The Principia

Underneath the platform is a graph of your operation — not a database of it.

Most operational software answers one question well: where is this, right now. Each fact sits by itself — current, accurate, and disconnected from everything around it. That's tracking. It isn't understanding.

The Principia is a graph, not a table. Every carrier, hub, facility, shipment and asset is an object; every relationship between them carries its own attributes, source and history — assembled from the systems you already run, across ERP, EAM, MES, WMS, TMS, CMMS, SCADA, industrial IoT, EDI and live telemetry.

01

Model the operation as it actually is.

Objects and relationships, not rows. Every entity type declares the properties it may carry — each with a type, a unit, and whether it was derived or supplied by an operator — and every relationship is a permitted triple with a cardinality and a readable inverse. So what can connect to what is a documented answer rather than tribal knowledge. Both the automated derivation path and the manual write path validate against that same published definition: a misspelled property, or a relationship pointing the wrong way round, is refused with an explanation instead of being stored as a fact indistinguishable from a real one.

02

Extend the model without losing the audit trail.

The graph isn't read-only. Operators can correct and extend it directly, and every change is validated, attributed to its source, and versioned.

03

Take the feed as it actually arrives.

A source that carries columns nobody predefined — a dwell time, a fuel surcharge, a damage count, a temperature reading — isn't dropped for failing to match a known field name. Theralon profiles it, classifies what kind of measure it is, attaches it to the object it describes, and reasons over it exactly as it would a built-in field. A feed missing a field isn't discarded either: it's graded on what it can drive, so a source with no receipt data still populates the network and scores lanes, and a source with no carrier column still contributes its transit measurements. Nothing usable is thrown away for being incomplete against someone else's checklist.

04

Read the EDI, not just the order.

Theralon parses real X12 and EDIFACT interchanges and reconciles advance ship notices against receipt data automatically. The model therefore reflects what actually moved rather than what was planned to move — which is usually where the first honest signal of a problem appears.

05

Trace dependencies across the whole network.

Ask what fails alongside a partner, how a carrier connects to a hub, or where the network is most concentrated. Multi-hop questions are ordinary queries, not bespoke reports.

06

Rank what an object is really coupled to.

A single facility can reach hundreds of other objects within a few hops; drawing all of them produces a hairball, not a reading. Theralon surfaces only the strongest handful and states, in words, how strongly each one is coupled, on what measured basis, and what happens if that coupling is left unaddressed — the part a diagram can show but never say.

07

Reconstruct any past state.

Every materially-changed object and relationship is snapshotted to an append-only log, so the whole network can be rebuilt exactly as it stood at any past instant — not one object at a time. An audit, an insurance review or an incident investigation starts from a date, not a list of object IDs.

08

Keep every model its own.

Each organisation's graph is private and persistent, never blended with another's. Access within it is scoped per principal and enforced centrally, across risk items, compliance data and the graph itself.

09

Read the graph, not just render it.

Twenty accurate figures with no statement of which one matters is a data dump — the reader's real question goes unanswered while surrounded by evidence. Every object carries how it's performing, which way it's heading, and how much strain each connection is under, and every score decomposes into the factors that produced it. Theralon returns a short, ranked set of findings in plain language, each bound to the measurement behind it. No finding is published without evidence, and no threshold is invented: comparisons come from your own stated risk appetite, or from the network's own distribution. When nothing rises to a finding, it says so and names what it checked — a panel that renders empty without saying it looked is indistinguishable from one that's broken.

All of it is rendered on real geography — facilities, carriers and lanes in the physical space they actually occupy — on mapping infrastructure Theralon builds and hosts itself, with no third-party service anywhere in the request path.

"Competitors selling visibility are, at bottom, selling a very good list. Theralon is selling the map the list was missing."
See the platform built on it → Every domain intelligence reads and writes against this same graph.