Challenge
As already mentioned, Lyst integrates over 27,000 brands. Naturally, the main challenge was the scale of the data and the complexity of the existing architecture. Furthermore, we were working with large, legacy code that had already proven reliable but required a deep understanding of the Lyst website structure and the relationships between microservices. Key tasks included:
- Implementing a large-scale experiment to improve product taxonomy.
- Creating new dynamic attribute sections and expanding color filters.
- Ensuring seamless integration of new modules with the existing infrastructure without compromising stability, as we are dealing with millions of users.
Solution
Given the importance of the solution, our project touched upon all the website's pillars, from the backend logic to the user interface. The work was carried out iteratively, as we prioritized both deployment security and code quality.
- Deep Integration and A/B Testing: The foundation of our project was the implementation and refinement of A/B tests related to filtering. This allowed us to make decisions based on real data rather than hypotheses, gradually improving the user experience.
- Backend Development (Python): Our team implemented complex service logic and data processing in Python. Grinteq integrated new algorithms into the existing microservices system, paying special attention to stable operation under high load.
- Frontend Development (React): Intuitive filtering interfaces were created for fashion shoppers. We developed new dynamic attribute sections, significantly expanded the available color options, and implemented an improved navigation tree.
- Working with a legacy system: Diving into the existing code and architecture was a critical step. We refined the search functionality without disrupting the established relationships between systems.
Tools and Security: To ensure quality, we used pytest for the backend and Jest for the frontend. Automated workflows in GitHub Actions ensured continuous integration. To avoid any risk to the live site, Grinteq ensured secure deployment methods with clearly documented rollback procedures.
Impact
Thanks to the Grinteq team, Lyst successfully completed a large-scale taxonomy improvement experiment. A new filtering system and improved search algorithms allowed users to more effectively navigate the catalog of over 27,000 brands. The introduction of dynamic attributes and an expanded color palette increased search results relevance, which positively impacted conversion. The project was implemented while maintaining high platform stability.
