The Challenge
An automotive performance parts retailer operated a custom ecommerce platform with well over 250 million product variants distributed across multiple search indexes. The scale and complexity of its catalog made parts catalog search performance a central engineering concern.
Facet calculations were particularly demanding. Helping shoppers narrow results across a large assortment required an architecture that could process product attributes efficiently while supporting relevance, scalability, and availability.
The retailer needed specialist guidance during implementation and continued assistance as it expanded the platform’s capabilities. Its goals included improving query performance, developing more effective ranking, and establishing ways to evaluate search changes.
Innovent’s Approach
Innovent’s core effort focused on query performance. The team redesigned the search architecture to distribute product categories into separate indexes, improving how aggregation and facet workloads were handled across the catalog.
Alongside the architecture work, Innovent helped implement relevance engineering techniques that combined field based matching and weighting, metrics based ranking, and classification boosting. These approaches addressed how product information and additional ranking signals influenced the results shoppers received.
The team also helped implement query intent discovery to better account for what shoppers were trying to find. This connected query interpretation with matching and ranking decisions rather than applying the same treatment to every search.
The broader engagement included ingestion redesign and keyword injection, along with A/B testing strategy, implementation, and measurement. These efforts connected changes to searchable product data and search behavior with methods for evaluating their effects.
Innovent also provided application performance monitoring and platform upgrade guidance, and participated in natural language processing and vector search initiatives. These activities supported the retailer’s evaluation and development of additional search capabilities.
Ongoing autocomplete guidance addressed opportunities to improve how shoppers were assisted as they entered queries, extending the work beyond the results page to the beginning of the search experience.
The Result
The retailer gained a revised index architecture designed to improve aggregation and facet performance across a catalog containing well over 250 million product variants.
Implemented relevance capabilities incorporated field weighting, metrics, classification, and query intent. The engagement also brought measurement and experimentation into the process of evaluating further search improvements.
Innovent’s work connected architecture, data ingestion, relevance engineering, and operational performance across the platform’s development. Continued autocomplete guidance and participation in emerging search initiatives supported the retailer’s ongoing improvement efforts.