Methodology Statement

Every Navos Intelligence brief carries a number and a view a client can plan against. This page sets out how we arrive at them: where the facts come from, how we build the forecasts, what our confidence words mean, how we frame risk, and how the work is produced and disclosed. It is the same standard behind every edition, so a reader can check the method against standards they already recognise.

How we source what we report

We grade every source by the authority it holds over the thing being reported, and we grade every claim by how well it is corroborated. The two are judged separately.

Source authority

• Mandated issuers. The institution with the statutory mandate on the figure, such as ELSTAT, the Bank of Greece, the ECB, Eurostat, the IMF, or the OECD.
• Financial press of record. Wire services and global financial press with specialist macro desks and a verify-before-publish discipline, such as Reuters, Bloomberg, and the Financial Times.
• Specialist and national press. Sector trade press, national financial press, and sell-side research.
• Commentary. Opinion, analyst notes, and social posts. These colour a read but never anchor a number or a confidence rating.

A claim that anchors a headline number needs at least one mandated-issuer source. A forward-looking read needs realised data from a mandated issuer plus independent corroboration for the path. A claim that stands on a single source is flagged as such, not presented as settled.

What our confidence words mean

We state two things separately in every judgement: how likely the outcome is, and how much confidence we hold in the evidence behind it. We keep them in separate sentences so a reader can tell a likely-but-uncertain call from an unlikely-but-well-evidenced one.

Likelihood of the outcome

This is the calibrated scale used by the Intergovernmental Panel on Climate Change. We use its exact wording so the terms carry a fixed, checkable meaning.

Confidence in the evidence

These confidence levels follow the analytic standard of the US intelligence community (Intelligence Community Directive 203).

How we build forecasts

We do not invent a number. We anchor each forecast to a panel of named institutions that publish under documented methods, such as the IMF, the OECD, the European Commission, the ECB, and the Bank of England, and to live market prices where a traded market exists. We take the middle of the panel as the base case and show the spread across it, rather than blending disagreeing views into a single figure that hides the disagreement.

Every forecast states the assumptions it rests on, such as the oil price, the exchange rate, and the policy-rate path. If one of those assumptions breaks, we revise the forecast and say what changed, rather than widening the range to absorb the surprise. We own the number we publish and stand behind it, and we show at least one named house above our view and one below, so a reader sees where we sit in the field and why.

How we frame risk

Each brief opens its risk section by stating whether the balance of risk is tilted to the downside, tilted to the upside, or broadly balanced. Every individual risk is a named, monitorable condition, never a vague category. For each one we state its direction, its probability on a Low, Medium, High scale, its impact on the same scale, how it would feed through, and the single event that would tell you it is happening.

This follows the risk-assessment convention the IMF uses in its country reports, so the register reads the way an institutional partner expects and stays comparable across editions.

How each edition stays current

Navos briefs revise when the evidence warrants, not on a fixed quarterly calendar. Each new edition inherits the prior edition's structure and view, then lands the changes that new data and events require. Every edition is complete on its own: a reader picking one up cold sees the full view without needing the previous week. A dedicated section sets out what changed since the last edition and why.

How Navos uses AI

Navos briefs are generated by the Navos AI analysis system, working under the methodology set out here and the specific contract for each Module. Source data is pulled from named statistical authorities and financial press. The analysis and the prose are AI-authored, then checked at publication by automated gates that enforce the source, calibration, risk, and language standards on this page.

Human control sits at the methodology, contract, and configuration layer, and published output is reviewed periodically. Navos does not claim that a person reviews and signs off every individual edition. Instead we disclose AI generation plainly, consistent with the EU AI Act. Navos AI, the publisher, holds editorial responsibility for every edition. Forecast accuracy is measured against what actually happened, and reviewed each quarter.

Important information