MyHackathon

User sees an item in the video (red handbag).
TikTok recommends the item in the next card.
User can swipe right for more options.
The shopping card will take the user to cart out.

TikTok 2024 TechJam

Team: MyHackathon Problem Statement: #3 Enhancing Tailored Discovery on TikTok Shop

Introduction

This document outlines the technical enhancements to the TikTok Shop, aiming to personalize user shopping experiences through an innovative "For You Page" card. By integrating swipeable shopping cards akin to Tinder's interface, we aim to collect user preferences and utilize machine learning to tailor product recommendations dynamically.

Front-End Design

User Interface

Swipeable Cards: Implement swipeable card interfaces within the TikTok feed. These cards will appear intermittently between standard content cards (videos).
Action: Each card will have swipe actions:
- Swipe Right: Indicates interest in the product.
- Swipe Left: Indicates disinterest.
- Tap: Redirect to TikTok Shop product detail page to access more details about the product and potentially purchase directly.

User Interaction

Feedback Loop: User swipes are collected in real-time to adjust upcoming product recommendations.
Navigation: Easy navigation options to redirect users to the TikTok Shop for a detailed view or to make a purchase.

Back-End Design

Data Handling

Preference Collection: Store swipe data and user interactions to build a profile of user preferences.
Product Matching: Integrate data collected to existing algorithms to match products based on the accumulated preference data.

Integration

Existing Systems: Ensure seamless integration with current TikTok Shop databases and APIs to pull product data and handle transactions.

Model Design

Model Construction

Recommendation Engine: Develop or modify the existing machine learning model that adapts to user preferences from swipe interactions.
Algorithm: Utilize collaborative filtering techniques combined with real-time learning algorithms to enhance the personalization of product suggestions

Personalization

Dynamic Content Generation: The content of each card is generated based on the user’s previous interactions and preferences, providing a unique and personalized shopping experience.
Adaptive Recommendations: Continuously adapt the recommendations based on user feedback through swipes.

Image Recognition Algorithm

Spotlight Item Detection: Our image recognition algorithm can identify spotlight items in user videos (e.g. a red handbag). This feature allows the system to recognize and catalog various products showcased in the videos users watch.
Similar Item Recommendations: Based on the spotlight items detected, the shopping card will recommend similar items available on TikTok Shop. For instance, if a user watches a video featuring a red handbag, the algorithm will identify this item and suggest comparable products in the shopping cards.
Enhanced User Experience: By linking the visual recognition of products in videos with the shopping recommendations, we create a seamless and engaging shopping experience. Users can discover and purchase items that catch their interest directly from the videos they watch.

Assumptions and Constraints

User Engagement: Assume users are familiar with swipeable interfaces.
Performance: Ensure that the back-end can handle real-time data processing without lag.

Summary

The proposed enhancements to TikTok Shop are designed to transform the shopping experience by integrating personalized, actionable shopping opportunities directly into the user's social feed. This approach is expected to increase engagement and conversion rates by leveraging sophisticated machine learning techniques to cater to individual preferences.