Designing the “AI Chat Personalized Shopping Assistant” for Amazon
User Research
Objective:
To know how users interact with large e-commerce platforms.
Methodology:
- Observation: I went to Media Market, a big electronics retailer, with 2 of my friends to observe how customers navigate and choose products in a physical shop. I interviewed them in the process of shopping to understand how they go about making decisions when presented with an enormous number of choices.
- User Interviews: I interviewed 3 users via phone calls who shop online to discuss their experiences with e-commerce platforms, specifically to understand the issues they are having in finding the right products.
- Competitive Analysis: I took big e-commerce and technology companies like Amazon, Walmart, Trendyol, Media Market, and Aliexpress to learn their solutions.
Observation Questions:
- How do you decide which product to buy in this store?
- What factors do you consider most important when comparing products?
- How do you deal with the amount of available options you are given?
- What form, if any, of assistance would you desire in-store when making a choice?
- What frustrations have you faced when shopping in big-box retailers?
Online Consumers Questions:
- How often do you shop online and what type of products have you typically purchased?
- What has been your issues when trying to find products online?
- How do you usually compare similar products?
- Which features do you think should be really included in the online shopping experience so that people could be helped?
- Has there been a case when you have felt overwhelmed by the number of choices that existed while shopping online?
Key Insights:
- Real-World Observations: When customers shop in big retail stores like Media Market, they usually seek the help of sales assistants to compare products and find the best match. They need personalized guidance in line with their requirements, be it budget, technical features, or brand preference. Some of them mentioned they check the prices in different online stores. Also, while they are in the market, they read the users’ comments on the web about the product. Sometimes they feel good discussing a product with the seller.
- Online Shopping Needs: Online shoppers expressed frustration with the time-consuming process of filtering and sorting through thousands of products.
- AI Assistance Demand: In-store and online users expressed interest in AI assistance systems that can quickly understand their needs and suggest personalized options.
Competitive Analysis:
I did a competitive analysis regarding how leading e-commerce platforms treat personalized shopping and AI-driven assistance. The results are summarized below:
Feature | Amazon | Walmart | Trendyol | Media Markt | AliExpress |
Personalized Recommendations | Advanced AI-driven recommendations based on past purchases, browsing history, and preferences. | Offers personalized suggestions, but primarily based on previous purchases and browsing behavior. | Provides personalized recommendations based on past purchases, preferences, and browsing history; focuses on localized promotions. | Basic recommendations based on browsing history, lacking deep personalization. | AI-driven personalized recommendations, sensitive to search history, preferences, and location—different for each user. |
Chat Support/Assistance | No AI-driven shopping assistance. | No AI-driven shopping assistance. | Chatbot is available for customer support, but mainly focused on general inquiries. | Limited to basic customer service chat, no AI-driven shopping assistance. | No AI-driven shopping assistance. |
User Experience | Powerful personalization features, but the interface can be overwhelming and cluttered for new or less tech-savvy users. | Clean interface, but personalization is less integrated and user-driven. | Streamlined interface with personalization and localized content; tailored promotions. | Simple interface, but limited in guiding users through a personalized shopping experience. | Tailored experience with localized content; interface can be cluttered and overwhelming for new users. |
This table compares the personalized shopping features of Amazon with other major platforms.
User Persona and Journey Map:
To ensure that my design decisions are user-centered and effectively address real-world user needs, I created a proto-persona for a group of people and the user journey map.
Wireframe
I did some wireframing about the user flow and, more generally, the ‘Personalized Shopping Assistant’ interface. In some ways, this worked as a blueprint for the feature with low-fidelity sketches, which guided me in visualizing how the layout and interactions work before moving on to high-fidelity designs.
Related Links:
High-Fidelity Design and Prototype
A high-fidelity design was created to reflect the final look and feel of the AI assistant. The followings are the main frames of the redesigned pages and the new feature.
Related Links:
Design Rationale
Built the “AI Chat Personalized Shopping Assistant” for replicating the in-store experience, wherein customers make a lot of decisions based on personal contact with sales assistants. Add an AI chat interface, and we can make product discovery easier, enable better decision-making, and reduce decision fatigue.
- Conversational Interface: Chat-based interaction makes it more natural for users to express their needs, removing friction from navigating complex menus and filters.
- Real-Time Comparisons: Products are compared real-time to facilitate informed decisions at a great value for the tech-savvy user and those making complex purchases.
- Efficiency and Personalization: It learns from user behavior, preferences, and feedback in order to fine-tune recommendations over time for higher accuracy and relevance.
This approach fits pretty well in an already developed Amazon app to address customer satisfaction by providing a personalized, fast, and rich shopping experience. Since the feature is implemented in the form of a chatbot, it runs independently of the rest of the app and does not change or conflict with it; hence, the overall user experience remains consistent and familiar.