Product → Who it's for → Pricing → About → Trust → Get a free sample

About

Product data should carry its sources

Specs are still typed in from PDFs by hand. AI can do it, with a source for every value.

30,000 values in a 1,000-product catalogue, every one needing a source

Every value in a catalogue should be one click from where it came from.

Why we built this

Product data is the one thing in the chain that nobody owns. The manufacturer has the specs, written by the engineers who designed the part. The distributor needs them as fields a buyer can filter on. Turning one into the other is nobody’s job, so it happens by hand, late, or not at all.

We came at it from the other side. Reading a datasheet and filling a fixed set of fields is close to what language models are now best at. The reason it had not already been done is not the reading. It is that nobody could check the result. A thousand products across thirty fields is thirty thousand values, and no one is going to verify those by eye.

So the checking came first. Every value an agent writes carries the page it was read from, and anything the documents do not settle arrives as a question rather than a guess. That constraint is the product.

How we work

Four things that hold on every project.

  1. We start on your products.No demo catalogue and no sandbox. The first thing you see is twenty-five of your own articles, finished.
  2. Your documents are the only source.Agents read the datasheets and catalogues you already have. A value your own papers do not support does not get written.
  3. You approve before anything is published.Records reach a data pool when you say so, and not before.
  4. The score is agreed in writing.We set a target fill rate with you before the work starts, and the pool measures it, not us.

Illustration with example values.

What we don’t do

Every job has edges. These are ours.

Illustration with example values.

Prices and stock

Those live in your ERP and change weekly. We leave them alone.

Article numbers and GTINs

We work with the identifiers you already use. We do not create new ones.

Marketing copy

We fill the fields a distributor filters on, not the description that sells the product.

Software to run

Nothing to license, install or log into. You get finished data in formats your systems already accept.

Where we are

Founded

2026

Based in

Netherlands

First data pool

2BA

Classification

ETIM 10

We start with lighting and electrical installation, where the ETIM classes run deepest and the gaps are widest.