Diagnostic Intelligence
for Hospital Material Flow

Your ERP has the data. We make it speak.

Every hospital has the data.
No hospital can read it.

Materials are received, moved, consumed, and billed across disconnected systems. The gaps between these systems are where revenue leaks.

Seven systems. Three languages.
Two decades of schema drift.

Warehouse
Ward
OR
Billing

Materials management doesn't see billing. Billing doesn't see logistics.
The CFO gets a variance report with no explanation.

From raw exports to certified findings

1
Ingest
Load every row from any ERP. Nothing is judged, nothing is dropped.
2
Translate
Map vendor-specific columns to one canonical schema. Once.
3
Validate
Flag every row valid or invalid with a human-readable reason.
4
Diagnose
Run probes, score entity health, test business hypotheses.

Three layers. One truth.

Bronze
Raw data, faithfully loaded. Every row, every column, exactly as it arrived. Tagged with source file and line number.
Silver
Validated truth. Every row flagged valid or invalid with a human-readable reason. Nothing silently dropped.
Gold
The certified product. Only valid rows. Versioned contract. Everything downstream codes against Gold.

Invalid rows aren't deleted.
They're diagnosed.

is_valid: true
Case exists. Material in master file.
Quantity positive. Status recognised.
→ Passes to Gold.
is_valid: false
invalid_reason:
"case_token not found in silver_cases"
→ Stays in Silver. Queryable. Countable.

Data quality is not a prerequisite for analysis. It is analysis.

One pipeline. Any ERP.

OPALE
SAP
Navision
Canonical Model

Vendor translation happens once, at the boundary.
From that point forward, every hospital looks the same.
Onboarding a new ERP is a translation layer — not a rebuild.

A standing infrastructure.
Always running. Always scoring.

Action Lists
prescribed remediation
Diagnosis
root cause analysis
Assessments
entity health scores
Probes
detect symptoms in data

From the outside world. A hunch. A concern.

“I have a hunch that we don't invoice everything we are actually using in a case…”
— a CFO who can't sleep
activates
Assessments
Probes
verdict

The hypothesis drives the inquiry. The engine delivers the evidence.
The verdict emerges where they meet.

The engine answers.
In one of four ways.

Confirmed
Strong evidence supports the concern
Plausible
Some evidence, warrants investigation
Not Observed
The data does not support the concern
Insufficient
Not enough data to judge

Every verdict is backed by weighted evidence — traceable to specific probes and findings.

Automated auditors that never sleep

We test our own detectors.

Plant
defects
Known errors
at known rates
Run
probes
Probes don't know
the answer key
Measure
accuracy
Precision & recall
per probe

When we say a probe catches a certain percentage of cases,
that's a reproducible, testable claim — not a marketing number.

Not a prototype.
A working system.

Tested on real Swiss hospital data — real cases, real materials, real billing. Multiple sites, multiple years. The probes find things. The findings have real-world consequences.

The hospital doesn't have one supply chain.
It has two.

Warehouse logistics and patient-level consumption operate as parallel systems. Some materials appear in both. Many appear in only one. Nobody had measured the gap — until now.

Every material falls into one of three groups

Both Systems
Tracked in warehouse transfers and patient consumption. The only group where the I/O coefficient is meaningful.
Transfer Only
Delivered to wards, consumed from floor stock. Nurses take without scanning. Expected for commodities.
Usage Only
Dispensed directly from pharmacy. Never touches the warehouse. Correct by regulation for specialty drugs.

This isn't a defect. It's how hospital logistics actually works.
But nobody had measured it before.

Same product. Two package sizes.
Completely different tracking.

The smaller size, used in the operating theatre, tracks perfectly — every unit delivered is recorded as consumed.

The larger size, used on the ward, shows a fraction. Stable month after month, year after year.

Not an error. A structural feature of hospital logistics — now visible for the first time.

The coefficient is not a KPI to optimise.
It's an architectural constant.

It measures how tightly coupled two parallel systems are.
The operating theatre shows what perfect tracking looks like.
The ward shows what normal looks like.

The question nuMetrix answers:
which materials should look like the OR — and don't?

Every hospital is an island.

Hospital A
schema_a.*
Hospital B
schema_b.*
Hospital C
schema_c.*
Hospital D
schema_d.*

Same pipeline. Same probes. Same models.
But the data never touches. The isolation boundary is absolute.
A platform layer unions across hospitals for benchmarking — when permitted.

Everything speaks your language

DE
FR
EN

Every probe finding, every hypothesis verdict, every PDF report.
German. French. English. From the start. Not bolted on later.

We never see patient names.

All identifiers are pseudonymised tokens. The system tracks materials, costs, and processes — never individuals. Each hospital's data is fully isolated.

Three ways to see your data

Explorer (Browser)
Per-hospital drill-down into dimensions, findings, and hypothesis verdicts. Any modern browser, no installation.
Explorer (Desktop)
Native desktop app. Opens the database locally, works offline. No server, no browser dependency.
PDF Reports
Data Quality. Financial Risk. Health Cards. Tri-lingual. Ready for the audit committee.

Every finding is traceable.
Every result is reproducible.

Probes have version numbers. Findings have evidence chains.
An external auditor can re-run the same pipeline on the same data
and arrive at the same numbers.

We sit outside the ERP.
We read exports. That's it.

No interface development. No middleware. No installation.
No access to the production system. No risk to operations.

One data export is the starting point.

We need one thing from you.

A CSV export from your existing ERP.
We will build your validated dataset.
We will run the probes.
We will hand you a report — in your language,
with your materials and departments named.

For every material consumed,
was it properly tracked and billed?

We answer that question. At scale. Across systems. Today.

nuMetrix