How Northline research is made
What a reader can rely on, and what a reader should check. These standards describe how the publication works today; where a safeguard is automated rather than done by a person, we say so.
Every figure traces to a dataset or a source
Not typed in by hand.
- Figures in research are inserted from the published datasets by reference, with the period and unit; a figure that cannot be resolved stops the build rather than being shown blank or guessed.
- Statements that come from a report, budget or release are registered with the publisher, title, date and the exact wording, and are checked by script against the publisher's page where it can be fetched. Some pages cannot be fetched automatically; those are listed as unchecked, not as confirmed.
- Rankings and comparisons are computed from the data. Nominal and real values, levels and growth rates, and different periods are kept apart.
Prepared with AI assistance, reviewed by a person before publication
Disclosed on every research document.
- Research drafts are prepared with AI assistance. Every research document carries that disclosure.
- Automated checks (reference resolution, linting, source matching) are quality controls. They are not human editorial review and are never described as such.
- A document is shown as approved or published only after a named person has recorded that decision against the exact text. Changing the text afterwards withdraws the approval.
What appears in this edition
Approved documents only.
Only documents that a named person has reviewed, approved and published appear in this edition. Documents that have not been through that step are not part of it, and nothing on this site describes a machine check as a human review.
Facts, interpretation and forecasts are labelled
They are not blended.
- Where the data show a pattern, the research says what the data show and what they cannot show. Interpretation is marked as interpretation; correlation is not presented as cause.
- Northline does not publish economic forecasts. Scenario and sensitivity work, where it appears, is conditional on stated assumptions and is not a prediction. Experimental modelling is restricted and labelled.
- Descriptive economic research only. Nothing here is investment, tax, legal or real-estate advice.
Sponsors and clients do not shape findings
Sponsorship is labelled.
- Sponsored content, when it exists, is labelled and kept apart from editorial content. Sponsors and clients cannot influence findings or which stories are covered.
- This site publishes no client names, testimonials or audience figures, because none are available to report.
Errors are corrected in the open
A substantive correction to a published document is logged on the corrections page with the date, what was wrong and what changed. To report an error, see contact. Methods and sources are described on the methodology page; data credits are on the data attribution page.