How ONX reasons

Confidence is composed.
Not asserted.

Most software shows you data and leaves the reasoning to the meeting. ONX is built the other way round: five architectural rules that make its recommendations explainable, challengeable and accountable to reality.

01

Evidence has a hierarchy

Every fact in ONX carries its evidence state: measured, modelled, inferred, stated or unmeasured. A margin promise built on an unmeasured assumption looks exactly as fragile as it is, before you commit, not after.

02

Confidence is composed, never asserted

Nobody types a confidence number into ONX. Win, delivery and commercial confidence are computed from the evidence each function publishes, then composed into one Business Confidence, weighted by evidence quality, explainable to the line item.

03

The worst constraint decides

A commitment is not ready when the average looks good. It is ready when every binding constraint clears. ONX converges every domain's clear-by date and recommends the earliest date that clears them all. One unvalidated launch assumption can hold the whole commitment, and should.

04

Patterns must be falsifiable

When ONX notices something nobody asked about, demand colliding with capacity, a claim exceeding the verified evidence, the pattern ships with its falsification conditions: the facts that would prove it wrong. Intelligence you cannot challenge is just opinion with a dashboard.

05

Outcomes close the loop

Every recommendation is scored against what actually happened. Overrides are recorded, not resented, and when reality disagrees with the model, the calibration changes. The system gets harder to fool every quarter you run it.

The evidence ladder

Five states, attached to every fact on the spine. The platform never lets an assumption impersonate a measurement.

MeasuredObserved in production, with the numbers on record.
ModelledDerived from a model whose inputs are known.
InferredRead from adjacent signals, plausible, unproven.
StatedSomeone said so. Recorded as exactly that.
UnmeasuredNobody knows yet. The most dangerous state, and the most honestly labelled.

The Intelligence module

Every other module is a system of record or understanding. Intelligence is the system of learning. It computes Business Confidence, detects patterns across closed decisions, and recalibrates the platform against real outcomes.

Business Confidence

The composite of Win, Delivery and Commercial Confidence: the single “should we do this?” score, weighted per deal profile.

Pattern detection

Segment, override, seasonal and cross-module patterns learned from closed Decision Outcomes: significance-gated, never guesswork.

Threshold recalibration

Every confidence threshold tested against real false-positive and false-negative rates, and tuned per segment over time.

Learning feed

Named learnings surfaced to leadership: “deals signed against this profile produced −4.2% margin”: the moat no competitor can buy.

Intelligence reads every module and every closed decision, then writes the composite score and recommended actions back, the connective tissue that makes ONX an operating system, not eleven tools.

Reasoning you can argue with.

Every conclusion answers “why do you believe that?”, with its evidence, its confidence and the conditions that would prove it wrong.

The whole philosophy, in one essay: Enterprise Decision Intelligence