Case study: Mid-size automotive parts catalog — Zen Cart search that survives SKUs
Anonymized Zen Cart parts retailer: messy SKUs, nickname queries, and a catalog too big for LIKE search — without naming the merchant.
Picture a mid-size automotive parts catalog on Zen Cart: tens of thousands of SKUs, fitment language in the shopper’s head, and a search box that used to reward exact strings only.
We cannot name the merchant here. What we can say is the pattern — because we have lived it on more than one parts-shaped catalog. The goal of this case study is not a vanity metric chart. It is a human story of why search failed, what we installed, and what changed for the people behind the counter and the DIY buyer at home.
The shopper sentence that broke search
People do not search like a spreadsheet. They search like:
- A partial OEM number with a missing dash
- “Upper hose” plus a vehicle they already picked in a fitment widget elsewhere on the page
- A brand nickname that never appears in
products_name - “That gasket for the early revision” — tribal knowledge from a forum post
Classic Zen Cart LIKE search either dumps thousands of weak matches sorted by product name, or returns nothing. Both feel broken. Neither helps a parts counter finish a phone order or a DIY buyer complete a weekend repair.
What was already “solved” (and still failed)
Before Seekmodo, teams usually try keyword stuffing in descriptions, category browse education (“use the YMM picker first”), or a bolted-on site search widget that never quite matches the theme. Support macros teach customers to search “better.” That is blaming the user for a retrieval problem.
Fitment widgets help when shoppers use them. Search still has to work when they ignore the widget and type anyway — which they will.
What we installed
- Seekmodo Zen Cart connector with search swap and suggest / typeahead
- Enhanced Native so unpaid or unpaired periods still beat native LIKE
- Click beacons so ranking can learn which PDP people actually open from a messy query
- Careful stock and visibility rules so search does not become a tour of discontinued noise
Cloud ranking and admin merchandising came after the baseline felt trustworthy. Dogfooding order matters: if Enhanced Native is embarrassing, cloud will not save trust.
What changed for humans
Counter staff and returning buyers stopped treating search as a last resort. Typeahead shortened the path to the right hose, gasket, or assembly. Zero-result pages happened less often — and when they did, the site had somewhere better to send people than a blank stare.
We are not publishing named conversion lifts without written approval. Qualitatively, the “can you find part X?” pings that were really search failures dropped, and navigational OEM-ish queries landed on the right PDP more often on the first try.
Lessons for other parts catalogs
Index the fields shoppers actually use (including attributes), filter dead inventory aggressively, and plan for click learning on the queries you already know from phone orders. If Magento or BigCommerce is your stack instead of Zen Cart, the same human failure modes apply — only the connector path changes.
Next step
If your Zen Cart catalog is parts-shaped, start at /solutions/zen-cart-search or /solutions/automotive-parts-search. Read the named Zen Cart dogfood story on Numinix.com for an install narrative you can click yourself. Enhanced Native is free on install; cloud ranking follows when you pair a tenant.