Latest articles
No results for this search. Try another term.
Heurix turns your customers' intent into relevant results. Typos, jargon, or technical references — it always finds the right product.
What a developer sees — the real API response, on the "m8x20 stainless" example:
curl -X POST https://api.heurix.fr/v1/index/mycatalog/search \
-H "Authorization: Bearer YOUR_API_KEY" \
-d '{"q": "m8x20 stainless"}'{
"query": "m8x20 stainless",
"tokens": ["m8x20", "stainless"],
"total": 1,
"hits": [{
"product": {"id": "V1", "name": "Hex head screw", "ref": "M8 x 20 — A2", "stock": 100},
"score": 37.5,
"in_stock": true,
"matched": [
"annotation #DIAM_M8", "annotation #LONG_20",
"annotation #MAT_INOX", "annotation #VIS_M8X20"
]
}]
}The matched field explains the ranking — useful in development, not just in demos. Full documentation →
M8x20 → diameter, length, material. The annotation cascade reads structure, not just letters.
2 typos per word tolerated, 1 on references. "red turtlneck swetaer" still finds it.
Index, search, you're live. No plugin, no rebuild, no IT project.
Every result says why it ranks. Designed in France, hosted at OVH, data in the EU.
The search bar is the only place where a customer tells you exactly what they want to buy. Most sites answer "no results". Heurix exists to turn that moment into a sale.
Our engine relies on regular expressions capable of recognizing the structure of a product reference, where most engines only compare text.
Search platforms are built for the enterprise. Heurix is built for large catalogs that have neither the means nor the need for a dedicated data team.
Every improvement translates into a quantifiable conversion gain — not a quarter-long development project.
The default search engine on most CMS platforms indexes the catalog, but understands neither typos, synonyms, nor the relevance order a buyer expects. The visitor sees a blank page or unrelated results — and closes the tab.
Fixing this requires neither a new platform nor months of project work. It requires an engine that reads your references the way your customers think them: Heurix plugs into your existing site, via API, and understands "M8x20" from the very first call.
Median engine response time, measured on a 10,000-product catalog. Every millisecond of latency costs conversions.
"red turtlneck swetaer" still finds the red turtleneck sweaters. Only one on references — a changed digit means a different product.
14 days free trial on the Search API: up to 2,000 requests and 2,000 products, no credit card. See pricing.
Heurix is not a layer on top of Algolia or Elasticsearch: it's our own engine, built in France, specialized in technical catalogs. Where generalist engines compare characters, Heurix reads the structure of your references.
The demo at the top of this page runs on this engine — the numbers on the right are measured, not promised.
Most engines — including well-known solutions like Algolia or Elasticsearch when poorly configured — treat a product reference like any other word. Two different formats for the same reference (M8x20 and M8 x 20) are then seen as two unrelated things. On a technical catalog, that's where a meaningful share of lost sales happens — not the site's design, but the engine's ability to understand what the buyer actually typed.
Heurix configures custom regular expression (regex) rules for your catalog: they recognize the structure of a reference — family, dimension, variant — rather than comparing characters one by one. A buyer typing a partial or slightly mis-formatted reference still finds the right product. This is fine-grained catalog understanding, encoded once and for all into the engine's rule packs — it takes understanding both your catalog and how your customers actually search.
⚡ You never write a single regex yourself. Our rule packs — hardware, fashion, industry — are pre-configured and activate with your industry choice, at indexing time.
The larger a catalog or content base, the more a well-tuned search matters — because native search fails exactly where precision is needed most.
Compound sizes (W32/L34), colors, materials, variants — and customers who type fast, with typos.
Tens of thousands of references, standardized product codes (DIN, ISO), need to filter by dimension, material, compatibility.
Search by manufacturer reference or equivalence, tolerance for multiple code formats depending on the supplier.
ISBNs, editions, formats, collections: the same title exists in dozens of versions — the demo on this page is the proof.
Precise technical specs (voltage, amperage, standards), catalogs that evolve continuously.
High volumes, professional buyers who search by exact reference more than by descriptive keyword.
Vintages, estates, appellations, formats (37.5 cl, magnum): references where every digit matters.
Articles, FAQ, knowledge base: the same typo tolerance that finds a product reference finds a poorly-worded question. (Try Ctrl+K on this page.)
If your customers search by reference, dimension, or variant, the engine adapts. Tell us about your catalog.
A glimpse of what the engine can do — every capability is in the product today, not on a roadmap.
The Heurix signature: leveled regex rules that understand M8x20 as a diameter and a length. How it works →
"red turtlneck swetaer"? Corrected. 2 typos per word, only 1 on references — a changed digit means a different product.
6205-2RS, W32L34, M8.5: your references stay whole, instead of being shattered into meaningless pieces.
Hardware, fashion, industry — shipped with the engine. Custom packs available on Scale plans.
"Screw" finds "bolt". API-managed groups, automatic merging, no false positives. API →
Weighted fields, availability as tiebreaker at equal score — an out-of-stock product never outranks an equivalent available one.
Every result says why it ranks: term found, typo corrected, shared annotation. No more black box.
Every "no results" your customers hit is logged and available in your console — the concrete list of missed sales, not an estimate. See the console →
≈ 10 ms response time: fast enough to query the API on every keystroke, with no perceptible lag for the visitor.
Engine designed in France, OVH hosting — your catalogs never leave the European Union.
Internal search often accounts for only 10 to 20% of a site's traffic — but for many e-commerce businesses, it drives 30 to 50% of revenue. It's the site's most valuable segment, and the least often optimized. Adjust the sliders to your situation: the calculation updates live.
Public examples observed on comparable sites — not a guaranteed average or a promised result for your catalog.
Adjust the sliders to reflect your e-commerce situation.
Cost estimated from an annual Search API subscription (Growth or Scale plan depending on catalog size). Impact assumptions are order-of-magnitude estimates, not a promise.
Concrete cases and technical observations, published monthly.
Why default engines fail on technical product codes, and how regex changes the game.
Read the article →The economic leverage of internal search, and how to estimate precisely what it's costing you today.
Read the article →Concrete symptoms to check on your own site, no technical tools required, in under ten minutes.
Read the article →Via API: you index your catalog (one call), you wire up search (one call), you're live. No plugin to install, no rebuild, no dependency on your CMS. The quickstart guide fits in 3 calls. Integration guides with real code are available for Shopify, PrestaShop, WooCommerce, and Magento.
A 14-day free trial (up to 2,000 requests and 2,000 products) to test without a credit card, then plans from €19/month. Counting rules are public and your usage is available in real time — no end-of-month surprise. Plan details are here.
Generalist engines compare characters; Heurix reads the structure of your references through its annotation cascade. If your catalog is technical (dimensions, standards, sizes, vintages, editions), that's precisely what Heurix is built for — with rule packs per vertical, stock-aware ranking, and explainable results. For a simple blog or editorial site, a generalist engine is plenty.
No. Heurix plugs into your existing site via API, whatever your CMS — Shopify, PrestaShop, WooCommerce, Magento, or a custom build. Your catalog stays with you; the engine only replaces search.
The matched field of every result explains why it ranks, and the ROI simulator on this page gives you an order of magnitude based on your traffic and average basket. The most reliable method remains a before/after test on your zero-result queries — measurable in your analytics tool within the first week.
Three rule packs ship with the engine (hardware, fashion, industry) and the mechanics apply to any catalog searched by reference, dimension, or variant — books, wine, spare parts, electronics... If your references are very specific, tell us about your catalog: custom packs are available on Scale plans.
In France, at OVH (Gravelines or Strasbourg). Your catalogs never leave the European Union, and the engine itself is designed and built in France. For GDPR requirements, it's the simplest scenario to document with your DPO.
Never an abrupt cutoff: overage is billed at the usage rate shown on the pricing page. Counting rules are public (1 search = 1 request, 1,000 indexed products = 10 requests) and your usage is available in real time via the API.
Yes — it's the Heurix engine logic, ported to JavaScript and embedded in the page with a generated demo catalog: same normalization, same typo tolerance, same annotation cascade, same scoring. Developers can query it from the console: window.heurixDemo.search("...").
The 14-day free trial is enough to get started without writing to us. If you have a question before diving in, or your references are specific enough to warrant a custom rule pack (Scale plans), describe your catalog in a couple of lines — reply within 48 business hours.
Prefer to write directly? contact@heurix.fr
Up and running in 3 steps.
No plugin, no rebuild: your catalog, one API call, and the engine does the rest.
Index your catalog
One API call is enough — the catalog is created automatically, with your industry's rule pack (hardware, fashion, industry).
The cascade annotates every product
"M8 x 20 — A2" becomes
DIAM_M8+LONG_20+MAT_INOX. Your customers' queries go through the same rules: matching happens on meaning, not spelling.Your customers find
Despite typos, despite different spellings — sorted by relevance then availability. And every result explains why it matches.
The full endpoint reference is in the API documentation. Integration guides with real code available for Shopify, PrestaShop, WooCommerce, and Magento.