Fordel StudiosAI-Powered Product Search and Discovery
This case study describes a real engagement. Client identity, proprietary details, and specific metrics are anonymized or approximated under NDA.
What needed
solving
Keyword-based site search returning irrelevant results. 40% of search sessions ending without a click. No semantic understanding of product relationships — a search for "office chair with lumbar support" would return results based on keyword overlap rather than product characteristics.
The primary challenge was building an embedding strategy that captures product characteristics accurately across a catalog with significant attribute inconsistency. Product descriptions and specifications were submitted by multiple merchants with different terminology, levels of detail, and formatting conventions — the same product attribute might appear as "lumbar support", "ergonomic back", or "lower back cushion" depending on the merchant. Embedding quality is directly dependent on input text quality, requiring a normalization pass before embedding generation. The personalization re-ranking layer needed to be fast enough to stay within the overall latency budget, which ruled out approaches that required expensive model inference at re-rank time.
How we
built it
- 01
Analysed six months of search session data to identify the highest-frequency zero-result and low-click queries — these became the test set for validating that semantic search was solving the actual problem, not just improving the easy queries.
- 02
Built a hybrid retrieval system combining semantic embeddings for conceptual similarity with structured filters for product attributes, so a search for "ergonomic desk chair" ranks by relevance rather than keyword frequency.
- 03
Embedded the full product catalogue including descriptions, attributes, reviews, and category hierarchy to give the retrieval model a rich semantic surface to match against.
- 04
A/B tested the semantic search against the existing keyword search on a 20% traffic split before full rollout, measuring click-through rate, add-to-cart conversion, and search session abandonment as primary metrics.
This engagement replaced a legacy keyword-based search system with a vector retrieval architecture covering 180,000+ active product SKUs. The system combines dense vector retrieval using OpenAI embeddings with sparse BM25 retrieval in a hybrid ranking layer, which outperforms either approach alone on the query types that dominate the platform's search logs. Product embeddings are pre-computed and stored in Pinecone; query embeddings are generated at request time. Re-ranking applies a lightweight personalization model that adjusts result ordering based on category affinity derived from the user's session and purchase history. Total P95 search latency at production load is 180ms, including embedding generation, retrieval, and re-ranking.
What we
delivered
Vector search with embedding-based retrieval, query understanding layer, and personalized re-ranking using session and purchase history signals. Replaced the existing keyword search index with a hybrid retrieval system that handles semantic queries, misspellings, and attribute-based filtering.
Measurable
outcomes
- Search click-through rate increased 2.4× — the primary metric indicating that results became more relevant to query intent.
- P95 search latency reached 180ms including embedding and vector retrieval, meeting the performance requirement for a responsive search experience.
- Conversion rate for search sessions increased 31%, attributed to users finding the products they were looking for rather than abandoning after irrelevant results.
“The search click-through rate change was visible in analytics within the first week. The semantic understanding is actually working — customers are finding products they would not have found with keyword matching.”
— Head of Product, E-Commerce PlatformReady to build
something like this?
Tell us what you are building. We will scope it, price it honestly, and give you a clear plan.
Start a ConversationFree 30-minute scoping call. No obligation.