Digital Advertising and Customer Movement Analysis Using BLE Beacon Technology and Smart Shopping Carts in Retail

Abstract

This paper proposes an innovative, intelligent shopping cart system with an interdisciplinary approach using Bluetooth low energy (BLE) beacons. The research integrates online and offline retail strategies by presenting campaigns and ads to the customers during in-store navigation. In a testing environment, BLE beacons are strategically positioned to monitor the purchasing process and deliver relevant insights to retailers. The technology anonymously logs customers’ locations and the duration of their browsing at each sales shelf. Through the analysis of client movement heatmaps, retailers may discern high-traffic zones and modify product placement to enhance visibility and sales. Additionally, the system provides an additional revenue model for store owners through location specific targeted ads displayed on a tablet mounted on the cart. Unlike previous BLE-based tracking solutions, this research bridges the gap between customer movement analytics and real-time targeted advertising in retail settings. The system achieved an accuracy of 82.4% when the aisle partition length was 3.00 m and 91.7% when the aisle partition length was 6.00 m. This system, which can generate additional income for store owners by generating 0.171 USD in a single test simulation as a result of displaying ads to three test customers in a two-partitioned aisle layout, offers a new and scalable business model for modern retailers.

1. Introduction

Technological advancements have transformed customer interactions and marketing strategies in the retail sector. To increase the shopping convenience of consumers, location-based technologies [1], such as global positioning systems (GPSs), wireless fidelity (WiFi), and radio frequency identification (RFID), offer solutions for tracking customer movements [2]. However, these methods suffer from limitations, including high costs, limited accuracy in indoor environments, and privacy concerns. GPS via satellites, which are very useful in our daily lives, can be used only in outdoor environments, as it requires access to location signals from satellites from an accessible open area, making it an inconvenient technology for indoor regions [3,4]. The advent of Bluetooth low energy (BLE) beacons has provided a cost and energy-efficient alternative for collecting real-time customer movement data, including in the retail industry.

While BLE technology has been widely studied in the context of location tracking, most existing research focuses on optimizing signal accuracy rather than leveraging BLE-based insights for real-time marketing applications. It has many applications, including targeted location-based advertising, tracking the elderly and disabled, and navigation in low-visibility environments [5]. BLE-based location position systems are widely preferred in the retail industry because of their low purchase and maintenance costs, energy efficiency, and ease of application. Current systems, such as mobile coupon redemption systems in countries where customers are detected by their location and proximity to points of sale [1], fail to integrate customer movement analytics with targeted advertising strategies, leaving a gap in how in-store tracking can directly contribute to revenue generation and store layout optimization.

For example, Faragher and Harle [4] demonstrated the feasibility of using BLE for location fingerprinting by achieving an accuracy of 2.6 m over 95% of the area using nineteen BLE beacons in 600 m². However, their approach focused primarily on signal optimization rather than integrating consumer interaction features. Similarly, Assayag et al. [6] proposed a particle swarm optimization model for BLE localization but did not address how this technology could be monetized through targeted advertising. Unlike these studies, our study fills a significant gap in proximity marketing research by combining BLE-based customer tracking with an in-store advertising system.

Table 1. Comparison of the results with those of previous studies.

Study	Technology	Purpose	Limitations	This Study’s Contribution
Faragher and Harle [4]	BLE	Indoor positioning	No marketing integration	Integrates BLE tracking with real-time advertising
Paolanti, Liciotti, Pietrini et al. [7]	UWB	Indoor positioning	No marketing integration	Integrates BLE tracking with real-time advertising
Spachos and Plataniotis [8]	BLE and other beacons	Beacons and innovative city applications	Review article	Integrates BLE tracking with real-time shopping and advertising application
Muddinagiri et al. [9]	BLE, SmartPhone, and IoT	Proximity marketing in retail	The customer’s identity is decrypted from the smartphone.	The customer does not need a communication device; the shopping cart performs anonymization.
Assayag et al. [3]	BLE + PSO	Signal optimization	No revenue model	Adds a monetization strategy via targeted ads
Hoang et al. [10]	BLE + Blockchain	Privacy-preserving tracking	No customer engagement strategy	Incorporates customer engagement through real-time promotions
Prathessh and Rajalakshmi [11]	No Technology	Customer navigation and retail store organization	No revenue model, theoretical research	In our research, the navigation map can be calculated electronically
This Study	BLE	Smart Shopping Cart System	It is just a prototype and has not been tested in a real retail store.	Combines BLE tracking, heatmaps, and revenue generation

In some retail stores, such as Amazon Go, next-generation technologies can be used intensively to produce solutions to current problems. In Amazon Go, which works with the concept of “Just Walk Out Shopping”, cashiers, crowded checkouts, and customers putting the products they buy in their bags or shopping carts and then returning to the checkout to pay, etc., can be solved with this new generation of retail. However, customers are closely monitored after using this intensive technology [12]. Almost all movements in the store are monitored, payments are collected automatically, and they require a subscription to applications prepared by Amazon [13]. Unlike Amazon Go, which tracks customer movements intensively, our system ensures anonymity by avoiding personal data collection while still providing location-based insights. Furthermore, the customer does not need to use any additional software or hardware in addition to the shopping cart.

provides a comprehensive comparison of the focus of our research and previous studies.

This study aims to build and assess a smart BLE-based shopping cart system intended to enhance the in-store customer experience and provide useful insights for businesses. The technology we suggest here enables advertisers to engage with store customers in their purchasing process while offering leverage to store owners so that they can analyze customer behavior. Thanks to the technology suggested, retailers or shop managers can develop such data-driven marketing strategies as providing coupons or discounts to generate interest in under-visited product displays. Moreover, technology enables store proprietors to make supplementary income from advertisers via an advertising strategy predicated on ad impressions. This research is explicitly concentrated on the subsequent objectives:

• Evaluate the precision and practicality of Bluetooth Low Energy (BLE) indoor tracking within a regulated retail setting.
• Integrating digital advertising with personalized location-based technologies to explore potential applications in retail.
• Investigating how Bluetooth Low Energy (BLE) can enhance store placements and utilize customer mobility information to augment revenue.

By integrating BLE-based tracking with real-time advertising, this study contributes to both the indoor positioning research literature and the digital marketing literature. Unlike previous works that focused solely on signal processing, this study aims to meet the following objectives:

• Bridging the gap between in-store tracking and marketing analytics.
• Providing a scalable and low-cost alternative for personalized marketing strategies.
• Offering empirical insights into the commercial viability of BLE-based advertising in retail settings.

The rest of this paper is structured as follows: Section 2 reviews the literature on proximity marketing, indoor positioning systems, BLE devices, and location-based technologies’ impact on marketing. Section 3 describes the methodology of the research. Section 4 presents the design, technical details, and implementation of the proposed innovative shopping cart system. Section 5 presents the findings and evaluations of the test cases planned in the previous section. Section 6 presents the general discussion and then presents the overall conclusions and evaluation of the study in Section 7. Finally, Section 8 summarizes the study’s limitations and indicates future work directions.

2. Literature Review

Recent research has shown that online retail in the U.S. accounts for approximately 21.3% of total retail spending in the U.S., an increase of 44% from the previous year [14]. A new generation of intelligent apps with location-based services (LBSs) is now integral to our daily lives [15]. As brick-and-mortar retailers look to drive more customer traffic to their stores to increase sales, location-based advertising on mobile devices has become an important marketing tool. The effectiveness of such advertising campaigns depends on the ability to collect and analyze data to target the right customers at the right time and place. A previous investigation showed that unanticipated in-store coupons boost unplanned purchases and overall spending [16]. Proximity marketing, which helps businesses provide location-specific advertisements, is crucial to contemporary retail techniques. Such diverse technologies as RFID, NFC, and WiFi have been investigated for proximity marketing, each of which has unique strengths and weaknesses. For example, in Amazon Go stores, where many new-generation technologies are used [13], in-store coupons can be given to customers, but the use of RFID tags per product has created problems. Some limiting issues include round-boxed products, products that must be kept in wet or cold environments, and the cost of attaching RFID tags to each product [12]. In this respect, the use of BLE beacons for in-store coupon delivery to customers or special campaign delivery to aisles stands out owing to their low power consumption and widespread adoption in mobile devices, so our research was conducted on an application using BLE beacons. The following sections discuss developments in indoor positioning systems and their role in improving proximity-based advertising.

2.1. Proximity Marketing

The places where the buyer meets the business and becomes the recipient of marketing communications can be defined as customer touchpoints. This contact can take place in real and virtual spaces [17]. Proximity marketing involves a local advertising approach using wireless access devices. Technically, it is a method that can be used for situations involving advertising to potential users in a specific location [9].

In most applications, proximity marketing can be divided into proximity-based and location-based services in context. In both services, beacons can be placed fixedly or attached to moving objects [8].

Proximity-based positioning focuses on detecting whether objects are in a particular area rather than their exact location. This method, especially with technologies such as Apple iBeacon, allows objects to be detected when approaching a point of interest (PoI). Proximity detection is defined by adjusting the signal strength (Tx-strength) [18]. The approximate location of the object is estimated via machine learning algorithms (such as MLE and KNN). Although this method has difficulty determining the exact location due to signal interference, errors can be reduced with data fusion and motion analysis techniques. Proximity-based methods are effective for announcing specific data or determining the approximate location of objects in IoT services [19]. Proximity marketing can be implemented through various technologies, such as Barcode, RFID, NFC, WiFi, and BLE [9].

Although proximity technologies (Bluetooth, WiFi, NFC) are used in retail stores to track customer behavior and run personalized campaigns, data security and ethical issues arise due to their centralized structure. This leads to issues such as hit inflation and violations of customer privacy [10].

While proximity marketing techniques offer new opportunities for marketers, they also raise privacy concerns. In this research, since the customer does not need to use any device of their own, the proposed system produces a solution that addresses privacy concerns [20].

2.2. Indoor Positioning Systems (IPSs)

The focus of our research is based on determining the customer’s location in the store. In this way, it is possible to track the customer’s location in the store and to deliver advertisements and campaigns for products sold at the customer’s indoor location with anonymous targeting. The following explains current technologies and approaches for location detection in indoor spaces.

Various technologies, such as BLE, WiFi, RFID, and ultrawideband (UWB), are widely used to communicate in geolocation applications in indoor spaces [2,3]. These systems calculate the position by considering several parameters, such as the strength and angle of the signals coming from the beacons [19,21,22]. Owing to these systems, businesses can offer various solutions, including physical in-store positioning and customer tracking.

Indoor positioning systems have great potential to track people’s movements, analyze which sections they are in or how much time they spend, and integrate these data into marketing strategies. Indoor positioning refers to determining a target’s location indoors [23]. It has many applications, including targeted location-based advertising, tracking the elderly and disabled, and navigation in low-visibility environments [5]. BLE-based location position systems are widely preferred in the retail industry because of their low purchase and maintenance costs, energy efficiency, and ease of application.

Positioning technologies in closed areas are separated into infrastructure-based and infrastructure-free technologies. The infrastructure-based method works by converting the signal strength received from the auxiliary devices placed in the environment into distance measurementz and calculating the location with algorithms; for example, the BLE beacon signal strength indicator (RSSI) [6] is used. On the other hand, the infrastructure-less method uses environmental information such as preexisting WiFi access points. For high accuracy, the locations of the access points must be known in advance, which increases the cost. Fingerprint-based positioning determines mobile device locations [3] by creating a database of signal strength at reference points in the offline phase and matching it with this database online. This method is gaining widespread attention because it reduces hardware requirements and efficiently uses available resources [22].

Indoor positioning systems [6] empower marketing strategies by enabling store owners to offer location-based personalized advertisements and campaigns to customers. Technologies such as Bluetooth beacons (such as iBeacons) make it possible to detect customers’ locations in real-time and send them advertising messages or special offers at appropriate moments. These technologies allow store owners to inform customers of deals in their immediate vicinity and encourage unplanned purchases. For example, customers may receive immediate discounts for related products when they approach the detergent sales shelf. IPS technologies enable retailers to learn more about client behavior, facilitate the delivery of more relevant and timely marketing, increase customer happiness, and boost sales. These, as a result, make indoor positioning systems essential in contemporary retail and marketing.

WiFi reduces costs when existing infrastructure is used and provides centimeter-level accuracy with advancements such as WiFi 6. Nevertheless, high-power consumption and interference issues can occur in busy environments. UWB offers a high accuracy of less than 10 cm and is not strongly affected by obstacles but is limited by high infrastructure costs and energy consumption. On the other hand, BLE stands out due to its low cost, energy efficiency, and easy deployment advantages, making it an attractive option for indoor positioning [24]. Although fingerprint-based methods offer high accuracy, they are complex due to the challenges of data collection and processing [25].

BLE beacons facilitate and enhance communication by delivering data to compatible devices over radio waves. This technology monitors customer behavior in real retail environments, facilitating tailored marketing communications and navigation assistance. Beacons engage at many points of the consumer journey, enhancing the customer experience through discounts, product recommendations, and educational messages [23]. The aggregated data amalgamates online and offline behaviors, facilitating the more accurate creation of customer profiles. Thus, brands enhance client loyalty, secure a competitive edge, and deliver services that align more closely with consumer requirements [17].

The project, established in a recent study, seeks to enhance the shopping experience through the construction of an innovative IoT-based shopping cart system. The system is engineered to autonomously identify and monitor products incorporated into the cart with RFID readers, weight sensors, and barcode scanners. The objective is to furnish shoppers with real-time information, including the cumulative cost of items in their basket and their location inside the store, so facilitating automatic invoicing and payment, and thereby eradicating waiting times in queues. The technology will synchronize with the store’s inventory management to monitor stock levels and provide tailored promotions. This solution aims to provide customers with a more convenient and faster shopping experience and provide store owners with valuable data on inventory management and customer engagement [26].

Our study’s focus was combining retail merchandising with targeted advertising, determining customers’ in-store navigation routes, and providing applicable insights to store managers via this information. In this context, unlike previous studies focusing on signal accuracy in indoor position systems, our research proposed a system that would generate additional income for the store owner from advertisements displayed to customers and increase store sales through various promotions and campaigns through an application conducted in a controlled test area simulating a store.

2.3. Bluetooth Low Energy (BLE)

In light of the development of technologies, BLE beacons can be used effectively in marketing because they are economical and easily integrated into applications. In this research, we use BLE beacon technology to instantly determine the location of customer shopping in a retail store, send targeted advertisements to customers via these location data, and record customer movements. Detailed explanations of the BLE beacons are provided below.

Electronic location beacons have become an essential tool for hyperlocal retail promotions, allowing marketers to correlate consumers’ online and offline behavior and run campaigns to target them [27]. BLE technology is compatible with low-cost and energy-efficient beacons [15,28] and has made proximity marketing more accessible by facilitating one-way signal transmission to consumer devices. The widespread adoption of smartphones and BLE technology has helped transfer data gathered from beacons to remote servers, allowing advertisers to analyze customer behavior and provide personalized advertisements [23]. This technology is employed across several sectors, including stadiums and retail establishments, to deliver consumers push alerts, coupons, and navigational assistance [27]. BLE beacons are wireless devices that periodically broadcast signals in the 2.4 GHz ISM band and operate with low power consumption [29]. Multiple devices can read these beacons without pairing and are used for distance estimation and confined space positioning, especially in IoT applications [4]. BLE is part of the Bluetooth 4.0 standard and provides communication over 40 channels, making it an energy-efficient and cost-effective technology [18]. In confined space positioning, BLE usually estimates distance via RSSI, but this method can be affected by obstacles and signal interference. For this reason, accuracy is often improved by combining the BLE with other technologies, such as UWB or inertial measurement units (IMUs). According to Han et al., triangulation and trilateration are characterized by low-cost implementation and high accuracy, particularly within room-scale environments [25]. Fingerprint-based positioning is one of the most accurate methods of confined space positioning [3]. It involves storing the data collected in a database at reference points such as the RSSI [2], magnetic field, or channel status information (CSI). It also compares online measurements with offline data to specify the target’s location [5].

The ease of integration, especially between off-the-shelf BLE beacons and smartphones, has spurred a variety of IoT use cases that require less human effort to perform any task, especially among emerging unmanned IoT applications. BLE beacons have been used in a wide range of IoT innovations, such as improving the shopper’s experience, museum guidance [8], indoor positioning and tracking [30], assisting the visually impaired or disabled, energy-saving smart offices, managing smart homes and warehouses, and locating BLE devices with beacons via fingerprints, etc. [28].

Bluetooth 4.0 or higher innovative protocols send information such as text, UID, or URL to phones. This information can be used in a variety of ways through smartphone apps. There are two standards known in the industry, “Eddystone” and “iBeacon,” with BLE standing out as the most popular proximity marketing technology with minimal restrictions [9].

Recently, BLE beacons in indoor positioning applications have increased in popularity because of their high availability, low cost, low power consumption, and ease of deployment. Beacons can work with or without button batteries [31,32]. Bluetooth 5.1 provides high accuracy with the AoA and AoD techniques [33]. On the other hand, WiFi is suitable for human positioning systems because of its broad coverage and high data rate. Both technologies are affected by signal interference and ambient conditions, but accuracy can be improved with the correct algorithm and filtering techniques. These technologies offer ideal solutions for human and object tracking [30], navigation, and context-aware services in IoT applications [19].

Considering the studies that mention the importance of knowing the consumer [1], BLE beacons offer an effective solution for indoor positioning with the advantages of low cost and easy installation [34]. For example, systems have been developed for tourists showing bus stop timetables or distances to metro stations. It has also been used to provide information about museum artworks [8] or to serve proximity-based ads to store users. BLE beacons provide contextual information by measuring users’ proximity to objects [18] and offer a wider interaction range than QR codes or NFC. BLE beacons monitor users’ activities and support health monitoring systems by integrating wearable devices [28,35].

2.4. Commercial Applications and Retail Optimization with Location-Based Technology

The importance of advertising and advertising activities carried out under the promotion title in the marketing mix is relatively high. Owing to the nature of advertising, a fee must be paid. In this respect, delivering advertising and campaigns to the right people at the right time with proper techniques benefits the marketer. The costs of campaigns with incorrect targeting mean lost profits for the marketer. In 2017, Amazon GO launched a cashierless retail store initiative. This concept, which does not employ cashiers or personnel, uses QR labels, multiple cameras placed on the ceiling, computer vision, weight sensors on the shelves, and intensive artificial intelligence techniques. When the store’s location is determined, owing to this intensive technology, promotional coupons can be delivered while the store is shopping [27]. In addition, all customer movements while shopping can be recorded. Although the use of such intensive technology provides benefits for marketers in terms of efficiency, customer data collection, sustainability, and increased customer satisfaction by eliminating the waiting process at the checkout, it also creates new problems for customers in terms of privacy and the ability to record personal information [12,13,36].

A recent study conducted among customers of a supermarket chain operating in Greece revealed that factors such as proximity to stores, variety, and quality of products, cleanliness, and the attitudes of employees are the primary factors in the formation of customer preferences, whereas low prices compared with competitors and the existence of promotions and discount coupons for products are also very important [37]. Accordingly, the system proposed within the scope of our research will solve an essential problem in determining the times when customers are in the store or determining the frequency of customers visiting the store and determining discounts and other promotions for when customers come less or do not come at all.

Especially during the COVID-19 pandemic, the retail sector’s transition to online platforms accelerated. As a result, storing and analyzing many different types of data on a large scale within the scope of e-commerce analytics has become necessary [38,39].

Predictive analytics uses historical data to predict consumer behavior and leverages advanced data analysis methods such as machine learning in e-commerce. The potential to provide customized services based on customer habits via technology will increase in the future. In contrast, through reinforcement learning, systems can improve themselves based on environmental feedback. This process increases the efficiency of e-commerce by covering many areas, from advertising to post-sales logistics [39].

Understanding shopper behavior is one of the keys to success in the retail industry. In an experimental study conducted in an actual store environment, by integrating the IoT, the system was able to provide the consumer with product location information, including a walking route map and/or aisle number where the product is in the retail store, by predicting the shopper’s next shelf/category attraction. The results showed accuracy exceeding 76% in tracking the shopping cart and basket with a UWB device [11].

The literature review reveals that marketers tend to produce cost-cutting solutions, such as staff reduction and turning to online marketing. On the other hand, marketers should conduct high-cost market and customer analysis studies to determine customer behavior and shopping habits. Processes and technological developments show that new-generation hybrid marketing environments that combine e-commerce and traditional retail may soon occur in our lives. In this respect, the system proposed by our research becomes a valuable model where traditional retailing and online e-commerce can be combined, and customer behavior data can be collected anonymously and quickly.

3. Methodology

This section describes the methodological approach to developing and evaluating the proposed BLE-based innovative shopping cart system. This research has three main parts: (1) experimental setup, where the test environment and BLE marker placement are defined; (2) determination of test scenarios of three customers in the experimental environment; (3) data collection and processing, which includes recording customer movement patterns and advertising interactions; and (4) data analysis, where the collected data are processed to evaluate the system performance in terms of tracking accuracy, advertising interaction, and store layout optimization. Each stage is explained in detail in the following subsections.

3.1. Experimental Setup

To evaluate the performance of the proposed system, we designed a controlled environment that simulates an actual supermarket aisle. The test area is 1.50 m wide and 12.60 m long, and three BLE Bluetooth 5.0 beacons are placed at strategic locations. The selection of three beacons is based on previous research suggesting that at least three reference points are necessary for accurate triangulation. A laptop was used to simulate a tablet computer mounted on a shopping cart, thus providing real-time data collection and advertisement display functionality. During the experiment, the test cart moved along predefined paths and waited in the planned area for three different customer navigation times, as shown in Tables 1–3. Three BLE beacons in the aisle emitted signals at 160 ms intervals, and the received signal strength indicator values were recorded by a laptop PC placed in the shopping cart. Advertisements planned to be sent to the shopping cart’s location area were sent to the laptop PC screen of the cart entering the designated area and displayed for 20 s. The number of advertisement impressions was recorded.

3.2. Case Study: Simulated Retail Store Environment

For the test case study, the shelf navigation durations of three customers were determined. Although these times were different in terms of total time for the three customers, it was ensured that the times were the same for two-, four-, and six-partitioned scenarios where the shelf sizes were varied.

To create a realistic shopping experience, a selection of widely sold supermarket products was included in the simulation. These products covered essential grocery items, household supplies, and personal care products. The advertisements shown to test participants were designed to reflect real-world promotional campaigns, including product discounts, limited-time offers, and store loyalty program incentives. These details provide a more precise context for customer interactions and movement patterns.

In accordance with these preplanned times, customer navigation scenarios were implemented one by one within the test area with a shopping cart and a laptop PC on it, adhering to the prescribed times, and a record of all interactions made by the system was kept. The experiment lasted 15 min for a single customer and included several test case repeats for three. As a result, it took about 45 min to finish all of the exams.

3.3. Data Collection and Processing

The essential stage just after the establishment of the controlled environment and the specification of BLE beacon placement was the ascertaining of the data gathering mechanism. This stage involved the selection of appropriate tracking parameters, configuring signal emission intervals, and defining movement paths to ensure a structured and repeatable experiment. Data collection begins with the real-time location information obtained by tablet PCs mounted on the handles of shopping carts to be used by customers on the basis of the signal from the BLE beacons and sent to the database via WiFi. When customers enter an aisle with their shopping cart, their approximate location is calculated on the basis of the RSSI measurement value from the three closest BLE beacons and sent for recording with a timestamp. Since all the collected data are anonymous, stakeholders do not share customer identity information.

A Kalman filter is used to mitigate noise and signal variances so that BLE locating precision is improved. The system employs RSSI-based distance estimation with environmental factor calibration. These strategies improve tracking accuracy, guaranteeing that client movement data remains dependable even in fluctuating retail environments.

When a customer logs into the targeting zone of an ad in the existing ad pool, the system displays the promotional ad specific to that department on the tablet screen in the customer’s shopping cart. This impression information is also recorded so that the store owner can earn additional advertising revenue. Ads specific to the sales shelf area were shown every twenty seconds. From the recorded data, the time spent by customers on each sales shelf, the frequency of visits, and the navigation routes are analyzed. During this analysis, location information reported from the shopping tool computer is used, which divides a shelf section into three equal parts. Since the entry time stamp is taken for each shelf section, the duration of stay in the relevant area can be calculated. The formula used in the calculation is given in Equation (1).

$$sumT=\sum _{i=1}^n{t}_i.$$

(1)

3.4. Data Analysis

Once the movement data and advertising engagement metrics are collected, they require systematic processing to derive meaningful insights. The data analysis phase focused on assessing tracking accuracy, evaluating the effectiveness of BLE-based advertising, and identifying key performance trends in different store layouts.

The three-phase data analysis made is as follows:

• Using the mean absolute error (MAE), customer movement accuracy was assessed to compare actual positions to estimated positions.
• For advertising impression performance, the number of impressions per shelf area was documented.
• To evaluate the revenue model, advertising revenue was determined based on the impression price per advertiser.

4. Design of the Experimental System and Technical Specifications

This part addresses the design and preparation phases of the experimental system utilized for the investigation. It elucidates the procedure for implementing test scenarios post-installation.

4.1. General Structure of the Prototype System

An aisle the size of an average supermarket aisle was used to test the proposed system. The research used three beacons because at least three markers are required to determine the location according to BLE-based beacons with the triangulation method. In addition to the Python programming language standard libraries for the parts of the proposed system that require coding, such as displaying digital advertisements on the shopping vehicle and recording the coordinates of the shopping carts to the database, the Python Bleak library was used to communicate with BLE beacons. A schematic view of the system is presented in Figure 1.

Figure 1 illustrates the overall architecture of the BLE-based smart shopping cart system, highlighting the interaction between customers, advertisers, and analytical reporting mechanisms. The system operates by utilizing strategically placed BLE beacons within the store to track customer movements in real-time.

When a customer approaches a designated advertising zone, the shopping cart’s mounted tablet detects the BLE signals and displays relevant advertisements based on predefined criteria set by advertisers. The advertisers provide promotional content, which is managed through a centralized system and displayed dynamically depending on the customer’s location.

On the right side of the figure, the analytical reporting section represents the data processing mechanism, where customer movement patterns, ad impressions, and engagement metrics are collected and analyzed. These data allow retailers to assess high-traffic areas, optimize store layouts, and refine targeted advertising strategies for improved marketing effectiveness. The reports generated from this system can be used to adjust ad placement, modify product positioning, and enhance customer shopping experience based on real-time insights.

4.2. Design of the Experimental Evaluation Area

A 1.50 m wide and 12.60 m long area and three BLE beacons were used to simulate the sales shelves. The signal of the selected BLE beacons can usually reach up to approximately 30 m [34]. However, since high shelves negatively affect the distance measurement, the BLE beacons in each shelf aisle are triangular. The test environment was examined by partitioning the shopping area into two, four, and six equal parts. A schematic view of the scenario that divides the test environment into two equal parts is given in Figure 2.

The first test area was partitioned into two equal parts. Each part in this section is 1.50 m wide and 6.30 m long. The following scenario involves the test area being partitioned into four equal parts. The shape of this partitioning is presented in Figure 3.

The test area was partitioned into four equal parts in the second scenario. Each part is 1.50 m wide and 3.15 m long in this partition. The test area is partitioned into four equal parts in our last scenario. The shape of this partitioning is shown in Figure 4.

The third and last test area was partitioned into six equal parts. Each part is 1.50 m wide and 2.10 m long in this partition.

An image of the test environment of the proposed system is given in Figure 5 and Figure 6.

While testing the prototype system, the same area was used for all three test scenarios to simulate sales shelves in a supermarket environment.

4.3. Workflow Diagram of the Prototype System

According to the designed workflow diagram, when the customer starts shopping, the tablet PC in the shopping cart starts searching for the signal of the registered BLE beacons coming from the nearby environment. When at least three registered BLE beacon signals are found, the triangulation method performs the coordinate calculation using the distance values calculated on the basis of the RSSI values coming from these beacons. The coordinates found are reported to the database via the tablet PC’s WiFi connection, which records location information by adding a database timestamp. If an advertisement is specific to the sales store close to the location in the existing ad pool, it shows this ad to the customer on the tablet screen. If there is more than one ad, the advertiser’s advertisement with the highest bid is sent to the customer’s vehicle. The workflow diagram of the system proposed within the scope of the research is given in Figure 7.

Some academic studies mention difficulty in measuring the distance of signals from beacons in closed spaces. In one of these studies, these variable measurements were studied to make them consistent, and three Bayesian filtering techniques were applied. According to previous research, filters have produced excellent results, approaching 40% in proximity estimation [31]. The Kalman filter (KF) is a popular method for tracking and analysis because of its simplicity, optimality, robustness, and ease of use [40]. Proximity estimation was conducted via RSSI values, which were calculated via standardized Equation (2) while considering environmental factors and the BLE signal. The calibration result in the test environment was taken as TxPower –49 dB and envFac as 3.$\mathrm{D}\mathrm{i}\mathrm{s}\mathrm{t}\mathrm{a}\mathrm{n}\mathrm{c}\mathrm{e}=\mathrm{d}\mathrm{x}\mathrm{e}\mathrm{n}\mathrm{v}\mathrm{F}\mathrm{a}\mathrm{c})$

$$Distance=100x\left({10}^{\left({\displaystyle \frac{\left(TxPower-RSSI\right)}{\left(10xenvFac\right)}}\right)}\right).$$

(2)

In our research, the triangulation method, a widely accepted technique frequently used in similar studies, was used to determine the customer’s location. Once the mobile device knows the distance from three known beacons, triangulation is performed to determine its coordinates. Three circles, centered at each beacon with a radius equal to the distance between each beacon and the mobile device, are drawn in Figure 8. The triangulation location is the centroid of the ABC triangle, which consists of cords of the intersection part of three circles [41]. Since the Kalman filter technique has been suggested in the literature to improve measurement accuracy, this technique is also used in our study of the customer coordinate calculation function.

4.4. Bluetooth Low Energy (BLE) Beacons

The NRF52810 Bluetooth 5.0 BLE device produced by HolyIoT (Shenzhen, China) was used for the BLE beacons in the proposed system because it has a long battery life and low maintenance requirements. A typical BLE device operates for at least 205 days with a CR2032-type battery [28]. This device, which is preferred for use, uses one CR2032 type 3-volt 320 mAh capacity lithium battery and can work uninterruptedly for approximately 180 days (6 months) with this power supply. The advertisement signal is made nearly six times per second (160 ms). An image of the relevant BLE device is given in Figure 9.

4.5. Data Collection Tools

The software developed by the researcher via Python v.3.12.5 was used to collect data and display targeted advertisements to the customer. This software is run on a laptop PC, which represents the tablet PC mounted on the shopping cart. The navigations made by the test customers along the aisle with this shopping cart are initiated by the software, which receives signals from three BLE beacons in the aisle. The software records the calculated coordinates in the database via the store WiFi receivers. If advertisements are specific to the customer’s reported location information, they are retrieved from the database and displayed to the customer. Each stage of these processes is recorded with a time stamp for analysis.

4.6. The Case Study for Testing System Accuracy

To test the proposed system, the following scenarios were implemented in the test area for three customers following the predefined periods presented in Table 1,

Table 2. The table of shopping times of two-partition scenarios tests customers in sales shelf areas.

Sales Self Areas	Test Customer (1)	Test Customer (2)	Test Customer (3)
AreaSS-12-1	70 s	40 s	40 s
AreaSS-34-1	50 s	85 s	50 s
Total	120 s	125 s	90 s

and

Table 3. The table of shopping times of four-partitioned scenarios tests customers in sales shelf areas.

Sales Self Areas	Test Customer (1)	Test Customer (2)	Test Customer (3)
AreaSS-12-1	40 s	20 s	10 s
AreaSS-12-2	30 s	20 s	30 s
AreaSS-34-1	20 s	40 s	30 s
AreaSS-34-2	30 s	45 s	20 s
Total	120 s	125 s	90 s

. These durations are the shopping scenarios defined by the author to represent the time spent by a customer while shopping in the store for selecting products between the sales shelves. For example, Test Customer-1 simulates a real customer in a two-partitioned scenario who spends 70 s selecting products in front of a 6.00 m long sales self, AreaSS12-1, and then 50 s in AreaSS34-1. Table 2 shows how long a customer has to wait at the sales shelves for the shopping simulation in the scenario partitioned into two equal parts.

Table 3 shows how long a customer has to wait at the sales shelves for the shopping simulation in the scenario partitioned into four equal parts.

Table 4. The table of shopping times of six-partitioned scenarios tests customers in sales shelf areas.

Scheme 1.	Test Customer (1)	Test Customer (2)	Test Customer (3)
AreaSS-12-1	40 s	0 s	10 s
AreaSS-12-2	0 s	20 s	0 s
AreaSS-12-3	30 s	20 s	30 s
AreaSS-34-1	0 s	30 s	0 s
AreaSS-34-2	20 s	25 s	30 s
AreaSS-34-3	30 s	30 s	20 s
Total	120 s	125 s	90 s

shows how long a customer has to wait at the sales shelves for the shopping simulation in the scenario partitioned into six equal parts.

For the test scenario, ads with various types of impression-based pricing for three different advertisers were planned, and the shelves belonging to these advertisers were determined. This distribution of the ad plan and the prices per ad impression are given in

Table 5. Table of advertising numbers by test area sales shelves.

Sales Self Areas	Ad Counts	Advertiser (1)	Advertiser (2)	Advertiser (3)
AreaSS-12-1	3	0.050 USD	0.055 USD	0.052 USD
AreaSS-12-2	1	No ads	0.028 USD	No ads
AreaSS-12-3	1	0.035 USD	No ads	No ads
AreaSS-34-1	3	0.048 USD	No ads	0.049 USD
AreaSS-34-2	3	0.044 USD	0.042 USD	0.041 USD
AreaSS-34-3	1	0.042 USD	No ads	No ads

.

According to the test scenario, the system will produce customer-specific department reports, each lasting 335 s. It displays advertisements every twenty seconds specific to the areas where customers are located. These tests are performed by configuring BLE beacons to emit signals at 160 ms intervals. The test results compare the data generated by the scenarios partitioned into different numbers of areas.

5. Findings

This section presents the findings of implementing the test environment designed in the previous section for three different customers over different navigation scenarios. According to the test scenario, shopping times simulating shopping in front of the shelves with varying timings for three customers were foreseen, and purchases were tested according to this plan. The data recorded by the system are analyzed, and the findings are presented in the relevant section.

5.1. Customer Navigation Metrics

The shopping process was simulated during the test process at the times specified in Table 2, Table 3 and Table 4. The findings in

Table 6. Comparison of two-partition scenario customer test cases and measurement information of the system.

Sales Self Areas	Test Customer (1)		Test Customer (2)		Test Customer (3)
	Test Dur.	Meas.	Test Dur.	Meas.	Test Dur.	Meas.
AreaSS-12-1	70 s	65 s	40 s	51 s	40 s	34 s
AreaSS-34-1	50 s	49 s	85 s	72 s	50 s	55 s
Total	120 s	114 s	125 s	123 s	90 s	89 s

were achieved by partitioning the scenario into two equal parts.

The findings presented in

Table 7. Comparison of four-partition scenario customer test cases and measurement information of the system.

Sales Self Areas	Test Customer (1)		Test Customer (2)		Test Customer (3)
	Test Dur.	Meas.	Test Dur.	Meas.	Test Dur.	Meas.
AreaSS-12-1	40 s	33 s	20 s	24 s	10 s	18 s
AreaSS-12-2	30 s	34 s	20 s	27 s	30 s	19 s
AreaSS-34-1	20 s	25 s	40 s	36 s	30 s	39 s
AreaSS-34-2	30 s	28 s	45 s	33 s	20 s	12 s
Total	120 s	120 s	125 s	120 s	90 s	88 s

were achieved by partitioning the scenario into four equal parts.

The findings presented in

Table 8. Comparison of six-partitioned scenario customer test cases and measurement information of the system.

Sales Self Areas	Test Customer (1)		Test Customer (2)		Test Customer (3)
	Test Dur.	Meas.	Test Dur.	Meas.	Test Dur.	Meas.
AreaSS-12-1	40 s	11 s	0 s	6 s	10 s	9 s
AreaSS-12-2	0 s	25 s	20 s	12 s	0 s	7 s
AreaSS-12-3	30 s	27 s	20 s	16 s	30 s	14 s
AreaSS-34-1	0 s	23 s	30 s	42 s	0 s	36 s
AreaSS-34-2	20 s	13 s	25 s	24 s	30 s	19 s
AreaSS-34-3	30 s	18 s	30 s	25 s	20 s	4 s
Total	120 s	117 s	125 s	125 s	90 s	89 s

were achieved by partitioning the scenario into six equal parts.

As a result of the test being applied to three different partitioned areas, the measurement consistency between the regions is given in

Table 9. Customer navigation test metric accuracy table.

Sales Self Areas	Under Measurement Error	Over Measurement Error	Average Measurement Error
Two-partitioned area	7.46%	9.17%	8.31%
Four-partitioned area	22.92%	12.22%	17.57%
Six-partitioned area	39.38%	100.00%	69.69%

.

When the obtained metrics are examined, the areas that give the most precision regarding under- and overmeasurement are the two, four, and six-partitioned test areas, respectively. A graphical visualization of these summary metrics is presented in Figure 10.

5.2. Advertising Impression Performance Analysis

This topic will examine the metrics and effects of advertisements displayed by the system specific to the customers’ location as the test customers walk along the aisle.

Table 10. Two-partitioned area test times and number of ads shown to customers.

Sales Self Areas	Test Customer (1)		Test Customer (2)		Test Customer (3)
	Test Durations	Displayed Ads	Test Durations	Displayed Ads	Test Durations	Displayed Ads
AreaSS-12-1	70 s	2	40 s	1	40 s	1
AreaSS-34-1	50 s	2	85 s	2	50 s	2
Total	120 s	4	125 s	3	90 s	3

presents the shopping times of customers moving through two-partitioned aisle areas and the number of advertisements presented to them on a customer basis.

Table 11. Four-partitioned area test times and number of ads shown to customers.

Sales Self Areas	Test Customer (1)		Test Customer (2)		Test Customer (3)
	Test Durations	Displayed Ads	Test Durations	Displayed Ads	Test Durations	Displayed Ads
AreaSS-12-1	40 s	1	20 s	1	10 s	0
AreaSS-12-2	30 s	1	20 s	1	30 s	1
AreaSS-34-1	20 s	0	40 s	1	30 s	1
AreaSS-34-2	30 s	2	45 s	1	20 s	1
Total Seconds	120 s	4	125 s	4	90 s	3

presents the shopping times of customers moving through four-partitioned aisle areas and the number of advertisements presented to them on a customer basis.

Table 12. Six-partitioned area test times and number of ads shown to customers.

Sales Self Areas	Test Customer (1)		Test Customer (2)		Test Customer (3)
	Test Durations	Displayed Ads	Test Durations	Displayed Ads	Test Durations	Displayed Ads
AreaSS-12-1	40 s	0	0 s	1	10 s	1
AreaSS-12-2	0 s	0	20 s	0	0 s	0
AreaSS-12-3	30 s	1	20 s	0	30 s	0
AreaSS-34-1	0 s	1	30 s	1	0 s	1
AreaSS-34-2	20 s	1	25 s	1	30 s	0
AreaSS-34-3	30 s	1	30 s	1	20 s	0
Total	120 s	4	125 s	4	90 s	2

shows the times representing the shopping durations of customers moving through the six-partitioned areas and the number of advertisements presented to them per customer. According to the information obtained, the advertisement display in the six-partitioned test area was unsuccessful. For example, although 40 s of shopping time was spent in AreaSS-12-1, the first area for Test Customer-1, no ads were displayed during this period. In other scenarios, there have been cases where the customer was in the sales shelf section for a sufficient amount of time to be shown an advertisement, but the ad was not shown because the position of the shopping cart was not determined correctly.

When the ad display metrics given in Table 10, Table 11 and Table 12 are analyzed, the following conclusions are drawn. The ads displayed in the two-partitioned area were correctly positioned, and the accuracy rate was 100%. When testing the four-partitioned area, customer-1 failed to display at least one ad in AreaSS-34-1. Therefore, according to the test scenario where 12 ads were expected to be shown, eleven ads were shown, and the accuracy rate was 91.66%. In the six-partitioned scenario, at least one ad was shown in AreaSS-12-1 for customer-1, one ad was shown when no ad should have been shown in AreaSS-34-1, one ad was shown when no ad should have been shown in AreaSS-12-1 for customer-2, no ad was shown when at least one ad should have been shown in AreaSS-12-2 and AreaSS-12-2, for customer-3, one advertisement was shown in AreaSS-12-1 and AreaSS-12-2 when no ad should have been shown, one advertisement was shown in AreaSS-12-3 when no advertisement should have been shown, one advertisement was shown in AreaSS-34-1 when no ad should have been shown, and finally, no advertisement was shown in AreaSS-34-2 and AreaSS-34-3 when at least one advertisement should have been shown. Ten out of seventeen ads were displayed correctly. In this case, the ad display accuracy in the test scenario with a six-partitioned area was 58.83%.

Table 13. Advertising test metrics revenue table.

Sales Self Areas	Expected Ad Display Count	Actualized Ad Display Count	Accuracy
Two-partitioned area	8 ads	8 ads	100.00%
Four-partitioned area	12 ads	11 ads	91.66%
Six-partitioned area	17 ads	10 ads	58.83%

shows the ad impressions expected, actual, and accuracy metrics.

Table 14. Advertising test metrics average revenue table.

Sales Self Areas	Advertising Display Count Avg.	Total Advertising Revenue
Two-partitioned area	2.7 ads	0.171$
Four-partitioned area	3.7 ads	0.156$
Six-partitioned area	3.3 ads	0.155$

shows the average number of ad impressions and total revenue realized in ad impressions.

When the ad display metrics are examined, it is determined that most ads are shown in the four-partitioned area, the second most ads in the six-partitioned area, and finally, the two-partitioned area. A graphical visualization of this information is given in Figure 11.

When the ad impression price metrics are examined, it is determined that the highest-earning ad presents in the two-partitioned area, the second highest-earning ad in the four-partitioned area, and finally, in the six-partitioned area. However, although there is profit in the six-partition area, it is not possible to discuss the display of correctly targeted ads from the advertiser’s perspective because of the high error rate in the measurement metrics of this area. A graphical visualization of this information is given in Figure 12.

5.3. Customer Navigation Heatmap Analysis

After the test was conducted in the area where two-, four-, and six-partitioned areas were created for three customers, the system recorded the locations and shopping times of the customers. According to the test scenario, waiting was made straight along the corridor from the zero point, representing the test customers, following the previously planned shopping times. Within the scope of our study, the X-axis coordinate was taken as 4.75 m, and only the progress distances along the Y-axis were recorded. The store aisle representing the 12.60 m Y-axis was segmented along the minimum accuracy limit of 2.00 m for BLE beacons, which was suggested on the basis of literature studies. The customer shopping times recorded in the two-partitioned areas of the system are analyzed and presented in

Table 15. Customer navigation metrics for two-partitioned areas.

X-Axis	Y-Axis	Customer (1)	Customer (2)	Customer (3)	Total Wait
4.75 m	2.00 m	16 s	9 s	2 s	27 s
4.75 m	4.00 m	22 s	11 s	27 s	60 s
4.75 m	6.00 m	27 s	22 s	3 s	52 s
4.75 m	8.00 m	11 s	10 s	12 s	33 s
4.75 m	10.00 m	20 s	26 s	14 s	60 s
4.75 m	12.00 m	18 s	45 s	31 s	94 s

.

The shopping navigation times of the test customers along the two-bay corridor are presented in Table 15, and the heatmap visually created via the Python programming language and Matplotlib v.3.9.2 tool on the basis of these metrics is presented in Figure 13. On the basis of the information obtained, the customers shopped approximately 12.00 m of the corridor with the longest shopping time of 90 s, shopped approximately 4.00 m and 10.00 m with 60 s, and shopped approximately 6.00 m of the corridor with 52 s.

The shopping navigation times of the test customers along the four-partitioned aisle were recorded by the system, and these time metrics, grouped according to 2.00 m distances, are presented in

Table 16. Customer navigation metrics for the four-partitioned areas.

X-Axis	Y-Axis	Customer (1)	Customer (2)	Customer (3)	Total Wait Duration
4.75 m	2.00 m	8 s	10 s	2 s	20 s
4.75 m	4.00 m	25 s	12 s	15 s	52 s
4.75 m	6.00 m	21 s	16 s	10 s	47 s
4.75 m	8.00 m	19 s	14 s	13 s	46 s
4.75 m	10.00 m	19 s	28 s	31 s	78 s
4.75 m	12.00 m	28 s	40 s	17 s	85 s

.

The navigation times of the test customers along the four-partitioned aisle are presented in Table 16, and the heatmap created on the basis of these metrics is presented in Figure 14. On the basis of the information obtained, it was observed that customers shopped around the 12.00 m of the aisle with the longest shopping time of 85 s, shopped around the 10.00 m of the aisle with the second longest shopping time of 78 s, and shopped around the 4.00 m of the aisle with the shortest shopping time of 52 s.

The shopping navigation times of the test customers along the six-partitioned aisle were recorded by the system, and these time metrics, grouped according to 2.00 m distances, are presented in

Table 17. Customer shopping navigation metrics for the six-partition area.

X-Axis	Y-Axis	Customer (1)	Customer (2)	Customer (3)	Total Wait
4.75 m	2.00 m	9 s	4 s	3 s	16 s
4.75 m	4.00 m	2 s	5 s	10 s	17 s
4.75 m	6.00 m	42 s	13 s	6 s	61 s
4.75 m	8.00 m	21 s	24 s	19 s	64 s
4.75 m	10.00 m	17 s	32 s	30 s	79 s
4.75 m	12.00 m	26 s	47 s	21 s	94 s

.

The shopping navigation times of the test customers along the four-partitioned aisle are presented in Table 17, and the heatmap created on the basis of these metrics is presented in Figure 15. On the basis of the information obtained, customers shopped the most, with a shopping time of 94 s, approximately 12.00 m from the aisle; the second, with a shopping time of 79 s, approximately 10.00 m from the aisle; and finally, the last, with a shopping time of 64 s, approximately 6.00 m from the aisle.

The average time measurements of the test customers’ during shopping navigation along three different partitioned aisles made by the system are grouped according to 2.00 m distances and presented in

Table 18. Average customer navigation metrics across all three areas.

X-Axis	Y-Axis	Two Part.	Four-Part.	Six Part.	Avg. Shop Duration
4.75 m	2.00 m	27 s	20 s	16 s	21 s
4.75 m	4.00 m	60 s	52 s	17 s	43 s
4.75 m	6.00 m	52 s	47 s	61 s	53 s
4.75 m	8.00 m	33 s	46 s	64 s	48 s
4.75 m	10.00 m	60 s	78 s	79 s	72 s
4.75 m	12.00 m	94 s	85 s	94 s	91 s

.

The average shopping navigation time heatmap created according to Table 18, which shows the navigation times of the test customers along three different partitioned aisles, is presented in Figure 16. When the averages of the metrics obtained from all tests carried out in two-, four- and six-partition areas were examined, the test customers shopped approximately 12.00 m from the corridor for 91 s, approximately 10.00 m for 72 s, and finally approximately 6.00 m for 53 s.

5.4. Optimization of Store Layout and Its Business Implications

Recent research has shown that the configuration and design of retail establishments significantly affect consumer perceptions, emotions, and behaviors. It has also been revealed that an intuitive and emotionally resonant retail design can improve customer happiness and brand loyalty, while social interaction areas promote community engagement. Additionally, integrating digital and physical store experiences has been identified as a key factor in better meeting consumer expectations and providing a competitive advantage [11].

The heat maps created by examining the test customers’ aisle navigation time metrics provided vital information to store managers. When the two- and four-sectioned area metrics, which provide the most accurate information in terms of customer location accuracy, were examined together, it was determined that the customers were approximately 12.00 m from the aisle for 90 s, approximately 10.00 m for 69 s, and finally, approximately 4.00 m for 56 s.

It is important to note that these findings are based on a limited sample size of three test participants, and the results should be interpreted as preliminary insights rather than definitive conclusions. The navigation data collected is an example of how retailers can analyze customer movement patterns to optimize store layouts. Future studies should incorporate a larger sample size and real-world retail settings to validate these findings more comprehensively.

Additionally, the test environment included a selection of commonly sold supermarket products, including packaged foods, household goods, and personal care items, to create a realistic shopping scenario. Advertisements displayed on the smart shopping carts included promotions for discount products, seasonal offers, and loyalty-based campaigns. These details have been added to Section 3.2 to provide further context on the customer interaction scenarios.

The information obtained should be used to determine why customers are reluctant to purchase products that spend less time along the aisle. In addition, advertisers or store managers can make radical decisions to offer more attractive discounts or coupons for products located in areas where less shopping time is spent along the aisle.

6. Results and Discussion

In this section, the study’s findings are discussed in light of current academic research, the shortcomings of the study are pointed out, and suggestions for future studies are made.

6.1. Performance Evaluation

The proposed system has an accuracy of 91.7% in tracking customer movements in the two-partition layout, 82.4% in the four-partition layout, and 30.3% in the six-partition layout. In this case, if a similar application is to be made in the real world, the length of the product shelves to be advertised and promoted should not be shorter than 3.00 m. However, in previous studies, it was determined that the accuracy in determining the location according to the signals produced by BLE signals was between 2.00 and 2.60 m [4,42]. The results of our study confirm this finding. When implementing the system proposed in this research, it is recommended that the shelf length be 3.00 m or greater in terms of accuracy. These results resulted from the use of three BLE batches for the test area considered in our research. However, as in the studies by Faragher and Harle, using more beacons in a given area would increase the estimated positioning accuracy, and a shorter shelf distance could be used.

To better understand the financial implications of the proposed system, consider a hypothetical retail store with 100 shopping carts outfitted with BLE-based smart tablets. Assuming that each shopping cart generates an average of four targeted ad impressions per shopping session and that the store serves 500 consumers per day, the total daily ad impressions would be 2000. With an average income of 0.05 USD per ad impression, the store may earn around 100 USD each day, which has the incentive of adding 3000 USD to monthly revenue over a 30-day period. On a yearly basis, in-store digital advertising may generate an additional 36,000 USD in revenue. This scenario highlights the potential profitability of merging BLE-based tracking with targeted advertising in retail environments, providing store owners with a fresh revenue stream while increasing customer engagement.

6.2. Comparison with Previous Work

Recent studies on BLE-based indoor positioning and proximity marketing have focused on improving location accuracy or optimizing signal strength. For example, Faragher and Harle [4] demonstrated a BLE-based indoor positioning system with a 2.6 m accuracy but did not integrate targeted advertising or in-store analytics. Similarly, Assayag et al. [3] employed particle swarm optimization for BLE signal optimization; however, their study lacked a revenue model for retailers.

In contrast, our study bridges the gap between BLE-based tracking and digital advertising by incorporating real-time targeted promotions on the basis of customer movement patterns. Compared with Muddinagiri et al. [9], who explored BLE-based proximity marketing relying on smartphone interactions, our system eliminates the need for customer-side devices. It provides a fully integrated shopping cart solution. Additionally, while Hoang et al. [10] investigated privacy-preserving BLE tracking through blockchain, our study emphasizes practical store layout optimization and revenue generation without requiring complex cryptographic solutions.

Moreover, our findings indicate that BLE-based location tracking achieves an accuracy of 91.7% at the 6 m and 82.4% at the 3 m aisle partitions, which aligns with previous research suggesting that BLE accuracy is typically between 2.0 and 2.6 m [3,4]. However, our research extends beyond localization accuracy by quantifying advertising impressions, optimizing store layouts, and assessing revenue potential through targeted in-store promotions.

We take a more holistic approach by combining BLE-based tracking, revenue modeling, and marketing analytics, which previous studies have investigated separately. Our work offers an innovative and practical solution for the retail industry and provides actionable insights for advertisers and store owners.

7. Conclusions

Our study presents a new, intelligent shopping cart system that employs BLE beacon technology to improve the in-store customer experience while also providing retail operators with relevant insights. The system monitors customers’ in-store whereabouts and delivers localized promotions and adverts personalized by advertisers based on products near them. It also generates heat maps of consumer movements, which provide useful data for optimizing store layout and product placement. This technology could help to bridge the gap between consumer experience and shop analytics.

In this study, essential insights were gleaned from the simulated shopping experiences of test customers in an experimental environment. These insights have shown significant gains in the practical usefulness and scalability of the system for BLE-based proximity marketing techniques in real-world applications.

To fully reap the benefits of developing technologies, interdisciplinary research that goes beyond technological progress is required. This research is a prime example of an approach that successfully integrates electronics, computer science, and business science developments. While current studies have focused primarily on BLE beacon technology, energy efficiency, signal quality improvement, and distance measurement consistency, this study takes a broader perspective by proposing a system that combines these technological advances with practical business applications. The suggested approach overcomes technical problems while also providing a framework for merchants to increase customer engagement and generate additional revenue sources.

The results of this research are beneficial for the advertiser, the customer, and the store owner. The proposed system allows advertisers to reach customers directly in-store, creating a competitive environment where promotional offers can be presented to the right person at the most opportune moment, i.e., when the customer is in front of the product. This eliminates the need for customers to use additional devices, such as smartphones, to access promotional information, as all relevant advertisements are displayed on the shopping cart-mounted screen. Good integration of technology into the shopping process can increase customer comfort.

As a result of testing the system with different scenarios and in an area similar to typical supermarket sales shelves, some determinations were made. Our first determination is that using three BLE beacons for the test area, which is 12.60 m long, is insufficient. Although the installed system works as envisaged, it behaves independently in the entrance and exit areas. For this reason, it has been shown that it is necessary to use BLE beacons more frequently.

Another determination of our method is that targeted advertisements do not reach correctly in areas where the shopping area to be advertised is less than 3 m. Although more accurate targeting has been performed in the region partitioned into two parts, in practice, more than one category of products can be displayed in this space on supermarket shelves. However, it can still be said to be a success, given that the customer has not yet left the area.

Our solution provides store managers with a new business model via targeted advertising. Store managers can use revenue from these adverts to cover the system’s maintenance and operational costs, which may result in a self-sustaining solution. In addition, user privacy is protected by technology that does not disclose personal information to business owners or advertisers. This dedication to data protection is critical for gaining customer trust and complying with privacy requirements.

A pilot study should be undertaken in a genuine retail store with heavy consumer traffic to further assess the suggested system’s real-world applicability. In this scenario, BLE beacons would be installed in several store sections, and shopping carts would be outfitted with tablets for in-store navigation and targeted advertising.

One conceivable application scenario is to analyze client mobility patterns in high and low-traffic store locations. Using heat maps created from customer navigation data, store managers can relocate underperforming product shelves in more visible regions or launch clever discount campaigns to increase foot traffic.

Furthermore, the system might be examined during seasonal promotions or peak demand periods to determine its impact on consumer behavior and purchasing decisions. For example, during Black Friday discounts, the system may direct customers to reduced products while improving in-store advertisement distribution.

The system we propose here could revolutionize the retail industry by providing more personalized and efficient shopping experience. Retailers can better understand their customers’ behaviors and preferences by combining BLE beacon technology with powerful data analytics. This, in turn, may bring about more effective product placement, tailored marketing efforts, and higher customer satisfaction. The system’s potential to help earn extra cash through adverts can give shops a competitive edge in an increasingly tough business.

To summarize all the results, the proposed BLE-based shopping cart system demonstrated an accuracy of 91.7% in tracking customer movements in a two-partitioned scenario. In contrast, the accuracy decreased in four- and six-partitioned areas. The advertisement revenue model proved feasible as a secondary revenue for store owners. Customer movement heatmaps indicated that the most visited shelf areas were 12.00 m, 10.00 m, and 4.00 m, suggesting that strategic product placements can optimize sales.

Regarding the practical implications of the model proposed in the manuscript, store owners can leverage this system to optimize aisle layouts and product placements based on customer movement data. Advertisers can benefit from real-time targeted promotions within the store environment, increasing engagement. The system provides a scalable and cost-effective alternative to traditional retail analytics.

8. Limitations and Future Work

Although exhibiting promising results in a controlled environment, the proposed approach has numerous limitations that must be recognized. Initially, signal interference and ambient noise in actual retail settings may affect the accuracy of BLE localization. Elements such as congested aisles, metallic shelving, and interference from several beacons may cause inaccuracies in location monitoring.

One major limitation of this study is the small sample size of three test participants. While the results provide valuable preliminary insights into customer movement patterns, future studies should incorporate a more extensive and diverse group of participants to improve the generalizability of the findings. Furthermore, real-world retail stores have space and layout constraints that were not fully replicated in the controlled test environment.

Ultimately, even though the system effectively incorporates tailored advertising, the enduring effects of location-based promotions on customer engagement and store profitability remain uncertain.

Future applications should evaluate the effectiveness of the proposed system in different store layouts and varying customer densities. Furthermore, the testbed should be expanded to real-world retail environments with diverse customer behaviors. The impact of dynamic ad pricing models based on customer engagement metrics should be investigated. Finally, AI-based recommendation systems can be integrated to increase targeted advertising effectiveness and research can be conducted on the impact of various ad placement tactics on customer behavior over long periods of time.