Non-uniform codes
Different coding criteria between categories and suppliers. No unified logic applied over time.
Case Study · Medical gas equipment
Rosiglioni Impianti
When the catalog is an opaque system
Sector: Medical gas equipment
Type of work: Operational diagnosis, product system mapping, AI-assisted data enrichment
The signal
5,300 items. Twelve functional categories. Dozens of suppliers. An ERP system bearing the weight of daily operations.
But when Rosiglioni Impianti decided to build a digital configurator for medical gas installation design, a problem surfaced that almost no one had named out loud: the underlying data was not governable.
It wasn't wrong. It was opaque.
The diagnosis
Rosiglioni's item catalog had grown through stratification — a new supplier, an additional product, a variant never consolidated with the previous one. Year after year. The result was a system where knowledge was distributed unevenly: some areas well-maintained, others nearly abandoned.
Before touching anything, we did a structured reading of the existing system.
71% SPEC = GEN
The SPEC field valued as "generic" in 71% of cases — data that carries no useful information for the system.
ERP data separated
Prices and stock not present in the catalog. Managed separately in the company ERP. No integration.
2,763 items in "Accessories"
A category with over 2,700 items. The classic signal of a system that stopped classifying and started accumulating.
What wasn't visible from the outside
No explicit management of obsolete or replaced products. Classification knowledge that existed only in people, not documented anywhere. Descriptions written in completely inconsistent ways — some precise to the millimeter, others just three words.
This is not a judgment. It's what happens when a company grows in a healthy way and never had time to stop and sort things out.
The method
We didn't start with software. We started with structure.
Phase 1 — Observation
Quantitative analysis of the existing catalog: distribution by category, field completeness, patterns in legacy codes, data anomalies. The analysis produced a precise map of where the problems were and what their nature was.
Phase 2 — Mapping and stress testing
Working sessions with Sabrina Rosiglioni to validate the classification logic. Every decision was stress-tested: does this category hold when the configurator needs to automatically select components? Is this code memorable for someone who's never seen it before?
Non-obvious decisions were documented explicitly — with the reasoning and rejected alternatives. A knowledge asset as valuable as the catalog itself.
Phase 3 — Diagnostic output
New MACRO architecture in 12 functional categories. New SKU coding logic. Mapping of 214 suppliers with 3-letter codes assigned to the 24 main suppliers. Integration of ERP data. Explicit management of obsolete products with -OLD suffix. PENDING sheet for items still to be classified with the team.
The role of AI
On a catalog of 5,300 items, part of the work was automatable. Part was not.
We used AI to: infer the correct SPEC from text descriptions for 104 generically classified items; identify substitution patterns in legacy codes and internal notes; extract the commercial category (component, spare part, kit, consumable) from descriptions where readable; prepare the system for description enrichment from supplier technical PDF datasheets.
We did not use AI for: classification decisions (those require technical judgment and domain knowledge); validation of economic data (requires comparison with real commercial agreements); handling ambiguous cases (requires dialogue with the team).
AI accelerated the mechanical work. Judgment remained human. This distinction is not rhetorical — in a sector like medical gas, where a wrong item in an installation has concrete consequences, precision is non-negotiable.
What we learned
An opaque catalog is a symptom, not a cause.
The cause is almost always the same: a company that has rightly prioritized operational growth over data structuring. Not a mistake. A rational short-term choice. It becomes a constraint the moment you want to make the digital leap.
The equipment configurator Rosiglioni is building could not have relied on the previous catalog. Every automatic selection would have been unreliable. Every generated quote would have required manual verification.
The system diagnosis was the prerequisite. Not the accessory.
Recurring pattern: opaque data in industrial SMEs
Rosiglioni is not an isolated case. It's an archetype.
In almost every SME operating in technically complex sectors — equipment installation, spec-based manufacturing, technical distribution — we find the same dynamic: years of growth through accumulation, knowledge distributed among people, systems that hold but don't scale.
The critical moment comes when the company decides to make the digital leap. Configurators, advanced CRM, automated quoting: all require clean, structured, governed data. Without that foundation, technology doesn't help. It amplifies chaos.
Our intervention on this type of system includes:
01 — Structured reading
Analysis of existing data without bias and without pre-packaged solutions. Diagnosis precedes any intervention.
02 — Dialogue with the system
Sessions with those who live the system every day. Tacit knowledge is the most valuable data — and the hardest to recover once it's gone.
03 — Minimal architecture
Not the perfect system, but the one the team can maintain. Governability is worth more than perfection.
04 — Decision documentation
The value is not in the file. It's in the reasoning that produced it. Every choice is documented with reasoning and rejected alternatives.
The strategic and identity perspective
This document describes the technical and systemic dimension of the project: data diagnosis, structuring method, AI's role.
Kredo Marketing handled the strategic dimension: how the catalog restructuring connects to product identity, the company's repositioning, and preparation for international markets.
Kredo Diagnostics
Systemic diagnosis for B2B SMEs
We work with companies operating in technically complex sectors that need to structure their data before investing in technology.