Enabling India’s leading multi-brand online beauty retailer to leverage AI and achieve a 40% add-to-bag conversion rate

Developers at Daffodil began the engagement by integrating the shoppable video functionality into the web and mobile apps. The client wanted to fetch videos from its Youtube channel and integrate them into the web and mobile app with the ‘Watch and Buy’ feature. This would allow customers to add products to the bag in real-time while watching the videos, eliminating the need to search for products, even after the video is over. For building this new app feature, a custom video streaming player was developed that fetches content from Youtube. In order to provide a top-notch user experience, the videos were hosted on Akamai CDN .For product tagging in the videos, AI and ML algorithms were developed that fetched product names from Youtube, thereby removing manual intervention. An Embed API was developed, which facilitates users to share videos on third party websites, thereby increasing brand visibility and generating more traffic on their web and mobile apps. For synchronizing engagement stats such as likes and comments from Youtube, automation scripts were written that updates the stats every 10-minute.

For the blog section, a thorough code audit was done to understand the underlying factors affecting the client’s current blog performance. An outdated version of WordPress, malware files, and unnecessary plugins were some of the reasons that contributed to the poor performance. In order to fix the performance issues and introduce new functionalities into the blog, we upgraded the WordPress version from V3 to V4.5. We also performed database cleanup, deleted unnecessary tables, and added new tables to the database in order to support new functionalities. We were able to remove suspicious files (malware), unused/deactivated plugins, and plugins downloaded from unknown sources through a complete codebase cleanup and security audit. The UI/UX of the blog was also improved to ensure a delightful user experience on both the website and the mobile app by using plugins such as Smart Product Viewer, mobi, Awesome Header, and many more. Elasticsearch was used to store, search, and analyze the data in real-time enabling faster and seamless blog access to the users.

In addition to all that, a user community network was built using Java as the core technology for web and Android mobile apps, Swift was used for developing the network for iOS devices. The interactive user community was developed to engage users with beauty buffs, ask and answer beauty-related questions, give and seek advice, discover trends, and join conversations on topics of their interest. Separate Q/A and video advisory sections were developed in addition to a user posting section similar to a social media posting wall where users could create posts or ask questions. The network section also had a registration tab where users could signup and create their personal profiles to help increase user involvement. Best in class technologies such as MongoDB and MySQL were used to maintain the database for the network enabling the development of multiple classifications and categorizations of the data. Natural language processing tools were used to ensure seamless user experience.

The Laravel PHP framework was used for the development of the functionalities as it enables faster development and optimizes the application performance through its source code optimization, SQL request optimization, and caching features.

A few features that were integrated into the application were:

Natural language processing (NLP): The user community network was integrated with machine learning practice NLP to perform sentiment analysis of products mentioned. It also performed functions such as named entity recognition, noun phrase chunking, tagging part of speech, etc. to ease the functionality of the network for users. NLP was also used for the shoppable video section to enable machine translation in different languages and refined search for videos.

Computer Vision: This feature was incorporated in the shoppable video functionality to detect products while the video plays simultaneously and fetch product details for users allowing them to add the products to their cart at the same time

Smart recommendations and auto-tagging feature: Supervised learning technology was integrated into the app to get smart recommendations for videos, blogs, or user community content by using collaborative filtering techniques. This technology was also used for partial detection of products mentions by training the models for the community section and predicting the categories of the Videos for auto-tagging in the shoppable video section.

Live video streaming: The user community network was equipped with a live video streaming functionality wherein the client could stream live videos showcase expert advice, expert Q/A sessions, makeup and beauty routines and recommendations, and much more.

Content categorization: The client could categorize its content based on multiple categories and subcategories such as makeup, skincare, haircare, bath & body, and wellness, etc. for ease in accessibility through the integration of Elasticsearch for both blog and the user community section.