700,000 parts, and customers find the right one

The challenge

Industrial supply catalogs break conventional search. A contractor doesn't type a clean product name — they type a partial model number, a manufacturer's shorthand, or a plain description of the part they're holding. Across a catalog of more than 700,000 SKUs with inconsistent manufacturer naming, keyword matching returns either nothing or thousands of near-identical results, and neither one leads to a sale. Blackhawk Supply needed customers to find exactly the right part, immediately, on a catalog large enough that browsing was never a realistic path.

Our approach

We treated search as a data problem before an AI problem. Product data was normalized and enriched so attributes were consistent across manufacturers, which alone improved what any search system could do with it. Then we layered semantic understanding on top, so a natural-language description maps to the right SKUs even when it shares no keywords with the product title. Ranking accounts for availability and relevance, and results are cached so repeat queries return instantly rather than paying a model round trip. The search sits inside a full e-commerce platform we built alongside it, so finding the part and buying it are one uninterrupted path.

What we built

• A complete e-commerce platform for a catalog of more than 700,000 HVAC and plumbing products
• AI-powered search that resolves natural-language and partial-match queries to the correct SKUs
• Product data normalization and enrichment across inconsistent manufacturer sources
• Ranking, filtering, and recommendation logic tuned for a professional trade buyer, not a casual browser
• A caching and indexing layer that keeps results instant at full catalog scale

Results