How the research is done

Research methodology

How QuitWithoutNotice decides what counts as a verified source, what a jurisdiction has to clear to get a page, and what happens when the evidence runs out. This is the public version of the internal document the research programme is held to — the same rules, written for a reader rather than a contributor.

This page describes the research process itself, not the law of any one country. Country-specific claims live on that country’s own page, cited individually.

General information only. Employment rules depend on your contract and location.

Published by QuitWithoutNotice

The narrow definition

One kind of research, chosen on purpose.

This project publishes exactly one kind of research: a record of what a named authority says, with the URL, the source type, and the date it was last fetched and checked against the claim. On a topic where the answer affects a real decision, the tempting kind of research — a percentage, a survey finding, an “average notice period worldwide” — is the kind that gets quoted forever and cannot be checked by the person quoting it. A dataset whose every cell traces to a government page that was actually read is smaller, less quotable, and correct.

Two rules follow, and neither has an exception. No number enters this dataset from a secondary source — not a competitor, not an HR vendor’s blog, not an aggregator, not an AI-generated answer. And missing evidence is never converted into a conclusion: a country nobody has read the law for is labelled as such. “Probably follows the regional pattern” is not a finding; it is a guess wearing a finding’s clothes.

The source hierarchy

A source’s tier decides whether a claim may be published from it at all.

Four tiers of citable source, and one category that is never citable regardless of how it ranks or how often it is linked elsewhere.

Tier 1 Legislation: The instrument itself — a section of a civil code or an act. Quotable directly, where the text is actually served rather than rendered client-side.
Tier 2 Government agency guidance: The authority’s own explanation of the law. Publishable, attributed to the agency in its own terms.
Tier 3 Courts: A judgment. Publishable with the case named. None used yet in this dataset.
Tier 4 Named professional bodies: Supporting only, never the sole support for a claim. None used yet in this dataset.
— Commercial blogs, HR-SaaS content, law-firm marketing, aggregators, AI answers: Not citable and not registrable — regardless of ranking or how often another site links it. Usable, at most, as a lead to a Tier 1–2 document that must then be read directly.

Where a Tier 1 instrument exists but cannot be read from this project’s environment, the practice is to link the instrument and attribute the claim to the Tier 2 guidance that states it in those terms — never to quote a statute nobody opened.

Verification process

The same seven steps, every cycle, per country.

Skipping a step is how a competitor’s aggregator page ends up quoted as if it were the law.

Identify the authority — the body that publishes the rule, not the body that ranks for it.
Open the authority’s own pages from this project’s environment, and record the result. A source is not verified because it exists in an earlier document, because another site links to it, or because a search result mentions it.
Read the text that supports the exact claim to be published — not the page’s summary, not its title.
Write the claim so it can be checked against that text, with any qualifier the source itself carries.
Record the date the page was opened, and any “last updated” date the page states about itself. Both matter, and they are different facts.
Log the request — including a failure to load, which is a recorded network fact rather than a silent gap.
Set the status from what was actually read, using the four values below, and set the next review date.

Statuses

What each status means, and whether it gets a page.

These are the same four values the product itself renders — a record below “verified” carrying a published page would fail the build.

Verified: Every published claim names the source it was read from, and the date. Page published.
Partial: Something is verified, but the rules that decide notice sit below the national level or in the contract. No page published.
Source review required: The authority is known; its guidance has not been read this cycle. No page published.
Not covered: No research at all — the state of most of the world. No page published.

Inclusion and exclusion

What earns a place in the dataset, and what does not.

A country is included when an authority has been identified and at least one of its pages has been opened — even if the result was a failure to load, and even if nothing publishable came out of it. The failures are the part a reader cannot get anywhere else.
A country is excluded when no research has been attempted. It resolves to “not covered” and is described as such in aggregate rather than listed.
A rule is included in a country’s row when it comes from a Tier 1–2 source read this cycle, and it concerns an employee resigning.
A rule is excluded when it is employer-side notice presented as if it were employee-side — the most common error in this topic — when it comes from a secondary source, when it depends on a document the project cannot see (an individual contract, a collective agreement), or when the source describes a different employment status than the one the row claims.

The same test applies past employment law. Strategic or marketing material sometimes carries figures — search-behaviour statistics, click-uplift claims, survey rankings — and those are second-hand numbers from an interested party, unverifiable against a primary source. None of them may appear on the site, and none may inform a factual claim.

Update frequency

A planned review date, shown openly.

Trigger	Action
Planned review	Six months after a source was last checked — a plan, not a record that a review has happened.
A source states a newer “last updated” date	Re-read that page and re-date the claim, out of cycle.
A country moves status	The index is re-issued with a new version number.
A source URL 404s, moves, or starts failing	Logged, and the claim is dropped if no replacement is read.

Version numbering is major.minor. A minor bump is new rows or re-dated rows; a major bump is a change to the methodology itself.

Limitations

Stated once here, and repeated everywhere the data feeds.

A small number of verified countries. Any characterisation of this dataset as global, comprehensive, or an index of the world would be false. It is an index of what has been read.
Statutory floors are not what most people owe. In every country reviewed, the operative number usually comes from the employment contract or a collective agreement — documents this project cannot read for a reader.
Employee-side only. Employer notice obligations are a different set of rules and appear only where a row says so explicitly.
Network reachability shapes coverage. A jurisdiction can be unverified because its authority’s pages do not answer requests from this project’s environment — a fact about this project, not about the source.
No sub-national completeness. Where a country’s answer lives below the national level, one recorded state or province is one recorded state or province, and nothing more.
No practice data. Nothing here describes what employers actually do, how often notice is waived, or what buyouts cost — that would need a survey this project has deliberately not built.
Not legal advice, in any row, on any page, for anyone.

The Global Notice Period Index is where this methodology is applied — coverage, sources and verification dates for every jurisdiction with a record. The changelog tracks what has actually changed since publication.