Research overview
Research reason
In today's competitive market, businesses and retailers are heavily investing in enhancing their processes to develop superior business models that excel in performance and sales Key priorities include conducting market analysis, adapting marketing strategies, optimizing supply chains, managing costs effectively, and improving shopping experiences.
Enhancing the user shopping experience is crucial for successful retailing today, which revolves around prioritizing customer satisfaction Shopping automation plays a pivotal role in addressing key challenges that detract from this experience, such as long queues, overcrowded stores, and stressed sales assistants According to the Global Retail Automation Market report (2020-2025), the retail automation market was valued at USD 12.45 billion in 2019 and is projected to grow to USD 24.6 billion by 2025, reflecting a compound annual growth rate (CAGR) of 11.2%.
Automation in shopping has advanced, yet challenges persist in enhancing customer satisfaction while maintaining economic efficiency and scalability Businesses and technology firms are exploring solutions to determine which automated models can effectively replace traditional shopping methods Key considerations include identifying technologies that deliver a positive user experience, ensuring security and feasibility, and remaining competitive against other models Consequently, the development of "A mobile application for automated shopping" aims to create a shopping automation model that improves the user experience while ensuring economic viability for retailers.
Aims of the study
Successfully design a convenience store shopping automation model, apply suitable technologies to replace people in several operations, solving the problem of customer experience and economy problems for businesses
- Identify the stages that can be automated
- Determination of suitable technology and applicability
- Determine the factors that affect the success of the model
- Conducting technology research and building a complete model
- Implement and evaluate the results achieved.
Methods of study
An automatic shopping model utilizes technology and machines to enhance the shopping experience by replacing staff in various stages, transitioning from traditional payment methods to electronic payments, and reducing long wait times along with on-site customer service To be effectively implemented, this automatic model must adhere to essential foundational requirements.
- Buyers get exactly what they want, avoiding the situation of machinery and technology errors that prevent customers from purchasing valid products
- Buyers pay exactly what they chose, neither underpaid nor overpaid
- Seller receives the exact money buyers paid
- Security and economic efficiency criteria
To enhance online payment systems, research and development efforts focus on accurately identifying products available in stores This approach aims to improve user experience and streamline transactions, ensuring precision in product recognition and selection.
E-wallet payment is a payment method used so that buyer's money can reach the seller because of the popularity and reliability of e-wallet in the market QR code technology ku"crrnkgf"vq"eqpvtqn"ewuvqogtÓu"ceeguu"vq"vjg"uvqtg"cpf"hqt"ewuvqogt"vq"ocmg"rc{ogpv0
To ensure the security aspect of business owners, the model requires users to register and declare their identity on the HoApp self-developed application.
Outline
This thesis report begins by introducing successful automation shopping models, detailing their processes and technologies, along with the rationale behind their implementation and the benefits they offer The second section focuses on the design of the HoApp model, providing a comprehensive overview of its functionality and system architecture In the evaluation segment, the app's performance during testing will be assessed The report concludes with a summary and outlines future development directions for the thesis.
Related works
Some successful automation shopping model
Amazon Go represents the most promising automatic shopping model, poised to become a future trend Developed by Amazon, a leader in retail and technology, this innovative approach is set to revolutionize the shopping experience.
Amazon Go is a network of convenience stores in the U.S that allows customers to shop without traditional checkout processes Utilizing advanced technologies like computer vision, deep learning algorithms, and sensor fusion, these stores automate the purchasing, checkout, and payment experiences, providing a seamless shopping experience.
Figure 2.1 Amazon Go application interfaces [3]
In early 2018, Amazon launched its first fully automated Go store in Seattle after over two years of testing, allowing customers to simply walk out with selected items that are automatically charged to their Amazon Prime accounts.
Amazon Go has revolutionized the retail experience by effectively utilizing computer vision, deep learning algorithms, and sensor fusion for seamless purchase, checkout, and payment processes This innovative approach makes shopping easier and more convenient than ever before, turning dreams into reality for consumers.
A more specific description of the shopping process at Amazon Go:
- To shop at Amazon Go, you will need to first download the Amazon Go app and connect your Amazon account
To shop at Amazon Go, simply scan your app's barcode upon entering the store, allowing you to pick any items you desire and exit without the need for traditional checkout.
Amazon Go stores utilize advanced sensors to monitor items you pick up or return to the shelves Upon exiting, your Amazon account is automatically charged for the items you take with you, and you can conveniently verify your receipt through the app for accuracy.
Amazon Go features innovative smart carts equipped with built-in cameras, computer vision technology, a display screen, and a QR code reader This allows customers to shop, purchase, and pay directly from the cart, creating a familiar yet highly convenient and efficient shopping experience.
The high cost of implementing the Amazon Go model remains a significant obstacle to its widespread adoption, especially among small and medium-sized businesses Establishing a convenience store using this technology requires an investment of approximately $1 million, with Amazon projected to take around two years to achieve break-even As a result, the Amazon Go model is not a feasible option for smaller enterprises looking to replace traditional retail methods.
The high cost and challenges in achieving absolute accuracy in computer vision models, such as those used in Amazon Go, hinder their competitiveness against established RFID technology, which has earned significant trust among businesses over the years Furthermore, concerns over workforce reduction in certain Western countries, particularly the US, have also impeded the widespread adoption and replication of the Amazon Go model in various markets.
Another automated shopping model that has also achieved certain success and is facing enormous growth opportunities is BingoBox
BingoBox, a self-service convenience store brand founded by Chen Ziling in 2016, allows customers to enter by scanning a QR code at the door As of April, it has expanded to over 300 locations across 30 cities in China.
2018 It has also expanded to Taiwan, Malaysia, and South Korea and has also started expanding to Japan
BingoBox, a GGV portfolio company, is a 24-hour cashier-free convenience store that offers a seamless shopping experience through its smart counter Utilizing RFID technology and computer vision, BingoBox efficiently tracks items without the need for staff Customers can easily access the store by scanning a QR code and complete their purchases using Alipay or WeChat Pay.
To shop at BingoBox, customers begin by scanning a QR code to register on the website At checkout, they simply place their items on a counter, scan the QR code again, and complete the payment using WeChat Pay or AliPay A sensor at the door ensures that all items have been paid for, as it will only unlock once payment is confirmed for every item.
BingoBox has achieved remarkable success by utilizing a compact store design with only 800 SKUs, competitive pricing, and strategic locations in crowded areas The thorough implementation of RFID technology for item identification, tracking, and security has significantly reduced rental and labor costs, making operations one-fifteenth as expensive as traditional stores Additionally, BingoBox has introduced FAN-AI, a computer vision system that enhances the shopping experience by recognizing products at checkout and providing customer assistance when needed.
BingoBox presents an excellent business model for small and medium enterprises, as it operates in a market that still offers significant opportunities for growth With its current limited coverage worldwide, there is ample potential for new businesses to explore and adopt a similar model, making it an attractive option for those looking to enter this sector.
BingoBox model is considered to be very potential and highly feasible when applied in the Vietnamese market for the following reasons:
- The Vietnamese market is large and potential and has not really had a shopping automation model like BingoBox applied
- The compactness, labor and rental cost saving make it competitive with traditional convenient store
- The popularity of MoMo e-wallet and young people's willingness to experience the new will be advantageous when the model is deployed
2.1.3 Advantages and Disadvantages of Amazon Go and BingoBox
- Is it the ideal automation shopping system because of its computer vision techniques applying to all of the shopping processes
- Easy to understand, convenient, highly efficient and accuracy when applying
- Have a group of technicians and staffs to help with the technology and shopping progress
- Receipt is handled quickly and after shopping successfully, an email is sent vq"wugtÓu"crr" (in 24 hours)
- Have staffs to receive feedback and criticism
- There are smart carts to assist the shopping experience
- Iu"vjg"wpocppgf"oqfgn"*oqfgn"vjcv"fqgupÓv"pggf"human to operate)
- Quick payback time (about 10 months per store)
- Low building price for one store, affordable with small and medium enterprises
- Save on space rental cost because of small design and have no staffs
- Can move goods in store easily, even all goods to another store
- Easy to understand, can register an account in place About the disadvantages:
- Technique is complicated and high cost
- Not feasible for small and medium companies and stores when their economical efficiency is not as expected
- Pggf"c"nqpi"vkog"vq"jcxg"wugtÓu"tgeqipkvkqp"cpf trust
- Jkij"ghhkekgpe{"dwv"pqv"322'."tgn{kpi"qp"ecogtcÓu"ejgem"cpf"ugewtkv{ staffs
- Human rights issues in U.S when applying teejpqnqi{"vq"tgrnceg"jwocpÓu work
- RFID stickers have a quite high price, not efficient when buying in small amount
- Need staffs to stamp the stickers for every products
- EcpÓv"fgvgev"kh"uqogqpg"tgoqxg"vjg"uvkemgt0
- If a problem happening in store, there is no one in place to solve the situation
To have a more specific view between the two models, I have implemented the following table:
Table 2.1 BingoBox RFID vs Computer Vision check-out stand
Convenient store and grocery store Convenient store
- Using high tech to enhance the efficiency of old shopping environment
- Solving waiting lines during peak hours, which may lead to the customer going to another store
- Reduce labor costs since cashier in US is very highly paid
- Creating new shop and a new environment for consumers, a new way to buy products
- Creating a new channel, a high efficiency channel, an option other than the traditional model
- Aiming towards a small cost model by reducing labor and rental cost
- Cashier-free store (still have product consultants, on-site customer support staff, security)
- Computer vision, deep learning algorithms, and sensor fusion for the purchase, checkout, and payment steps associated with a retail
- RFID and computer vision to keep track of items, QR code to enter the store
- 10,400 square feet (~1000m 2 ) with 5000 items (Go Groceries)
- Medium and large size, with areas of 12.48m 2 and 15.6m 2 respectively with 500-800 items respectively
Use of technology in process
- QR code scanning at check-in and out
- Cameras and sensors track what customer take off or put back on the shelves
- Access and make online payment via
- Amazon Go app with Amazon account
- QR code scanning at check-in and when making payment
- RFID tags on each product, RFID reader to recognize product
- Small machine called BingoBox applying computer vision to recognize product (recently)
- Online payment via Wechat or Alipay
- 1m$ on hardware and 2 years for Amazon to break even
Unknown build cost but 10 month but
- the average payback time is around
Future potential With 1b$ investment each year,
- Amazon expect to reach 3000 Go stores in 2021
Taken over and supported from Alibaba,
- BingoBox aims to expand and reach 5000 stores worldwide
After thorough investigation, I believe the BingoBox model is viable and offers significant advantages over the Amazon Go model for application in Vietnam The strengths of BingoBox are more pronounced and better suited to the local market, making it a promising option for retail innovation.
Background
The HoApp model incorporates QR code technology, which is a two-dimensional barcode developed in 1994 for the automotive industry in Japan QR codes serve as machine-readable optical labels that provide essential information about the associated item Typically, they direct users to a website or application by containing data for locators, identifiers, or trackers QR codes utilize four standardized encoding modes—numeric, alphanumeric, byte/binary, and kanji—to efficiently store data, with the option for extensions as well.
The Quick Response (QR) system has gained popularity beyond the automotive sector due to its rapid readability and superior storage capacity compared to traditional UPC barcodes Its versatile applications encompass product tracking, item identification, time management, document management, and marketing strategies.
A QR code is a matrix of black squares on a white background that can be scanned by a camera It utilizes Reed-Solomon error correction to ensure accurate data interpretation The necessary information is extracted from the patterns found in both the horizontal and vertical elements of the code.
There are different types of QR code:
QR Code Model 1 and Model 2:
Model 1 is the original QR code The largest version of this code is 14 (73 x 73 modules), which is capable of storing up to 1,167 numerals
Model 2 is QR Code created by improving Model 1 so that this code can be read smoothly even if it is distorted in some way QR Codes that are printed on a curved surface or whose reading images are distorted due to the reading angle can be read efficiently by referring to an alignment pattern embedded in them This code can encode up to 7,089 numerals with its maximum version being 40 (177 x 177 modules)
Micro QR Code features a single position detection pattern, unlike standard QR Codes that require three corner patterns, making it more compact The largest Micro QR Code version, M4 (17 x 17 modules), can store up to 35 numerals In contrast, the iQR Code Model offers versatility with a wide range of sizes, from smaller codes to larger ones capable of storing significantly more data iQR Codes can be printed in various formats, including rectangular, turned-over, black-and-white inversion, or dot pattern codes, making them suitable for diverse applications The maximum theoretical version of iQR Code is 61 (422 x 422 modules), which can accommodate approximately 40,000 numerals.
A QR Code with reading restrictions can store both public and private data, with the latter accessible only through a specialized reader equipped with a cryptographic key, ensuring data protection The Secure QR Code (SQRC) maintains the same appearance as a standard QR Code, effectively preventing forgery and tampering.
The code features a customizable frame that can hold an image, allowing for flexible changes in shape and color This versatility enables a wide range of applications, as both text and images can be incorporated within the frame FrameQR is ideal for promotional purposes, authenticity verification, and various other uses.
For our project, we have selected QR Code Model 2 due to its widespread popularity and usage in modern applications This model offers an optimal size that ensures easy scanning for customers during check-in, unlike the smaller iQR codes Additionally, since our requirement is solely for a tracker that directs users to an application, such as a bank account, we do not require the more complex FrameQR, which is designed to accommodate images.
Kotlin is a versatile programming language designed for the Java Virtual Machine (JVM) and Android, integrating both object-oriented and functional programming paradigms It emphasizes interoperability, safety, clarity, and robust tooling support Additionally, Kotlin has versions that target JavaScript ES5.1 and native code through LLVM, making it suitable for various processors.
Kotlin originated at JetBrains, the company behind IntelliJ IDEA, in 2010, and has been open source since 2012
Until May 2017, Java and C++ were the only officially supported programming languages for Android development However, at Google I/O 2017, Google introduced official support for Kotlin, which became integrated into the Android development toolset with the release of Android Studio 3.0 For earlier versions of Android Studio, Kotlin can still be utilized by installing a plug-in.
Kotlin compiles to the same bytecode as Java and seamlessly interoperates with Java classes, sharing the same tooling This allows for efficient integration, making it easy to incrementally add Kotlin to existing Android applications written in Java without any performance overhead.
Why I choose Kotlin over Java and other programming languages
The debate between choosing Kotlin or Java for Android development has gained traction following the Google I/O announcement Kotlin is recognized for its safer and more concise code compared to Java, and it allows for seamless integration with existing Java files in Android applications This flexibility makes Kotlin not only suitable for new app development but also advantageous for enhancing current Java-based apps With its growing popularity and a wide array of programming utilities, Kotlin is poised to become a leading trend in the future, making it a wise choice to adopt now.
For our mobile application, I wug"Iqqing"Hktgdcug"cu"qwt"fcvcdcug"hqt"gcukgt"crrÓu" korngogpvcvkqp"cpf"cwvjgpvkecvkqp"rtqitguuÓu"ukornkhkecvkqp0
Hktgdcug provides a comprehensive suite of services that developers typically need to create applications, including analytics, authentication, databases, configuration, file storage, and push messaging These cloud-hosted services are designed to scale effortlessly, allowing developers to focus on enhancing the user experience without the burden of managing infrastructure.
HktgdcugÓu"vqqnu"hqt"dwknfkpi"HoApp
Firebase Authentication simplifies user login and identification, making it a crucial component for configuring other products, particularly when restricting access to user-specific data It facilitates secure logins, a process that can be challenging to implement correctly on your own.
Firebase Realtime Database and Cloud Firestore are powerful database services that allow developers to set up listeners for real-time data updates By utilizing the SDK, apps can automatically receive updates whenever changes occur in the specified data location, eliminating the need for manual polling This functionality enhances the efficiency of data management and provides a seamless user experience.
System Analysis and Design
Use case Diagram
Figure 3.1 Use case Diagram for HoApp
Use case Descriptions
Table 3.1 Use case descriptions for Customer
Customer User Registration Register an account to use Account is compulsory when using the app Can be done after filling all important information
User Login Use an existing account to login and access the crrÓu"ujqrrkpi"hwpevkqpu0
Change Password Modify the password in case they forgot or wanting another password
User Functions Access functions that are available to users
Search Products Gpvgt"vjg"rtqfwevÓu"pcog"vq"vjg"ugctej"dct"vq" search its information
View Receipts Show the receipts that are generated after users made a purchase Each of the receipts can be viewed with more detail
View Products Information Ujqy"vjg"rtqfwevuÓu"information in detail
Add Products to Cart Uecp"vjg"rtqfwevuÓu"ST"eqfg"vq"cff"vjgo"vq" vjg"crrÓu"ectv0
Edit Profile Change some attributes in Profile screen by enkemkpi"éGfkvẹ0 View User Ranking Show the ranking of customers of the app based on their purchased products
Check Out Make a payment, interact with products before paying
Modify Products in Cart Change attributes of products in cart
Change Products Amount in Cart
Increase or Decrease the amount of each products in cart
Delete Products in Cart Fgngvg"rtqfwevu"d{"enkem"éFgngvgẹ"qt"fgetgcug" the amount to 0
Select Payment Method Choose the payment options for the order
Pay by MoMo in Store Use MoMo to pay for the order when shopping in store
Pay by MoMo online Use MoMo to pay for the order and the store will deliver the products to you
Pay by Cash in Store Use cash to pay for the order when shopping in store
Pay by Cash at home Pay for the order by cash when you receive it
Table 3.2 Use case descriptions for Staff
Staff members can log in using their accounts to access essential functions within the system They can manage online requests by retrieving orders that need to be prepared for customers before their arrival.
Accept Payment Uecp"wugtÓu"ectv"cpf"ceegrv"rc{ogpv"cu"OqOq" or using cash
Table 3.3 Use case descriptions for Admin
Admin Login Use a admin account to login and access the crrÓu"admin functions
Admin functions include essential access for managing products, allowing admins to search for product information efficiently They can scan and retrieve details about existing products in the database Additionally, admins have the ability to add new products by filling in the required information, ensuring the database is up-to-date and comprehensive.
Access detailed product information and edit specific product details as needed Monitor product sales data by month and year, and analyze sales performance across various product categories.
Flow Chart Diagrams
Figure 3.4 Flowchart for Change password
Figure 3.5 Flowchart for Search Products
Figure 3.6 Flowchart for View Product Information
Figure 3.7 Flowchart for Add Product to Cart
Figure 3.8 Flowchart for View Receipts
Figure 3.9 Flowchart for Edit Profile
Figure 3.10 Flowchart for View User Ranking
Figure 3.11 Flowchart for Modify Products in Cart
Figure 3.12 Flowchart for Check Out
Figure 3.13 Flowchart for Get Request from Customer
Figure 3.14 Flowchart for Accept Payment from Customer
Figure 3.15 Flowchart for Admin Search Products
Figure 3.16 Flowchart for Admin Add a product
Figure 3.17 Flowchart for Cfokp"Xkgy"RtqfwevÓu"Kphqtocvkqp
Figure 3.18 Flowchart for Admin Edit Product Information
Figure 3.19 Flowchart for Admin View Products Sales
Entity Relationship Diagram
Figure 3.20 Entity Relationship Diagram for HoApp System
Table 3.4 Entity Relationship Diagram Description
- user_ID: identifier of user
- user_email: email of user, uses for login
- user_password: password of user, uses for login
- wugtakocig