Customers arrive with a problem, not a part number.
CogQuery turns your catalog into a conversational, self-improving search experience, so customers find the right part the first time, even when they describe it in their own words, upload a photo, or ask a question.
Built for catalogs where search is hard
If your support team fields "which filter fits a Kirby Sentria?" calls, or a search for "Hoover belt" returns 400 results, CogQuery was designed for you.
Parts retailers
Appliance, vacuum, auto, HVAC, marine. Customers know their machine model, not the SKU.
B2B distributors
Deep technical catalogs where compatibility and cross-references are everyday work.
Repair-driven shops
Customers are mid-fix, photographing a broken part with their phone for the fastest answer.
Specialty retailers
Inventory full of industry jargon the average customer will not Google correctly first try.
Not generic search. Not DIY infrastructure.
Most catalog search is either shallow (plugin-grade) or powerful but domain-blind (Algolia, Elastic). CogQuery sits in a different category.
It learns from your customers
Every click and cart-add is a signal. Rankings improve automatically over time, no manual tuning required.
It understands compatibility
Natively models "fits", "replaces", "substitutes for" - the questions parts customers actually ask.
It handles images, not just text
A customer photographs the broken part on their counter and gets matches back, even with dust or a bad angle.
It speaks natural language
"What bag fits UH30600?", "upgrade for DC41", "kirby belt" all work - no special syntax to learn.
It is multi-tenant by design
One deployment serves many merchants with isolated data, branding, and learning. Zero cross-contamination.
It is API-first
Storefront, mobile app, partner portal, voice, or headless commerce - all the same endpoints, all the same intelligence.
Everything a parts customer needs to find it
Natural-language understanding
Smart query routing decides if a question is a product hunt, a compatibility question, or both. Phonetic + fuzzy matching auto-corrects "Hover" to "Hoover"; synonyms map "sack" to "bag".
CogFinder fitmentNEW
"Oil filter for my 2020 Ford F-150" parses year/make/model from the query and returns only products that fit. Define your own cascading taxonomy; fitting products soft-boost to the top.
Search with a photo
Reverse image search by visual similarity, with a built-in crop tool. Handles HEIC, JPEG, PNG. Model/brand context narrows matches; similarity scores prevent forcing a bad result.
SearchandisingNEW
Merchandise the results page like a storefront. Build campaigns tied to specific search terms - pin, boost, or bury products, feature a promotion, or surface a seasonal bundle - then A/B test each campaign and keep the version that lifts revenue.
Self-improving rankings
Click-through learning ranks the products customers actually click higher; add-to-cart counts more than a passing click. Admins see every learned pair and can block or pin it. A "★ Learned" badge shows the system working.
AI admin assistNEW
A nightly Zero-Result Coach classifies failed searches and proposes one-click fixes that write real catalog entries. The Results Narrator turns every A/B test into plain English: "Treatment won, +12% revenue, ship it."
A/B testing + revenue liftNEW
Test learning on/off, ranking weights, and routing with session-bucketed assignment and revenue lift at 95% confidence intervals. A 3-step wizard lets a non-technical admin configure a real experiment in under 90 seconds.
Analytics that close the loop
Executive dashboard for searches, hit rate, CTR, conversion and latency percentiles - all with period deltas. Zero-result queries come with one-click "teach the system" right from the failed search.
Search that gets better with use
Unlike traditional search that degrades as the catalog grows, CogQuery improves as more customers use it. Switching away means losing months of accumulated learning.
Customer searches
A query comes in - typed, spoken, or a photo. SmartSearch routes and returns results.
They click & buy
Clicks and cart-adds are captured against a tracking token. Conversions count more than clicks.
It learns nightly
Rankings update once per night above configurable thresholds, so patterns stabilize without chasing noise.
Next search is better
The right products surface higher for the same query - and the loop compounds with every customer.
Buy once, deploy everywhere
Turn your catalog into answers.
Bring a slice of your products and a handful of the searches your customers actually type. We'll show you CogQuery finding the right part the first time.