AI Search for Multi-Brand Auto Parts Retailer
Challenge
The retailer was experiencing a high number of returns caused by customers purchasing parts that appeared compatible with their vehicles but failed to fit during installation.
The existing Year/Make/Model filtering system could identify basic vehicle compatibility but did not consistently account for configuration-level details such as engine type, drivetrain, trim, transmission, and body style.
The merchandising team also had to manually process fitment feeds, OEM and interchange numbers, electric-vehicle data, and newly released model-year information. As the catalog expanded, maintaining accurate compatibility data became increasingly time-consuming.
Additionally, “did not fit” return reasons remained isolated within the support system. They were not systematically used to correct catalog information, improve search rules, or prevent similar compatibility errors from affecting future customers.
The retailer was experiencing a high number of returns caused by customers purchasing parts that appeared compatible with their vehicles but failed to fit during installation.
The existing Year/Make/Model filtering system could identify basic vehicle compatibility but did not consistently account for configuration-level details such as engine type, drivetrain, trim, transmission, and body style.
The merchandising team also had to manually process fitment feeds, OEM and interchange numbers, electric-vehicle data, and newly released model-year information. As the catalog expanded, maintaining accurate compatibility data became increasingly time-consuming.
Additionally, “did not fit” return reasons remained isolated within the support system. They were not systematically used to correct catalog information, improve search rules, or prevent similar compatibility errors from affecting future customers.
Solution
WiserBrand developed a dedicated fitment intelligence and intelligent search layer while keeping Magento Open Source as the retailer’s primary order management platform.
The new solution connected product and order data, supplier fitment feeds, OEM and interchange references, internal compatibility rules, and historical return codes. A Solr-backed search pipeline and Kafka-based processing workers normalized incoming records, identified conflicting data, and kept the search index updated.
The system introduced configuration-level matching based on engine, drivetrain, trim, transmission, and body style. This allowed customers to narrow down compatible products earlier in the shopping journey instead of discovering incompatibility after checkout.
WiserBrand also improved the interpretation of customer search queries. The search engine could recognize part numbers, OEM references, interchange numbers, vehicle chassis codes, automotive slang, and symptom-based searches and translate them into structured catalog criteria.
AI-assisted enrichment was used to identify missing attributes, recognize emerging vehicle and part terminology, and expand catalog coverage for electric vehicles and new model years. Compatibility-critical information was still validated through trusted data sources, QA rules, and human review workflows before being displayed to customers.
Finally, WiserBrand created a return feedback loop. “Did not fit” return codes were routed back into fitment QA, allowing recurring issues to trigger catalog corrections, search-rule updates, attribute improvements, and manual review tasks.
WiserBrand developed a dedicated fitment intelligence and intelligent search layer while keeping Magento Open Source as the retailer’s primary order management platform.
The new solution connected product and order data, supplier fitment feeds, OEM and interchange references, internal compatibility rules, and historical return codes. A Solr-backed search pipeline and Kafka-based processing workers normalized incoming records, identified conflicting data, and kept the search index updated.
The system introduced configuration-level matching based on engine, drivetrain, trim, transmission, and body style. This allowed customers to narrow down compatible products earlier in the shopping journey instead of discovering incompatibility after checkout.
WiserBrand also improved the interpretation of customer search queries. The search engine could recognize part numbers, OEM references, interchange numbers, vehicle chassis codes, automotive slang, and symptom-based searches and translate them into structured catalog criteria.
AI-assisted enrichment was used to identify missing attributes, recognize emerging vehicle and part terminology, and expand catalog coverage for electric vehicles and new model years. Compatibility-critical information was still validated through trusted data sources, QA rules, and human review workflows before being displayed to customers.
Finally, WiserBrand created a return feedback loop. “Did not fit” return codes were routed back into fitment QA, allowing recurring issues to trigger catalog corrections, search-rule updates, attribute improvements, and manual review tasks.
Results
Within six months of implementation:
- Wrong-fitment returns decreased by 35%.
- Manual merchandising and fitment QA workload fell by approximately 60% per 10,000 onboarded SKUs.
- Fitment conflicts that previously required two to three weeks to identify could be detected near real time.
- Zero-result searches for electric vehicles and newly released model-year vehicles declined as catalog and taxonomy coverage improved.
- Customer return data became an active source of search and catalog improvements rather than remaining isolated within the support system.
Within six months of implementation:
- Wrong-fitment returns decreased by 35%.
- Manual merchandising and fitment QA workload fell by approximately 60% per 10,000 onboarded SKUs.
- Fitment conflicts that previously required two to three weeks to identify could be detected near real time.
- Zero-result searches for electric vehicles and newly released model-year vehicles declined as catalog and taxonomy coverage improved.
- Customer return data became an active source of search and catalog improvements rather than remaining isolated within the support system.