Home / AI and automation / AI search
Approsearch · AI search
AI search for technical catalogues: Approsearch
In a B2C store, when search fails, customers browse the categories. In a B2B catalogue with 40,000 products, when search fails, customers pick up the phone or buy somewhere else.
A 60-minute session. We review what people search for on your site today, what they fail to find and how much of it ends up as a phone call.
Searches that return zero results today
3RT2016-1BB41The manufacturer’s code, not your internal SKU.0 resultsM8 x 40The same size, written differently.0 resultsAllen boltThe name used on the shop floor.0 resultsREF-4512A discontinued part with no replacement mapped.0 resultsrodamiento 6204Another language, another country, same part.0 resultsWith Approsearch, all five lead to the product.
The problem
Why search fails in your technical catalogue
A text-matching search engine finds what is written exactly the same way. Industrial buyers almost never write it the same way. These five cases account for most zero-result searches in a technical B2B eCommerce catalogue.
The customer searches by manufacturer code
They type 3RT2016-1BB41 because that is what is printed on the equipment in front of them. Your product page indexes it as CONT-9A-24V, your internal SKU. Literal search finds nothing, even though the product is in stock.
The customer writes the size differently
M8x40, M8 x 40, 8x40 mm. The same part and three different text strings. A literal index treats each variant as a different query and only matches the one that is written like your product page.
The customer uses trade vocabulary
They ask for an Allen bolt and your catalogue describes it as a DIN 912 hexagon socket head cap screw. Neither is wrong. They are two vocabularies: the shop floor’s and the standard’s.
The customer searches for a discontinued part
They type REF-4512, which they bought three years ago and you no longer make. They need to know what replaces it. A conventional search engine returns zero results and the customer concludes you don’t have it.
The customer makes a typo or uses another language
baering, rodamiento 6204, roulement 6204. One extra character or a term from another country and the result disappears.
They all end the same way
Each of these searches ends in a call to your sales team or an order that goes to another supplier.
Measure how many you lose on your siteSemantic search
What makes AI search different
A conventional search engine compares letters. A semantic search engine compares meanings. The second finds the product even if the customer doesn’t use a single word from your product page.
To do this, every product in the catalogue is turned into an embedding: a numerical representation of its meaning that places products meaning the same thing close together, however they are written. When a query comes in, it is also turned into an embedding and the closest products are retrieved.
Synonyms and trade vocabulary
It recognises that an Allen bolt and a DIN 912 screw are the same thing, and learns your industry’s vocabulary from your catalogue, your technical documentation and your customers’ real searches.
Equivalents and cross-references
It links your internal SKU to the manufacturer code and to discontinued parts: a query for any of the three leads to the current product.
Normalised sizes and units
It reads M8x40, M8 x 40 and 8x40 mm as the same specification and matches them against the product’s technical attributes.
Typos, languages and images
It corrects typos, understands queries in several languages and lets customers search with a photo of the part, with autocomplete as they type.
Our own product
Approsearch
An AI search platform developed by eComm360 for complex catalogues and industrial environments. It is built on Elasticsearch and includes semantic and visual search.
What you configure. Industry vocabulary, equivalents between parts, result ranking rules and permissions by customer profile. Day-to-day use needs no technical knowledge: everything is managed from the dashboard. For advanced customisation, the eComm360 technical team works alongside you.
Where it works
What’s included
- Autocomplete with contextual suggestions and automatic typo correction.
- Dynamic filters that adapt to the profile of the user searching.
- Unified search across the catalogue, PDF datasheets, documentation, CMS and blog.
- Automatic content tagging and image search.
- Results that adapt to each customer’s behaviour.
- Embedded AI agents for assisted search inside the search engine itself, built with the same method as the AI agents that carry out tasks in the ERP.
- A dashboard with performance indicators and zero-result search analytics.
Open architecture
Without the limits of a closed solution: it adapts to your catalogue, your processes and your architecture.
Built in-house
We develop and maintain it ourselves, with full control over its features.
Designed for B2B
Customer-specific pricing, approved catalogues and technical documentation: what a professional buyer needs.
Specialist support
A technical team that knows your industry and keeps improving search based on what the analytics show.
Integration
Search that knows your customer
Finding the product is half the problem. The other half is showing the right price, real availability and only what that customer can buy.
Approsearch runs on Integrafy, eComm360’s integration middleware with more than 100 deployments. It syncs from the ERP the catalogue with its technical attributes, the PIM with datasheets and documentation, segmented price lists and multi-warehouse stock. Without that connection, search finds products. With it, search finds products the customer can actually buy.
How Integrafy AI connects data to searchThat customer’s price
Results show the negotiated price of whoever is searching, not the general list price.
Approved catalogue
Each customer sees the products in their contract. Anything else does not appear.
Stock at their warehouse
The availability shown is that of the warehouse serving that customer, not the total across all of them.
Measurement
What we measure
Four indicators tracked in the Approsearch dashboard and in the monthly review of the outsourced AI department.
Zero-result searches
The percentage of queries that return no products. Each one is a customer who didn’t find what they came to buy, and the best source for spotting gaps in the catalogue and vocabulary.
Search conversion
The share of sessions using search that end in an order, compared with those that browse the categories.
Catalogue depth
How many different products actually appear in results. A 40,000-product catalogue where only 3,000 are ever seen has an access problem, not a range problem.
Calls your sales team no longer takes
How many availability and part enquiries stop arriving by phone or email. It is the metric that turns search into hours freed up.
Deployment
How it is deployed: two common setups
Two ways to get started, depending on where your search problem is today.
Setup 01
Spare parts distributor with cross-references
- Situation
- The catalogue mixes internal SKUs, manufacturer codes and discontinued parts. Customers search by any of the three and search only recognises one.
- Setup
- Approsearch indexes the catalogue synced through Integrafy, adds equivalents as cross-references and normalises sizes and formats.
- What changes
- A search by manufacturer code or old part number leads to the current product. The zero-result searches that remain point to real gaps in the catalogue.
Setup 02
Manufacturer with scattered technical documentation
- Situation
- Specifications sit in PDF datasheets, the online catalogue only has a short description and the customer needs to check a dimension before ordering.
- Setup
- Approsearch indexes the catalogue, datasheets and documentation in a single search, with PIM attributes synced through Integrafy.
- What changes
- The customer finds the part and its specification in one place. The technical team stops forwarding datasheets by email.
FAQ
Frequently asked questions about AI search
Any other question can be answered in the free assessment.
Does it work with my current platform?+
Yes. Approsearch integrates with PrestaShop, Shopify, Magento, WooCommerce and any system via API, as well as with your ERP, PIM and custom feeds. There is no need to change platform.
How is it different from Doofinder or Motive?+
Approsearch has an open architecture: it can be fully customised, indexes external content such as PDF datasheets, documentation or the blog, and includes features designed for B2B, such as customer-specific pricing, approved catalogues and embedded AI agents.
Do I need to restructure my catalogue?+
No. Approsearch indexes the catalogue as it is, with the attributes it already has. Semantic search reduces the reliance on a perfect taxonomy, and zero-result search analytics show what is worth completing.
How does it learn my industry’s synonyms?+
From three sources: the catalogue itself and its technical attributes, the indexed documentation and your customers’ real searches. Specific vocabulary is configured and extended based on what the analytics show.
Does it respect each customer’s pricing and approved catalogue?+
Yes. With the ERP connected through Integrafy, results show that customer’s price, their approved catalogue and the stock at the warehouse that serves them, in real time.
Can I search inside my technical documentation?+
Yes. Approsearch connects to your CMS, your blog, your technical PDFs and any other content source relevant to your customers, and shows it in the same search as the catalogue.
How does image search work?+
The customer uploads a photo of the part and Approsearch shows the products that look like it. The same AI automatically tags the catalogue images so they can be found.
How long does it take to get started?+
It depends on the size of the catalogue and whether the ERP is already synced. With Integrafy in place, indexing and initial setup are the short part of the project. The exact timeline is set in the proposal after the assessment.
Free assessment · 60 minutes
See what people search for in your catalogue and don’t find
We analyse the real searches on your site, how many end with no results and how much of that volume ends up as a call to your sales team.
Barcelona · Valencia · Bilbao · Quito