corresponding to the calculated impression index data. As illustrated in FIG. 16, the processing device 10 (the generation unit 26) stores the impression index data and the time information in the storage unit 30 in association with each other.

[Third embodiment]
(1) Detected customer service behavior = clerk's accounting work (2) Impression estimation input data = detected clerk's utterance voice data during accounting work (3) Time information corresponding to impression index data = the clerk's The utterance time (4) to (6) = same as above The processing device 10 (second acquisition unit 23) uses the operation history (login history) information of each clerk of the cash register terminal and the utterance voice data of the clerk as clerk behavior data. get. The processing device 10 (detection unit 24) extracts the start time and end time of using the cashier terminal from the information of the operation history, and detects the period as the accounting work of the store clerk. The processing device 10 (impression estimation unit 25) applies the voice impression estimation model to the utterance voice data between the detected start time and end time of the detected accounting operation, so that the detected clerk during the accounting operation The index data of the impression estimated that the customer receives from the utterance is calculated (second estimation method). The processing device 10 (impression estimation unit 25) acquires time information of the speech data to which the voice impression estimation model is applied, as time information corresponding to the calculated index data of the impression.

  In the third embodiment, image data representing the detected facial expression of the store clerk during the accounting work may be used as the input data for impression estimation. In this case, the processing device 10 (second acquisition unit 23) acquires information on an operation history (login history) for each clerk of the cash register terminal and image data representing the clerk's facial expression as clerk action data. The processing device 10 (detection unit 24) detects the clerk's accounting work as described above. The processing device 10 (impression estimation unit 25) applies the face impression estimation model to the image data between the detected start time and end time of the detected check operation, thereby detecting the detected facial expression of the clerk during the check operation The index data of the impression estimated to be received by the customer is calculated (second estimation method).

[Fourth embodiment]
(1) Detected customer service behavior = store clerk's store guidance (2) impression estimation input data = clerk utterance voice data in the detected store guide (3) time information corresponding to impression index data = the clerk's The utterance time (4) to (6) = same as above The processing device 10 (second acquisition unit 23) acquires the position data history of the clerk in the store and the utterance voice data of the clerk as clerk action data. Moreover, the processing apparatus 10 (first acquisition unit 21) acquires a history of customer position data as customer behavior data. The processing device 10 (detection unit 24) continues the state where the clerk and the customer exist within a certain distance based on the clerk's position data history and the customer's position data history, and both move. Is detected as store guidance. The processing device 10 (impression estimation unit 25) applies the voice impression estimation model to the utterance voice data between the detected start time and end time of the store guide, thereby making it possible to detect the utterance of the store clerk in the store guide. Index data of impression estimated to be received by the customer is calculated (second estimation method). The processing device 10 (impression estimation unit 25) acquires time information of the speech data to which the voice impression estimation model is applied, as time information corresponding to the calculated index data of the impression.

Each above-mentioned embodiment and modification can be combined in the range in which the contents do not conflict.
Part or all of the above embodiments and modifications may be specified as follows. However, the above embodiments and modifications are not limited to the following descriptions.

1. An estimation means for calculating index data of impressions estimated to be received by the customer from the store clerk based on data indicating the store clerk at the time of customer service;
Data acquisition means for acquiring data indicating customer purchase behavior;
Generating means for generating information including the calculated index data of the impression and the acquired data indicating the purchase behavior of the customer in an associated state or an associated state;
A customer service data processing apparatus.
2. A detecting means for detecting the customer service behavior of the store clerk,
Further comprising
The estimating means calculates the index data of the impression using the data indicating the store clerk when the detected customer service behavior is performed as the data indicating the store clerk at the time of customer service,
1. The customer service data processing device described in 1.
3. The generating means calculates at least one of presence / absence of detection of customer service behavior of the store clerk, type of the customer service behavior of the detected store clerk, and content identification data identifying the content of the detected customer service behavior of the store clerk. Generating information related to the impression index data or the acquired customer purchase behavior data,
2. The customer service data processing device described in 1.
4). The detection means detects the clerk's utterance to the customer as the customer service behavior from the voice data including the clerk's utterance,
The estimation means, from the voice data of the detected clerk's speech, brightness, cheerfulness, strength, coolness, politeness, calmness, cuteness, honesty, coldness, fluency, youth, Calculating index data indicating the presence or strength of at least one impression of elegance and tenderness;
2. Or 3. The customer service data processing device described in 1.
5. The detection means detects, from the voice data including the clerk's speech, the clerk's speech corresponding to a specific expression representing a greeting or sales talk as the customer service behavior,
The estimation means calculates index data indicating the presence / absence or strength of at least one of brightness, energy, strength, politeness, and coldness from the detected speech data of the clerk's utterance. ,
2. To 4. The customer service data processing device according to any one of the above.
6). The detection means detects, from the image data including the clerk, a specific facial expression of the clerk for the customer as the customer service behavior,
The estimation means calculates index data of an impression estimated to be received by a customer from the facial expression, from image data including the facial expression of the clerk detected as the specific facial expression.
2. To 5. The customer service data processing device according to any one of the above.
7). The detection means detects the first customer service behavior of the store clerk for the customer,
The estimating means receives the customer from the second customer service behavior based on the data indicating the second customer service behavior of the store clerk, which is different from the first customer service behavior when the detected first customer service behavior is performed. Calculating the estimated impression index data,
2. To 6. The customer service data processing device according to any one of the above.
8). The estimation means calculates the index data of the impression from the data indicating the first customer service behavior of the store clerk and the data indicating the second customer service behavior when the detected first customer service behavior is performed. To
7). The customer service data processing device described in 1.
9. The data acquisition means indicates the purchase behavior of at least one of a product purchased by a customer, an amount purchased by the customer, a customer's store stay time, a customer's in-store travel route, a customer's store visit frequency, and a customer's store visit count. Obtain as the data,
1. To 8. The customer service data processing device according to any one of the above.
10. An output means for outputting the calculated index data of the impression and the acquired data indicating the purchase behavior of the customer in an associated state or an associated state;
1 is further provided. To 9. The customer service data processing device according to any one of the above.
11. The estimation means calculates index data of a plurality of impressions respectively indicating the presence or absence or strength of each impression based on data indicating a clerk at the time of customer service.
1. To 10. The customer service data processing device according to any one of the above.
12 Analyzing means for calculating a correlation coefficient between the calculated index data of the impression and the acquired data indicating the purchase behavior of the customer;
1 is further provided. To 11. The customer service data processing device according to any one of the above.
13. Advice generating means for generating advice information on customer service based on the calculated correlation coefficient;
Is further provided. The customer service data processing device described in 1.
14 Output means for outputting at least one of the generated correlation information and the correlation coefficient calculated for the impression;
Further comprising: The customer service data processing device described in 1.
15. The generation means associates identification information of a customer estimated to have received the impression with the index data of the impression based on a history of the position data of the customer and time information corresponding to the calculated index data of the impression, Based on a history of customer position data and time information corresponding to the acquired data indicating the purchase behavior of the customer, associating the identification information of the customer who is the subject of the purchase behavior with the data indicating the purchase behavior,
1. To 14. The customer service data processing device according to any one of the above.
16. The generating means further associates identification information of a store clerk who gave the impression to the customer to the calculated index data of the impression.
15. The customer service data processing device described in 1.

17. In the service data processing method executed by at least one computer,
Based on data showing the store clerk at the time of customer service, calculate the index data of the impression that the customer is likely to receive from the store clerk,
Get data that shows customer buying behavior,
Generating information including the calculated index data of the impression and the acquired data indicating the purchase behavior of the customer in an associated state or an associated state;
Service data processing method including things.
18. Detect customer service behavior
Further including
The calculation of the index data calculates the index data of the impression using the data indicating the store clerk when the detected customer service behavior is performed as the data indicating the store clerk at the time of customer service.
17. Service data processing method described in 1.
19. At least one of the presence / absence of detection of the customer service behavior of the store clerk, the type of the customer service behavior of the detected store clerk, and the content of the detected customer service behavior of the store clerk is the index data of the calculated impression Or generating information associated with the acquired customer purchase behavior data;
Further includes: Service data processing method described in 1.
20. The detection of the customer service behavior is performed by detecting a customer clerk's utterance as a customer service behavior from voice data including a clerk's speech.
The calculation of the index data is based on the detected voice data of the clerk's utterance, brightness, spirit, strength, coolness, politeness, calmness, cuteness, honesty, coldness, fluency, Calculating index data indicating the presence or strength of at least one impression of youth, elegance, and tenderness;
18. Or 19. Service data processing method described in 1.
21. The detection of the customer service behavior is performed by detecting an utterance of a clerk corresponding to a specific expression representing a greeting or a sales talk from the voice data including the utterance of the clerk as the customer service behavior,
The calculation of the index data is based on the detected clerk's utterance voice data, and indicates index data indicating the presence or strength of at least one of brightness, spirit, strength, politeness, and coldness. To calculate,
18. To 20. The customer service data processing method according to any one of the above.
22. The detection of the customer service behavior detects a specific facial expression of a store clerk for a customer as the customer service behavior from image data including a store clerk.
The calculation of the index data is based on image data including a store clerk's facial expression detected as the specific facial expression, and calculates index data of an impression that a customer is expected to receive from the facial expression.
18. To 21. The customer service data processing method according to any one of the above.
23. The detection of the customer service behavior detects a first customer service behavior of a store clerk for a customer,
The calculation of the index data is based on data indicating the second customer service behavior of the store clerk, which is different from the first customer service behavior when the detected first customer service behavior is performed. Calculate the index data of the impression that the customer is expected to receive,
18. To 22. The customer service data processing method according to any one of the above.
24. The calculation of the index data is based on the index of the impression from the data indicating the first customer service behavior of the store clerk and the data indicating the second customer service behavior when the detected first customer service behavior is performed. Calculate the data,
23. Service data processing method described in 1.
25. The acquisition of the data indicating the purchase behavior is at least one of a product purchased by a customer, an amount purchased by the customer, a customer's store stay time, a customer's in-store travel route, a customer's store visit frequency, and a customer's store visit count. As the data indicating the purchase behavior,
17. To 24. The customer service data processing method according to any one of the above.
26. Outputting the index data of the calculated impression and the data indicating the acquired purchase behavior of the customer in a state where they can be associated or in an associated state;
Further includes: To 25. The customer service data processing method according to any one of the above.
27. The calculation of the index data is based on data indicating a store clerk at the time of customer service, and calculates index data of a plurality of impressions indicating the presence or absence or strength of each impression,
17. To 26. The customer service data processing method according to any one of the above.
28. Calculating a correlation coefficient between the calculated index data of the impression and the acquired customer purchase behavior data;
Further includes: To 27. The customer service data processing method according to any one of the above.
29. Based on the calculated correlation coefficient, advice information about customer service is generated.
Further includes: Service data processing method described in 1.
30. Outputting at least one of the generated correlation information and the correlation coefficient calculated for the impression;
Further includes: 29. Service data processing method described in 1.
31. Based on the history information of the customer's location data and the time information corresponding to the calculated index data of the impression, the identification data of the customer estimated to have received the impression is associated with the index data of the impression,
Based on a history of customer position data and time information corresponding to the acquired data indicating the purchase behavior of the customer, associating the identification information of the customer who is the subject of the purchase behavior with the data indicating the purchase behavior,
Further includes: To 30. The customer service data processing method according to any one of the above.
32. Further associates with the calculated impression index data the identification information of the clerk who gave the impression to the customer;
Further includes: Service data processing method described in 1.

33. 17. To 32. A program that causes at least one computer to execute the service data processing method according to any one of the above.

10,100 Hospitality data processing device (processing device)
11 CPU
DESCRIPTION OF SYMBOLS 12 Memory 21 1st acquisition part 22,102 Data acquisition part 23 2nd acquisition part 24 Detection part 25 Impression estimation part 26,103 Generation part 27 Analysis part 28 Advice generation part 29 Output part 30 Storage part 101 Estimation part

Claims (13)

An estimation means for calculating index data of impressions estimated to be received by the customer from the store clerk based on data indicating the store clerk at the time of customer service;
Data acquisition means for acquiring data indicating customer purchase behavior;
Generating means for generating information including the calculated index data of the impression and the acquired data indicating the purchase behavior of the customer in an associated state or an associated state;
A customer service data processing apparatus. A detecting means for detecting the customer service behavior of the store clerk,
Further comprising
The estimating means calculates the index data of the impression using the data indicating the store clerk when the detected customer service behavior is performed as the data indicating the store clerk at the time of customer service,
The service data processing apparatus according to claim 1. The detection means detects the clerk's utterance to the customer as the customer service behavior from the voice data including the clerk's utterance,
The estimation means, from the voice data of the detected clerk's speech, brightness, cheerfulness, strength, coolness, politeness, calmness, cuteness, honesty, coldness, fluency, youth, Calculating index data indicating the presence or strength of at least one impression of elegance and tenderness;
The service data processing apparatus according to claim 2. The detection means detects, from the image data including the clerk, a specific facial expression of the clerk for the customer as the customer service behavior,
The estimation means calculates index data of an impression estimated to be received by a customer from the facial expression, from image data including the facial expression of the clerk detected as the specific facial expression.
The customer service data processing device according to any one of claims 2 to 3. The detection means detects the first customer service behavior of the store clerk for the customer,
The estimating means receives the customer from the second customer service behavior based on the data indicating the second customer service behavior of the store clerk, which is different from the first customer service behavior when the detected first customer service behavior is performed. Calculating the estimated impression index data,
The customer service data processing device according to any one of claims 2 to 4. The data acquisition means indicates the purchase behavior of at least one of a product purchased by a customer, an amount purchased by the customer, a customer's store stay time, a customer's in-store travel route, a customer's store visit frequency, and a customer's store visit count. Obtain as the data,
The customer service data processing apparatus according to any one of claims 1 to 5. An output means for outputting the calculated index data of the impression and the acquired data indicating the purchase behavior of the customer in an associated state or an associated state;
The customer service data processing device according to any one of claims 1 to 6, further comprising: The estimation means calculates index data of a plurality of impressions respectively indicating the presence or absence or strength of each impression based on data indicating a clerk at the time of customer service.
The customer service data processing apparatus according to any one of claims 1 to 7. Analyzing means for calculating a correlation coefficient between the calculated index data of the impression and the acquired data indicating the purchase behavior of the customer;
The customer service data processing device according to any one of claims 1 to 8, further comprising: Advice generating means for generating advice information on customer service based on the calculated correlation coefficient;
The customer service data processing device according to claim 9, further comprising: Output means for outputting at least one of the generated correlation information and the correlation coefficient calculated for the impression;
The customer service data processing device according to claim 10, further comprising: In the service data processing method executed by at least one computer,
Based on data showing the store clerk at the time of customer service, calculate the index data of the impression that the customer is likely to receive from the store clerk,
Get data that shows customer buying behavior,
Generating information including the calculated index data of the impression and the acquired data indicating the purchase behavior of the customer in an associated state or an associated state;
Service data processing method including things.   A program for causing at least one computer to execute the service data processing method according to claim 12.

Images