Sequence carries the finding
Every material conclusion above came from order and interval rather than totals. Balances conceal rhythm, and rhythm is where behaviour shows.
IndustriesForensic accounting and disputes
The ledger balanced.The value never moved.
An intercompany account that reconciles perfectly can hide the fact that nothing was ever transferred. Nexyron shows you which one you have.
Five tools that have never spoken to each other.
Tabular analytics for population testing. A review platform for documents. A charting tool for relationships, populated by hand. A capture tool for scanned statements. A spreadsheet holding the whole thing together.
Each one is good at its own layer. The question you are actually instructed on spans all five, so the joining is done by a person, at night, against a court deadline.
The finding is almost never inside one of those tools. It is in the interval between two of them.
An opinion on intercompany movement needs the ledger, the bank statements, the funding agreements, the correspondence and time. Five sources, four systems, one deadline, and the joining is the engagement.
What Nexyron does instead
Assemble it once. Then question it as many times as the matter needs.
Nexyron takes the ledger, the bank data, the agreements, the correspondence and the entity structure into one connected model in which time is a property rather than a separate system. The email and the payment sit in the same analysis, so the nine days between them is a computation instead of a discovery somebody makes at 2am.
What was measured, what was extracted from a document and what was inferred stay separable at every point, each with a query back to the evidence that produced it. An opinion that blurs those three does not survive cross-examination, and the engine is built so that you cannot blur them by accident.
Underneath, 63 graph algorithms and a semantic index over records and documents are one query surface. Community detection, shortest path and shared-neighbour weighting run against the same model that holds the postings.
Your review platform keeps privilege and production. Nexyron is where the analysis happens once everything is finally in one place.
A simulated engagement
Constructed to show the method. No client, no matter, no case study.
- 01
What the ledger omits
Fourteen entities, intercompany postings placed in time per entity pair rather than netted into balances. Two pairs change character seven months before the insolvency date: same volume, intervals lengthening from days to months.
- 02
Did anything actually move
Bank statements matched against those postings. One pair settles throughout. The other does not. From the same point, postings continue and transfers stop.
An account that reconciles on paper with no cash behind it is a different object entirely, and only the sequence distinguishes them.
- 03
What were they obliged to do
The funding agreements, guarantees and security documents read beside the transaction record. A set-off arrangement changes what the balances mean. A guarantee is engaged by the change in settlement behaviour.
- 04
Who knew, and when
Finance mailboxes and the payment record on one timeline. An exchange nine days before the change discusses the funding position. A second, two days after, instructs a change to the payment run. Both are ordinary correspondence. Both acquire meaning from their position in the sequence.
- 05
When it started
Each of 2,300 vendors assessed against its own history rather than a global threshold. Eleven show the same signature at the same time: first invoice within one quarter, unusual cadence, values just below an approval limit. Combined value modest, which is exactly why no threshold test ever found them.
- 06
Are the eleven related
Shared attributes weighted by rarity. Three share a bank account. Two share an address with a former employee. Six show nothing beyond timing.
That has not proven a scheme. It has produced a subset worth an interview and a subset worth a registry search, out of the group's own records, in an afternoon.
Assembled once. Questioned many times.
Documents are evidence, not attachments
The clause that changes the meaning of a balance sits inside the same analysis as the balance.
The subject is the population
Two thousand three hundred vendors, each against its own baseline. That is where a defensible period of loss comes from.
Fact types stay separable
Measured, extracted and inferred, distinguishable at every point, with the evidence attached to each.
The second question starts from the first
Reports, features and recorded analyses stay addressable. On a matter running eighteen months, that is the difference between compounding and starting again.
The dimension nobody in this category occupies
How much was lost, argued from the record rather than asserted.
The tools in this field are built to find and to present. None of them is built to estimate what would have happened otherwise, which is the question every damages schedule and every causation argument actually turns on.
Nexyron carries 42 causal and decision procedures on the same graph as the transactions. Counterfactuals constructed rather than asserted, comparable populations identified rather than assumed, and an explicit refusal where the data cannot support an estimate.
An expert who can show why a number is estimable, and say plainly where it is not, is in a considerably stronger position than one whose schedule rests on a modelling assumption nobody tested.
Built to be challenged
Published methods, and a complete trail.
Community detection, centrality, shortest path and shared-neighbour weighting are peer-reviewed techniques with decades of literature behind them. When method is challenged, the question is whether the technique is sound and generally accepted, independently of whose software ran it.
Every conclusion Nexyron produces carries the records behind it, the time basis, and the transformations applied to get there.
Running it
On the engagement's own hardware.
Privileged material becomes complicated the moment it moves through third-party infrastructure, and cross-border matters restrict where data may lawfully be processed at all.
Nexyron opens as an application on a machine you own and runs the entire engine there. The material never leaves, which removes an argument before opposing counsel gets to make it.
Take an engagement you closed this year.
Estimate the hours that went into joining evidence rather than analysing it, and how much of your conclusion was already visible in week one, had anybody been able to see all of it at once.