Standards

Methodology

How material reaches this platform, who checks it, how often it changes, and what happens when it turns out to be wrong. This document is the standard our editors are held to, published in full because a claim about method that cannot be inspected is not worth much.

Version 4.2 In force since 1 April 2026 Next review 1 April 2027 Supersedes 4.1 (2025)

Part 1

Principles

Five commitments that constrain everything else in this document.

We publish only what is already public. Every figure originates in a document that anyone can obtain: a statistical agency release, a ministry or municipal report, a regulatory filing, or peer-reviewed research. We conduct no surveys and purchase no proprietary datasets.
The limits travel with the figure. A number without its definition, reference date and known exclusions is not information. Those elements appear on the same panel as the number, not in a document linked from it.
We describe, we do not advise. We do not recommend policies, rank jurisdictions against one another, or characterise a result as good or bad.
We do not forecast. Every series ends at the last observed period. We publish no projections, models or scenarios.
Disagreement is published, not resolved. Where credible sources conflict, both appear with an explanation of what each counts. We do not average them.

Part 2

Information sources

What we accept, what we decline, and how each source is recorded.

Accepted sources

National, provincial and territorial statistical agency publications and their reference tables.
Ministry, department and agency reporting where the underlying collection method is documented.
Municipal open data portals and statutory reporting, where a schema is published.
Regulatory filings and the public summaries built from them.
Peer-reviewed research, used for method and context rather than as a source of headline figures.

Declined sources

Proprietary or subscription datasets that a reader cannot inspect.
Material supplied on condition that it not be attributed or that its limits not be described.
Commercial research where the method is not published.
Anything supplied by an organisation with a direct financial interest in how the figure reads, unless the same figure is independently published elsewhere.

The source register

Every source is recorded with eleven fields: publisher, title, publication date, reference period, geography, collection method, coverage or response rate, revision status, licence, retrieval date and the editor who entered it. The register is what makes the provenance line under each chart possible, and it is the first thing an editor consults when a figure is challenged.

The source register by publisher type

Municipal sources are the second largest group and by far the most work: they are the least standardised and the most likely to change schema between releases.

Source Whab source registerUpdated 1 April 2026Cadence Continuous

Part 3

Editorial review

Who checks what, and at which point.

Stage 1

Entry

One editor records the source against all eleven register fields and drafts the plain-language definition. Nothing proceeds without a documented collection method.

Stage 2

Independent check

A second editor repeats the entry from the original document without seeing the first. Discrepancies are resolved against the source, not by discussion.

Stage 3

Presentation review

A third reader who did not touch the data reads the panel cold and states what they think it says. If that differs from what it does say, the panel is rewritten.

Stage 4

Sign-off

The managing editor signs off. Sign-off is recorded with a name and a date and is retained for seven years.

The cold-read test

Stage 3 is the stage most often skipped elsewhere and the one that catches the most errors here. A chart can be technically correct and still leave almost every reader with a false impression, usually because of an axis choice, a missing denominator, or a title that asserts more than the data supports. A reader who knows nothing about the panel is the only reliable detector of this.

Part 4

Update process

How often each series changes, and what happens on a refresh day.

Series follow the cadence of their source, not a schedule of ours. A monthly source produces a monthly indicator; an annual survey produces an annual one. Every panel states its cadence and the date of its last update, so a reader never has to infer how fresh a figure is.

On a refresh day an editor retrieves the new release, checks it against the recorded schema, and looks for four things: a change in definition, a change in geography, a revision to previously published periods, and a change in coverage. Any of the four stops the automatic refresh and sends the series back to Stage 1. In 2025 this happened eleven times across thirty-six indicators.

Indicators by update cadence

The three irregular indicators depend on survey cycles that are not on a fixed schedule. Their panels say when the last cycle completed and do not imply a next one.

Source Whab publication scheduleUpdated 1 April 2026Cadence Reviewed quarterly

Part 5

Quality standards

The tests a panel must pass before it publishes.

Definitional completeness: The panel states what the measure counts, what it excludes, and the reference period. A panel that cannot state its exclusions does not publish.
Denominator visibility: Every rate shows, or links to, the count it was built from. Rates on small denominators carry an explicit caution.
Comparability: Series that combine jurisdictions include only those reporting on a common basis for the full period. Where that shrinks the coverage, the panel says how much.
Axis honesty: Bar charts begin at zero. Line charts may not, but a truncated axis is labelled as such. No dual axes unless the two series share a unit.
Colour independence: No information is carried by colour alone. Every series is also distinguishable by label, order or position, and every graphic passes a greyscale check.
Plain language: Panel text is written to be understood on first reading by someone with no background in the subject. Technical terms are defined where they appear, not in a glossary.

Part 6

Accessibility standards

What we build to, and how it is tested.

We build to WCAG 2.2 Level AA and test against it. For the material on this platform the requirements that matter most are the ones about charts, because a chart is where accessibility is most often abandoned.

Every graphic carries a text description that states what it shows, not merely what type of chart it is.
Every graphic exposes its underlying values as a data table that can be opened and read without a mouse.
Every data point is reachable with the keyboard and announced with its series, category and value.
Text contrast meets 4.5:1 in both the default and dark presentations, and a high-contrast mode is available from every page.
Text can be enlarged to 200 per cent without loss of content or function; the platform reflows rather than requiring horizontal scrolling.
Motion is limited and respects the operating system reduced-motion setting.

Testing combines automated checks on every build with manual keyboard and screen-reader passes on each new panel type. Two external reviewers audit the platform annually. Findings, including open ones, are listed in the accessibility statement.

Part 7

Corrections and revisions

Two different things, handled differently.

Revisions

A revision is a change made by the original source. Statistical agencies revise routinely as late returns arrive and as seasonal adjustment models are re-estimated. When a source revises, we adopt the revised value, mark the affected panel, and record the previous value in the revision log. A revision is not an error and is not treated as one.

Corrections

A correction is a change made because we got something wrong: a transcription error, a wrong unit, a misapplied definition, or a description that misrepresented what the figure showed. Corrections are published with the date, the previous value, the corrected value, and a plain statement of what went wrong. We do not quietly edit pages.

Reporting an error

Write to info@whab.digital with the page address and what you believe is wrong. We acknowledge within two business days. If the report is confirmed, the correction publishes within five business days of confirmation. If it is not confirmed, we write back explaining why, and that exchange is retained.

Revisions and corrections by year

We publish this chart because an organisation that never reports a correction is either not looking or not telling. The useful signal is the direction of the corrections series, not its absence.

Source Whab revision logUpdated 1 April 2026Cadence Annual

Part 8

Independence and funding

Who pays, and what that buys them.

Whab is funded by institutional memberships from libraries, school boards, municipalities and universities; by two philanthropic foundations under multi-year unrestricted grants; and by fees for accessibility and data-presentation training delivered to public bodies.

No funder reviews material before publication. No funder may commission a report on a subject in which it has a direct interest. Funding sources above five per cent of annual revenue are named in the annual disclosure. We accept no advertising and no sponsored content, and we do not sell reader data.

Staff may not hold paid roles with organisations whose material appears on the platform. Where an editor has a prior association with a source, the panel records it and a different editor performs the independent check.