Case study
Aronlight: the catalog that proved it.
A multi-brand lighting distributor in Portugal and Spain, and our first signed client. Their catalog and their inbox are what the product was built against — which is why the hard parts of it are the hard parts of lighting.
Sample data
The challenge
Orders and quote requests arrive by email across several mailboxes, in Portuguese and Spanish, written the way customers talk: a competitor’s code, a photograph of a handwritten list, “the usual driver”. Behind each one sits an Odoo catalog of more than 4,300 SKUs across two companies, each with its own price lists and VAT rules. Every line was resolved by a person, by hand, before a quote could go out.
What it does for Aronlight.
Each figure says whether it is measured from Aronlight’s production data, to September 24, 2026, or an estimate.
~3,000
Requests a month
Across all of Aronlight’s channels, by its own estimate.
~780
Read automatically each month
Measured: 614 orders and 166 quote requests in 30 days, read and matched without anyone retyping them.
70%+
Could need no human intervention
Our estimate of the share Ettore could handle on its own, once it covers all of them. Not yet measured.
3 → 1
Catalog sources brought together
Odoo, a 132-page price table and 20 technical-datasheet spreadsheets, now read as one, so every line is matched against all of it.
Why it works for a lighting distributor
01
It matches against their catalog, not a search box
Customers write a competitor’s code, a discontinued reference, or a description with no code at all. Ettore resolves each line against Aronlight’s own catalog and applies that customer’s pricelist, so the draft carries the price that customer actually pays. A near miss is flagged for a person rather than quietly swapped.
02
It reads lighting the way the trade writes it
Wattage, colour temperature, IP rating and beam angle arrive inside a sentence rather than in fields, and an accessory usually has to follow the fixture. Resolving that is the hard part of lighting, and it is the part the system was built against.
03
It gets better every time a rep corrects it
Reviewing is not just a safety net. When a rep changes a match, that correction feeds back, so the next request that looks like it resolves closer to right. The catalog’s habits — the shorthand, the substitutions, the customers who always mean one particular thing — are learned rather than configured.
What was built.
The whole path, from inbox to quote
Mail intake, triage, request understanding, catalog matching against the real product data, live stock and the customer’s own price, then review. Seven stages, all in production.
Connected to Odoo
Their ERP holds the catalog, the customers and the pricing, and keeps holding it. Ettore reads from it and writes approved quotes back.
Free text in, structured lines out
Bodies, PDFs, spreadsheets, images. What arrives as a sentence leaves as line items resolved to real SKUs — or as an honest statement that a line could not be resolved.
A person still presses send
Every delivery step needs a human click. That is architectural, not a setting — and it is written into the contract, which states the service does not send proposals autonomously to their customers.
The target.
The build set out to take quote turnaround from the 30 to 60 minutes a rep spends by hand down to under five minutes of review. That is the design goal the system was built to, stated at the outset — not a measured post-launch result, and we would rather say so than publish a number we cannot show you the working for.
Design goal, not a measured outcome.
Questions about this engagement.
Is this live, or a pilot?
Live. Aronlight is a signed client and the system is in production against their real inbox and their real catalog.
Does Ettore send quotes to Aronlight’s customers?
No. Every outbound step requires a person. That is how the system is built and it is also a term of the contract.
Would this work for a distributor that is not in lighting?
The shape generalises: a catalog of discrete SKUs in an ERP, meaningful request volume by email, a back-office team doing the matching by hand. Lighting is where it was proven, not the limit of where it applies.
How long did it take to connect?
About a month from start to live, including setup, a demo on their own requests, a training session, and a paid pilot before scaling up. The dependency that mattered most was how quickly ERP access was approved.
Your catalog is the interesting part.
Bring a real request and the catalog behind it. That conversation tells us both far more than a demo of someone else’s data.