How we verify
If you want to know why to trust a page here over the other eight results, this is the page that answers it.
Last updated 26 August 2026
The source hierarchy
Sources are ranked. A claim is supported at the highest tier available, and a claim that matters is not supported below tier two without saying so on the page.
| Tier | What it is | Used for |
|---|---|---|
| 1 | Primary legal instruments — acts, ordinances, regulations, gazette notices, regulator directives and circulars, court judgments | Any statement of what the law requires |
| 2 | Official institutional publications — regulator statistics, ministry reports, budget and monetary policy documents, official procedural guidance | Figures, procedures, institutional roles |
| 3 | Multilateral and international institutional data — IMF, World Bank, ADB, UN bodies, FATF | Macro data, comparative context, external assessment |
| 4 | Audited disclosures and exchange filings | Company and market data |
| 5 | Peer-reviewed and established research | Analysis and methodology |
| 6 | Reputable professional and news sources — named practitioners at identifiable firms, established news mastheads | Finding primary sources; reporting practice |
| 7 | Editorial interpretation — our own reading | Always labelled as ours |
The tier-six restriction, and why it exists
Tier six is deliberately narrow. When we surveyed the sources available on Nepali business and regulatory law, a large cluster of law-firm-shaped domains returned near-identical, unsourced, apparently machine-generated text on almost any query — the same paragraphs, the same confident figures, no citations, no named author, no date.
So tier six admits only named practitioners at identifiable firms and established news mastheads. A domain that looks like a law firm is not evidence that it is one. And a secondary source is used to find a primary source, never to substitute for one on a legal, fee or procedural claim.
The record kept for every source
Every source cited anywhere in the corpus has a record in the source register. Each record carries:
- Source title, in the original language and in English
- Issuing institution
- Publication date and, separately, effective date — these are frequently not the same
- URL, plus an archived copy, because Nepali government sites reorganise often
- Date accessed
- The specific claim or claims it supports
- Its tier in the hierarchy
- Whether expert review is still required on the claims it supports
- Amendment status, and what superseded it if anything has
- Every page relying on it — the field that makes re-verification tractable
That last field is the mechanism that lets a small desk maintain a hundred-page corpus. When an instrument is amended, the register produces the exact list of pages to re-check, rather than requiring us to audit everything and hope.
When we cannot confirm something
The page says so. It does not fill the gap from memory, from a plausible-sounding secondary source, or from what was true last year. An unconfirmed claim appears with a verification required flag, and the flag names the specific claim rather than hedging the whole page.
Where two official sources disagree — and in Nepal they sometimes do, including on headline macro figures — we publish both, attribute each, and say that they differ. Picking one and presenting it as settled would be the easier page and the worse one.
Three dates, three meanings
Every guide carries three separate dates, because collapsing them into one is misleading:
- Last verified — the date someone last checked the claims against their sources, whether or not anything changed.
- Last substantively updated — the date the content last changed in a way that affects what a reader would do.
- Next review due — the scheduled date, set by the page's own update cycle.
A cosmetic edit never touches dateModified. Freshness signals that are not earned
are a form of misdirection, and in a corpus whose entire value is currency, they are the
fastest way to lose the reader we are writing for.