JP6653413B2 - Purchase analysis device and purchase analysis method - Google Patents

Description

  The present disclosure particularly relates to a purchase analysis device and a purchase analysis method for analyzing a purchase behavior of a customer in a store.

  Conventionally, by calculating the residence time at each point in the customer's store and linking it with the product purchase information, information such as that the customer has not been purchased despite passing through the place where the product is placed There is a device that obtains the information and uses this information as a judgment material (for example, see Patent Document 1).

  Patent Literature 1 discloses a residence time measuring unit that measures a residence time in which a customer stays in front of each shelf in a store for each shelf, a storage unit that stores purchased commodity information indicating a commodity purchased by each customer at the time of settlement, and the like. Is provided.

  As a result, customer purchase information such as the number of purchasers who purchased the products on the product shelf and the average residence time and the number of people who did not purchase the products on the product shelf and the average residence time are stored for each product shelf. Based on a predetermined reference value set for each product category or for each shelf on which a product is placed, it is determined whether or not the stored average residence time has exceeded a reference value. It is configured to include a warning generating means for generating a warning when it is determined that the value has been exceeded. With this configuration, it is possible to estimate a product that the customer wanted but did not purchase at the store using the customer's flow line information and POS data. Specifically, shelves that have not been purchased yet and have a long average residence time and a large number of cases are extracted as shelves that need to be improved in terms of missing parts, assortment of goods, and the like.

Japanese Patent No. 4068505

  However, in the conventional system, since a certain standard is used for the residence time, for example, the difference between a person who frequently visits the store and a person who does not visit often may not be known, and the cause may not be accurately specified. . In the customer's purchase pattern, when a product category to be purchased and a product brand are determined before visiting the store (hereinafter referred to as planned purchase), the product category is determined but the product brand is determined by looking at the product when visiting the store. In some cases, and at the time of visiting the store, the product category or the product brand is not determined, and there is a case where the purchase intention is recalled in the store (hereinafter, referred to as unplanned purchase). In the case of planned purchase, the dwell time generally tends to be shorter than in the case of unplanned purchase. Since these purchase patterns are different for each customer, each product, and each visit, when a certain standard is used for the dwell time for each shelf, there is a possibility that the cause cannot be accurately identified as in the case of the visit frequency. .

  An object of the present disclosure is to provide a purchase analysis device and a purchase analysis method that can more accurately identify a cause from purchase information of individual customers and information of a camera.

A purchase analysis device according to the present disclosure is a purchase analysis device connected to a purchase information management device and a camera arranged in a store having a plurality of sales areas, and identifies a customer based on data received from the camera. A person specifying unit, a stay detecting unit that detects a stay time of each of the specified customers for each sales area, and purchase information including merchandise information during the store time of the specified customer from entering to leaving the store. A purchase information acquisition unit that reads the staying sales area in which the specified customer has stayed for the first time or more and the corresponding product information, and the purchase information, and the product information that is not included in the purchase information corresponds to the purchase information; and a non-purchase region detecting section that detects a non-purchase region of residence sales area, the residence time of the non-purchase area are classified into whether the second time or more residence time provided for each customer that the identified And a statistics unit for statistics.

The purchase analysis method according to the present disclosure is a purchase analysis method executed by a CPU of a purchase analysis device connected to a camera and a purchase information management device arranged in a store having a plurality of sales areas, wherein the data of the camera is Identifying the customer based on the following, detecting the residence time of each of the identified customers in each sales area, and purchasing information including the product information at the entry time from entry to exit of the identified customer. And reading the staying sales area in which the specified customer has stayed for the first time or more and the corresponding merchandise information, and the purchase information, and the merchandise information not included in the purchase information corresponds to the Detecting a non-purchase area that is a stagnant sales area; and determining whether or not the linger time of the non-purchase area is a linger time equal to or longer than a second time provided for each of the specified customers. And statistics.

  The purchase analysis device and the purchase analysis method according to the present disclosure can more accurately identify the cause from purchase information of individual customers and information of a camera.

FIG. 2 is a view for explaining the configuration of a purchase analysis system according to the first embodiment. The figure which shows the layout in a shop concerning Embodiment 1. FIG. 1 is a block diagram showing a configuration of a purchase analysis device according to a first embodiment. Flow chart of overall purchase analysis according to the first embodiment FIG. 4 is a diagram showing face search result data according to the first embodiment; Flow chart for creating face information data for each customer according to the first embodiment FIG. 3 is a diagram illustrating face information data for each customer according to the first embodiment; Flow chart for creating visit data from face information data for each customer according to the first embodiment FIG. 7 is a diagram showing visit data for each customer according to the first embodiment; The figure which shows the detection area in the store concerning Embodiment 1. The figure which shows the detection area data in the store concerning Embodiment 1. FIG. 7 is a diagram showing shelf position data in a store according to the first embodiment. FIG. 6 is a diagram showing merchandise category data for each shelf according to the first embodiment; Flowchart for detecting residence time according to the first embodiment FIG. 4 is a diagram showing residence time data for each visit of a customer according to the first embodiment; FIG. 3 is a diagram showing purchase data according to the first embodiment. Flow chart for linking residence time data and purchase data according to the first embodiment FIG. 8 is a diagram showing non-purchase residence time data for each visit of a customer according to the first embodiment; Flow chart for analysis according to the first embodiment FIG. 6 is a diagram showing stay time and purchase pattern data for each customer and each shelf according to the first embodiment. FIG. 7 is a diagram showing non-purchase count data for each shelf and each purchase pattern according to the first embodiment; The figure which shows the buying-order of unplanned purchase and the non-purchase frequency data for every shelf concerning Embodiment 1. FIG. 5 is a diagram showing a display example of an analysis result according to the first embodiment. FIG. 7 is a diagram showing purchase / non-purchase residence time data for each visit of a customer according to the first embodiment; FIG. 10 is a diagram showing stay time and purchase pattern data for each customer and each shelf according to the second embodiment. The figure which shows the number-of-times-of-purchase data at the time of unplanned purchase for every shelf concerning Embodiment 2, and the reduction rate of stay time. The figure which shows the planned purchase frequency and the unplanned purchase frequency for every shelf concerning embodiment.

  Hereinafter, embodiments will be described in detail with reference to the drawings as appropriate. However, an unnecessary detailed description may be omitted. For example, a detailed description of a well-known item or a redundant description of substantially the same configuration may be omitted. This is to prevent the following description from being unnecessarily redundant and to facilitate understanding by those skilled in the art.

  The applicant provides the accompanying drawings and the following description so that those skilled in the art can fully understand the present disclosure, and it is not intended that the invention be limited to the claimed subject matter. Absent.

(Embodiment 1)
Hereinafter, Embodiment 1 will be described with reference to FIGS.
[1-1. Constitution]
FIG. 1 is a diagram illustrating a configuration of a purchase analysis system 1 according to the first embodiment. The purchase analysis system 1 includes a purchase analysis device 2, a face detection camera 3, a stay detection camera 4, a payment device (POS, purchase information management device) 5, and a display device 6. The purchase analysis system 1 is configured to perform aggregation and analysis in the purchase analysis device 2 based on information from the face detection camera 3, the stay detection camera 4, and the settlement device 5, and display the aggregation and analysis results on the display device 6. It has become.

  FIG. 2 is a diagram illustrating an in-store layout according to the first embodiment. Inside the store, an entrance camera C31, which is one of the face detection cameras 3 for identifying a visitor above the front of the store entrance, for example, capturing an image in a specific direction, is installed. In addition, one of the face detection cameras 3 for identifying a customer who has performed the settlement, for example, pre-registration cameras C32 and C33 for photographing a specific direction are installed above the cash register R1 and cash register R2 in the store. Have been.

In the upper part of the store, there is one of the stay detection cameras 4 for detecting the flow line of the customer in the store and the stay time of each area. The stay detection cameras C41, C42, C43, C44, and C45 are installed. These staying detection cameras 4 determine the number of cameras and the arrangement position in accordance with the size of the store and the layout of the product shelves in order to detect the flow line in the store of the customer and the staying time of each area. In the store, there are a plurality of shelves for displaying products. Each shelf is assigned an identifier in advance. In the store according to the first embodiment, there are a total of 44 shelves from shelf T01 to shelf T44.
[1-1-1. Configuration of Analyzer]
FIG. 3 is a diagram illustrating a configuration of the purchase analysis device 2 according to the first embodiment. The purchase analysis device 2 includes an input / output unit 21 for inputting data from an external device and outputting data to the external device, and a person who specifies an individual based on data from the entrance camera C31 and the cameras C32 and C33 before the cash register. Identifying unit 22, staying detecting unit 23 for detecting the staying time of each predetermined area in the store based on data from staying detection cameras C41, C42, C43, C44, and C45; A purchase information acquisition unit 24 for acquiring information, a non-purchase area detection unit 25 for detecting a non-purchase area based on data from the person identification unit 22, the stay detection unit 23, and the purchase information acquisition unit 24; An analysis unit 26 for performing aggregation and analysis based on data from the area detection unit 25; and a storage unit for storing data from external devices and data detected, aggregated, and analyzed by each unit of the purchase analysis device. And a 7. The person specifying unit 22, the stay detecting unit 23, the purchase information acquiring unit 24, the non-purchase area detecting unit 25, and the analyzing unit 26 may all be performed by the same control unit (CPU), or may be a dedicated control unit. You may go in.
[1-2. motion]
The operation of the purchase analysis system 1 configured as described above will be described below. FIG. 4 is a flowchart showing an analysis process of the purchase analysis system 1. In the analysis process, first, the person identification unit 22 first extracts images captured from the entrance camera C31, the cameras C32 and C33 before the cash register, or metadata obtained by extracting feature points and the like (for example, age and gender) from the captured images. A customer identification process (S41) for identifying an individual is performed.

  Next, based on the image data from the stay detection cameras C41, C42, C43, C44, and C45, the stay detection unit 23 detects the stay time of each predetermined area for each customer visit and the sales area stay time. The detection process (S42) is performed.

Next, the purchase information acquisition unit 24 performs a purchase information acquisition process (S43) for acquiring purchase information of the customer from the settlement apparatus 5. Subsequently, the non-purchase area detection unit 25 performs a non-purchase area detection process (S44) for detecting a non-purchase area. Finally, using the non-purchase area detection processing result, the analysis unit 26 performs the detection data analysis processing (S45). Details of each process will be described below.
[1-2-1. Customer identification processing]
The customer identification process will be described with reference to FIGS.

  FIG. 5 shows the identifier of the detected face information (face information ID), the identifier of the camera that detected the face (camera ID), and the detection based on the images and the metadata from the entrance camera C31 and the pre-register cameras C32 and C33. It shows the date and time information, the age group and the gender estimated from the feature amount of the face at the time of face detection. The face information ID is assigned to each piece of data detected as a face from the image data. The face information ID and a face image (not shown) are linked and stored.

  When a customer comes to the store, first, the customer is detected by the entrance camera C31 arranged above the front of the entrance. When the shopping is completed and the payment is made, the customer is detected by the pre-registration camera C32 or C33 arranged above the cash register behind. At normal visit to the store, the face is detected at least twice by the entrance camera C31 and the pre-cashier camera C32 or C33. In the case of leaving the store without purchasing anything (leaving out of the store), it is detected only by the entrance camera C31. In addition, after the payment, if an additional product is purchased without leaving the store, the product is detected again by the pre-registration cameras C32 and C33. The face information detected by the entrance camera C31 and the pre-registration cameras C32 and C33 identifies the customer by face matching processing.

  FIG. 6 is a flowchart showing the operation of the customer specifying process. In the customer specifying process, first, face image information corresponding to the face detection data shown in FIG. 5 is read (S61). Next, comparison source face information is determined from the read face detection data (S62). Next, face comparison processing is performed using the comparison source face information and the comparison destination face information (S63). As a result of the face matching process, when it is determined that the persons are the same (YES in S64), the same customer identifier (customer ID) as the comparison source face information is assigned (S65). The operations from S63 to S65 are performed for all the comparison target face information (S66). When the matching process for the specific comparison source face information is completed, the comparison source image is changed, and a series of processes from S62 to S66 is executed.

  FIG. 7 shows face information data for each customer generated by the customer specifying process of FIG. In FIG. 7, face information data having face information IDs of 1, 4, and N are determined to be the same person as a result of the face matching process, and are given the same customer ID (= 1).

  Using the face information data for each customer in FIG. 7, data for each customer visit is created. FIG. 8 shows a flowchart for generating data for each visit from face information data for each customer. To generate data for each visit, first, face information data for each customer is read (S81). Next, the target customer ID is determined from the customer ID in the face information data (S82).

  Next, the face information data of the determined customer ID is rearranged in chronological order (S83). Then, the date and time information of the first face information data in FIG. 7 is set as the start date and time information, the number of visits is set to 1, and first visit data is generated (S84).

  Next, based on the previous data (the first visit data or the latest visit data), when the date and time information of the subsequent face information data is smaller than a predetermined value as compared with the start date and time information of the previous data (YES in S85). , The end date and time information of the first visit data is updated assuming the same visit (S86).

  On the other hand, if the date and time information of the subsequent face information data is larger than a predetermined value as compared with the start date and time information of the previous data (NO in S85), it is regarded as another visit and the date and time information is used as new date and time information as start date and time information. (Latest store visit data) is generated (S87). Note that the number of visits to the store is +1 to the number of visits to the previous store visit data.

  The subsequent face information is compared with the start date and time information of the latest visit data. The comparison processing of the date and time information is performed for all the face information data of the target customer ID (S88). By executing a series of processing for all customer IDs (S89), data for each visit is generated.

  The predetermined value in S85 is determined by the size of the target store and the average stay time, such as 15 minutes for a small store such as a convenience store, 60 minutes for a large store such as a supermarket, and the like. I do.

  FIG. 9 shows the data for each visit generated by performing the processing of FIG. 8 on the face information data of FIG. The data of the face information ID = 4 in FIG. 7 is the same visit as the face information ID = 1. On the other hand, the data with the face information ID = N is different from the face information ID = 1 as the visit data, and as a result, the number of visits is 2. Further, the camera ID in FIG. 9 is the original data when the first or latest store visit data is created, that is, the camera ID having the start date and time information. If the value is larger than the predetermined value as described above, new visit data is generated, so the end date and time information is blank.

The customer visit identification process is performed by a series of processes from FIG. 5 to FIG.
[1-2-2. Sales area residence time detection processing]
With reference to FIGS. 10 to 15, the processing for detecting the sales area residence time of the customer will be described.

  FIG. 10 is a diagram illustrating a stay detection area (sales area) set in a passage in the store 10. The stay detection area is set to detect that the customer stays in the detection area for a predetermined time. An identifier is assigned to each of the stay detection areas, such as A1, A2,... A27. In addition, as shown in FIG. 11, each area is stored in the storage unit 27 as stay detection area information together with location information within the store. In addition, the stay detection camera position information installed in the store (not shown) is also stored in the storage unit 27 in the same manner as the stay detection area information.

  FIG. 12 is shelf position information indicating position information of cash registers and product shelves related to each stay detection area. The shelf position information is used when the customer stays in the stay detection area for a predetermined time to specify a cashier waiting for the customer or a shelf in which the customer is interested. An identifier is assigned to each cash register and each product shelf, and is stored in the storage unit 27 as shelf position information together with the position information of each cash register and the product shelf in the store, similarly to the stay detection area information.

  FIG. 13 is a diagram showing product category information displayed on each product shelf. An identifier of a product category (product category ID) displayed on the shelf is stored for each product shelf. Each product handled in the store is provided with an identifier for each product and a product category ID indicating category information of the product. By using this information, it is possible to determine which product shelf each product is displayed on. This information is also stored in the storage unit 27, similarly to the stay detection area information and the shelf position information.

  Based on the information obtained by the customer specifying process shown in FIGS. 5 to 9, the information for each store shown in FIGS. 10 to 13, and the image information obtained from the stay detection cameras C41 to C45, the customer FIG. 14 shows a process for detecting the sales area residence time for each visit.

  In the sales area staying time detection processing, first, the target customer ID and the target visit date and time information (start date and time information) are determined from the visit data shown in FIG. 9 (S141). Next, the image information of the stay detection cameras C41 to C45 is read from the target visit date and time information (start date and time information) and the corresponding settlement date and time information (end date and time information) (S142).

  When the image information is read, in order to reliably detect the staying behavior of the target customer from entering the store to the end of settlement and closing the store, the target visit date and time information is set to about 10 seconds before the settlement date and time. The image information is read out until about 30 seconds after the information.

  Next, a tracking target is determined from the target visit date information and the image information read from the settlement date information based on the face image associated with the target customer ID determined in S141 (S143). When the tracking target stays in the detection area set in FIGS. 10 and 11 for a predetermined time (YES in S144), one of the cashiers related to the target detection area is obtained from the product shelf information shown in FIG. Or, it is detected whether or not it faces the direction of the product shelf (S145). If it is detected that it is suitable for any cash register or merchandise shelf (YES in S145), the corresponding shelf ID, stay start time, and stay time are recorded (S146).

  In this case, the predetermined time is set to a value that can be determined to be stationary based on the size of each detection area and the general walking speed. A series of processing from S144 to S146 is performed on the target visit from the entrance to the exit of the store (S147). By performing the series of processing shown in FIG. 14 for each customer visit, it is possible to detect the sales area residence time of all the target customers at the store visit.

  At the time of non-purchase exit, no settlement is performed, so that the face cannot be detected by the pre-registration cameras C32, C33 and the like, and the end date and time information cannot be known from the visit data in FIG. In such a case, the image information is read based on the visit date and time information (start date and time information), and the image information until the tracking target set in S143 leaves the store is additionally read. In addition, after the settlement, if the user stays in the store at the eat-in corner or the like and purchases the product again without leaving the store, the end date and time information is managed as another visit data in FIG. It will be blank. However, since the camera ID is different from normal entrance / exit, the camera ID can be distinguished here. In this case, the time from the settlement time (end date / time information) of the visit data before the target customer to the start date / time information is considered. Image information is read.

FIG. 15 shows the sales area residence time for each customer visit obtained by performing the processing in FIG. In the present embodiment, since the staying start time is recorded, even if the customer stays in front of a plurality of product shelves in one visit, the order of the stayed product shelves can be determined.
[1-2-3. Purchase information acquisition processing]
FIG. 16 shows the purchase information acquired from the settlement apparatus 5. The purchase information includes a store identifier, sales date / time information (payment date / time information), an identifier of a cash register that has performed the payment (checkout number), a receipt number for identifying the payment, a quantity, a sales amount, and a product for identifying the product. ID, a product category ID for identifying the product category, and the like are included.
[1-2-4. Non-purchase area detection processing]
Next, a non-purchase area detection process is performed based on the data obtained by the sales area residence time detection process and the data obtained by the purchase information acquisition process. FIG. 17 shows a flowchart of the non-purchase area detection processing. In the non-purchase area detection process, first, the sales area stay time information of the target customer ID is read from the sales area stay time data of each customer visit shown in FIG. 15 (S171). Next, merchandise category information for each shelf shown in FIG. 13 is read (S172). Subsequently, the purchase information shown in FIG. 16 is read (S173).

  Next, it is confirmed whether there is purchase information having the same or close to the settlement date and time information of the sales area stay time information (S174). If there is no purchase information corresponding to the settlement date information of the sales area stay time information as in the non-purchase closed store (YES in S174), the non-purchase flag is set in all the corresponding sales area stay time information ( S175), ends. If there is purchase information having settlement date and time information that matches or is close to the settlement date and time information of the sales area staying time information (YES in S174), the purchase information with the same receipt number is included in the target sales area staying time. It is checked whether or not the product (category) displayed on the product shelf included in the information is included (S176). If not included (YES in S176), a non-purchase flag indicating non-purchase is set in the corresponding sales area stay time information (S177). The processing of S174 to S177 is performed for all the sales area stay time information of interest (S178).

Through a series of these processes, a non-purchased area can be detected. FIG. 18 shows information obtained by performing the non-purchase area detection processing of FIG. 17 based on the sales area residence time information of FIG. 15 and the purchase information of FIG.
[1-2-5. Detection data analysis processing]
Based on the non-purchased area detection data obtained by the above series of processing, the analysis processing for specifying the non-purchased cause, the scale, and the corresponding priority is shown below.

  As described above, the purchase pattern of the customer includes the planned purchase and the unplanned purchase. In the case of planned purchase, the dwell time generally tends to be shorter than in the case of unplanned purchase. Note that the planned purchase means that the product category and the product brand to be purchased are determined, and thus includes the case where the purchase is made and the case where the purchase is not made. Similarly, unplanned purchasing includes purchasing and not purchasing. Further, these purchasing patterns differ depending on customers, products, and visits. In the present embodiment, the purchase patterns are classified based on the non-purchase residence time information for each customer, each visit and each shelf, and the cause, scale, and response priority are specified.

  FIG. 19 shows a flowchart of the purchase pattern classification process. In the purchase pattern classification process, first, a target customer ID is determined (S191). Next, non-purchase residence time information of the target customer ID is extracted from the non-purchase residence time information shown in FIG. 18 (S192). Next, a total is calculated for each shelf ID from the extracted non-purchase retention time information (S193). Next, based on the non-purchase retention time information totalized for each shelf ID, a threshold is determined for each customer and each product shelf. Based on the determined threshold, each non-purchase residence time is classified as a planned purchase or a non-planned purchase (S194). Aggregation of residence time for each shelf and classification of planned purchase / unplanned purchase are performed for all shelves included in the non-purchase residence time information (S195). By performing the series of processes shown in FIG. 19 for each customer, classification is performed for non-purchasing stays of all customers at all stores. In this embodiment, the average value of the non-purchase residence time is calculated for each customer and each product shelf, and the average value is used as a threshold.

  Here, the residence time, which is the threshold for the classification of planned purchases and unplanned purchases, changes due to changes in customer preferences, so it is necessary to perform this for each customer, not for the entire customer. There is.

  Therefore, in the present embodiment, the residence time for each customer and each shelf at the time of the past month visit is used as an object for calculating the average value. In addition, in the case where the frequency of leaving stores is low, and in the case of visit data for the past month, if the staying time information for calculating the average value is small, the average value is calculated by extending the target period.

  Therefore, if each non-purchase retention time is equal to or more than the average value, the non-purchase retention is determined by unplanned purchase, and if less than the average value, the non-purchase retention is determined by planned purchase. It is considered that the reason for the non-purchase in the planned purchase is that the product being purchased was out of stock. On the other hand, in the case of non-purchase in unplanned purchase, it is considered that the user was interested in the product category but did not purchase it, that is, there was a problem with the product lineup.

  FIG. 20 shows the result of the classification of the purchase pattern for the non-purchase retention of the merchandise shelves with customer IDs 10, 11, and 12 and shelf IDs T10, T28, and T36.

  FIG. 21 shows the result of FIG. 20 tabulated for each shelf ID and each purchase pattern, and tabulated for the number of non-purchases. FIG. 21 shows the estimated loss scale (yen) based on the number of non-purchases and the average purchase unit price of each shelf. For comparison, the average residence time and the number of non-purchase times for each product shelf performed in the conventional example are shown.

  In the case of the conventional example, the cause cannot be identified only by the correlation between the average residence time and the number of non-purchases, but as in the present embodiment, the number of non-purchases is classified by dividing into planned purchases and non-planned purchases. Can be.

  As a result, it is possible to estimate the cause, such as a missing item when the number of times of non-purchase is large in the planned purchase, and a cause of the assortment when the number of non-purchases is large in the non-planned purchase.

  Therefore, in the case of the present embodiment, the shelf T36 with the largest number of non-purchases in the planned purchase is determined to be the shelf with the most missing items in response to the purchase request. In other words, it is possible to determine that the shelf requires the most attention for inventory management and shelf storage. In addition, the shelf T10 having the largest number of non-purchases in unplanned purchasing can be determined to be the most important shelf in reviewing the product lineup.

  Further, in the case of the present embodiment, as shown in the sales area stay time detection processing section, since the stay start date and time information is recorded, each customer stays in front of a plurality of product shelves in one visit. In this case, it is possible to know the order of the stayed product shelves. The strength of willingness to purchase or interest in the product category appears in the buying order when visiting the store. Utilizing this, for example, as shown in FIG. 22, when the number of non-purchasers due to unplanned purchase is the same and the amount of loss is the same in a plurality of shelves, the shelf with the smaller average buying order for each merchandise shelf has the corresponding priority. Can be increased.

FIG. 23 shows an example in which the purchase analysis result is displayed on the display device 6. When analyzing the non-purchase cause, scale, and response priority, as shown in FIG. 23, select the target period, target gender, target age group, non-purchase information of planned purchase or non-purchase information of non-planned purchase. . Based on the selected information, using the non-purchase dwell time information shown in FIG. 18, the tallying / sorting process shown in FIG. 19 is performed, and the number of non-purchase times is estimated in the order of the product shelf with the highest priority. The loss amount is displayed (FIG. 23, lower left). The corresponding product shelves are displayed on the store layout shown on the right side of FIG. 23 so as to be distinguishable from other shelves.
[1-3. Summary]
As described above, in the present embodiment, the purchase analysis device 2 in the purchase analysis system 1
A purchase analysis device connected to a camera and a purchase information management device arranged in a store having a plurality of sales areas, and a person specification unit for specifying a customer based on data (image) received from the camera; A retention detector that detects the residence time of each customer in each sales area; a purchase information acquisition unit that acquires purchase information of each sales area in the entrance time from entering the store to leaving the identified customer; A non-purchase area detection unit that detects a non-purchase area that is a sales area where no purchase information exists among purchase information for each sales area of a customer, and a residence time equal to or longer than a threshold provided for each specified customer in the non-purchase area. And a statistics unit that divides the data into residence times smaller than the threshold value and performs statistics.

  This makes it possible to present the product category in which the opportunity loss has occurred, the cause, the scale of the loss, and the response priority from the purchase information and the in-store action information for each visit of each customer.

  In the present embodiment, in order to extract the cause of non-purchase and the number of times of non-purchase, as shown in FIG. 23, a period, a gender, and an age group are designated, but the present invention is not limited thereto. For example, the tallying may be performed only by customers who frequently visit the store, or the tallying may be performed only by customers having a high average purchase unit price. Also, instead of the number of times of non-purchase, the same customer ID may be treated as one person and the number of non-purchasers may be counted.

  Further, in the present embodiment, as a threshold for classifying the purchase pattern, the non-purchase stay time for each shelf of each customer is totaled and the average value is used, but this is not a limitation. For example, the non-purchase residence time for each customer shelf may be totaled, data may be arranged in descending order of residence time, or in ascending order, and a value between data having the largest difference may be set as a threshold. When a plurality of threshold value determination methods are incorporated and the difference between them is large as a result of actually calculating the threshold value, the threshold value may be determined by looking at the actual distribution of the non-purchase residence time.

  Further, in the present embodiment, the non-purchase residence time for each shelf is counted based on a period such as one month, but the present invention is not limited to this. For example, it may be set so that the non-purchase stay time information of a predetermined shelf becomes a predetermined number. In this case, the tallying period is different for each shelf. Also, if you want to see the effect of changing the assortment or the current assortment response at the store side, use the time when the assortment was changed as the base point and calculate the average value based on the residence time information after the base point. May be extracted. In this case, it is possible to analyze the influence of the product line on the shelf in which the product line has been changed, such as an increase / decrease in planned purchase customers and an increase / decrease in unplanned purchase.

(Embodiment 2)
Hereinafter, the second embodiment will be described with reference to FIGS.

  The difference between the second embodiment and the first embodiment is that in the first embodiment, the non-purchased area is subject to aggregation and analysis, whereas in the second embodiment, the purchased area is also aggregated and analyzed. It is a point that is. In the purchase behavior analysis processing shown in FIG. 4, the different points from the first embodiment are the non-purchase area detection processing (S44) and the detection data analysis processing (S45). In the second embodiment, not only the non-purchase area but also the purchase area is detected, and the tallying / analysis processing is performed for each of the purchase area and the non-purchase area. The non-purchase area detection unit 25 may perform the tallying and analysis of the information regarding the purchase, or a new purchase area detection unit may be provided.

  FIG. 24 shows information obtained by performing a purchase / non-purchase area detection process on the sales area residence time information of FIG. The difference from FIG. 18 in the first embodiment is that the non-purchase flag indicating non-purchase is set in the first embodiment, whereas the purchase (1) is performed in the present embodiment. The point is that an identifier indicating whether or not it is present or not purchased (0) is set.

  Also, in the purchase pattern classification process of the first embodiment shown in FIG. 19, the non-purchase residence time (S192) is targeted, whereas in the second embodiment, each of the purchase residence time and the non-purchase residence time is Perform purchase pattern classification for the target.

  FIG. 25 shows the result of the classification of the purchase pattern for purchase delays of the merchandise shelves with customer IDs 20 and 21 and shelf IDs T11 and T37.

  FIG. 26 shows the results of FIG. 25 tabulated for each shelf ID and each purchase pattern, and tabulated for the number of purchases. FIG. 26 also tabulates the staying time change rate for each unplanned purchase. The staying time change rate is calculated based on the change in the staying time at the time of unplanned purchase for each product shelf of each customer. For example, if the purchase residence time of the customer A on the shelf A at the time of the unplanned purchase changes from 30 seconds to 25 seconds, the residence time change rate becomes -5. Since the staying time change rate changes due to a change in the preference of the customer, in the present embodiment, the staying time at the time of visiting the store in the past month is used as an object for calculating the change rate as in the first embodiment. ing. In addition, if the visit frequency is low and in the visit data for the past month, if the staying time information for calculating the change rate is small, the average period is calculated by extending the target period.

  In the case of purchase by unplanned purchase, when the assortment is new, there are many options, and the purchase residence time tends to be long. On the other hand, if one gets tired of an assortment, the purchase residence time tends to be shorter. Unless the product lineup is reviewed, it may become non-purchased in the future. For this reason, by calculating the rate of change in purchase residence time in unplanned purchasing for each shelf, it is judged that many customers are tired of product shelves with a high rate of decrease, and the priority of reviewing the product lineup is raised. Can be handled. In the case of FIG. 26, the number of purchases by the unplanned purchase is the same for the shelves 11 and 37, but since the reduction rate of the staying time is higher for the shelves 11, the priority of reviewing the assortment of shelves 11 is raised. Judge that it is necessary.

As described above, in the present embodiment, the purchase analysis device 2 in the purchase analysis system 1 is a purchase analysis device connected to a camera and a purchase information management device arranged in a store having a plurality of sales areas, A person identification unit that identifies a customer based on data (images) received from the camera, a residence detection unit that detects the residence time of each of the identified customers in each sales area, and a residence detection unit that enters and exits the identified customer. A purchase information acquisition unit for acquiring purchase information including product information at the time of entering the store, a staying sales area in which the specified customer has stayed for the first time or more, and corresponding product information and purchase information are read; and a non-purchase area detection unit product information not included in the information to detect a non-purchase area in a residence sales area corresponding, on the residence time of the non-purchase area, a second time than provided for each customer identified And a, a statistical unit for statistically classified into whether the residence time. Further, the purchase analysis device according to the present embodiment further reads the staying sales area in which the specified customer has stayed for the first time or more, the corresponding merchandise information, and the purchase information, and includes the same in the purchase information. A purchase area detection unit that detects a purchase area that is a stagnant sales area corresponding to the product information to be acquired. The statistic unit determines whether the resident time of the purchase area is equal to or more than the second time provided for each specified customer. It is characterized in that it is classified as no or statistical.

  As a result, from the purchase information and in-store behavior information for each visit to each store, the product category and the cause of the opportunity loss, the loss scale, the response priority, and the product category in which the opportunity loss is likely to occur in the future, It is possible to present the corresponding priority.

  In the present embodiment, the staying time change rate is extracted based on the number of occurrences of staying purchase in unplanned purchase, but is not limited thereto. For example, it may be extracted on the basis of the information on the date and time of occurrence of purchase detention in unplanned purchase. Further, in the present embodiment, the non-purchase residence time for each shelf is counted based on a period such as one month, but the present invention is not limited to this. For example, if the store wants to see the effect of changing the assortment or the response of the current assortment, the change point is based on the time at which the assortment was changed and the dwell time information after the base point. May be extracted. This makes it possible to extract a shelf whose product lineup needs to be reviewed before the purchase information changes.

  In the first embodiment, the non-purchase cause, the loss scale, and the corresponding priority are detected based on the non-purchase retention time information and in the second embodiment, the purchase and non-purchase retention time information. It is also possible to evaluate based only on the residence time information without using it. In this case, the residence time of each customer for each shelf is totaled, and a purchase pattern classification process is performed. As a result, the number of planned purchases and the number of unplanned purchases for each shelf are extracted as shown in FIG. As a result, for each product shelf in the store, take measures such as increasing the priority of preventing out-of-stock items for shelves with many planned purchases, and increasing the priority of reviewing the assortment for shelves with many unplanned purchases. It becomes possible.

  The present disclosure is applicable to a purchase analysis device. Specifically, the present disclosure is applicable to a purchase analysis device that analyzes a customer's purchase behavior in a store.

DESCRIPTION OF SYMBOLS 1 Purchasing analysis system 2 Purchasing analyzer 3 Face detection camera 4 Stay detection camera 5 Payment device (POS)
Reference Signs List 6 display device 10 store 21 input / output unit 22 person identification unit 23 stay detection unit 24 purchase information acquisition unit 25 non-purchase area detection unit 26 analysis unit 27 storage unit

Claims (4)

A purchase analysis device connected to a camera and a purchase information management device arranged in a store having a plurality of sales areas,
A person identification unit that identifies a customer based on data received from the camera,
A stay detection unit that detects the stay time of each of the specified customer sales areas,
A purchase information acquisition unit that acquires purchase information including product information at a store entrance time from entering the store to exiting the specified customer;
The specified customer is read in the stagnant sales area where the stagnated sales area has stayed for the first time or more, the corresponding product information, and the purchase information are read. A non-purchase area detection unit that detects a purchase area;
A statistic unit that classifies and classifies the residence time of the non-purchase area into a residence time of a second time or more provided for each of the specified customers ;
Purchasing analyzer equipped with. The analyzer according to claim 1, further comprising:
The identified customer stays for the first time or more and reads the stagnant sales area and the corresponding merchandise information, and the purchase information. The merchandise information included in the purchase information is the corresponding stagnant sales area . A purchase area detection unit that detects a purchase area is provided,
The purchase analysis device, wherein the statistics unit classifies the stay time of the purchase area into a stay time equal to or longer than a second time provided for each of the specified customers and performs statistics. The analyzer according to claim 1,
The purchase analysis device, wherein the second time is determined based on a past residence time of the specified customer. A purchase analysis method executed by a CPU of a purchase analysis device connected to a camera and a purchase information management device arranged in a store having a plurality of sales areas,
Identifying a customer based on the camera data;
Detecting the residence time for each sales area of the specified customer,
A step of acquiring purchase information including product information at a store entrance time from entering the store to leaving the identified customer;
The specified customer is read in the stagnant sales area where the stagnated sales area has stayed for the first time or more, the corresponding product information, and the purchase information are read. Detecting a purchase area;
Classifying the staying time of the non-purchase area as a staying time of a second time or more provided for each of the specified customers and statistic;
Purchasing analysis method with.

Images