Evidence before action
Reporting methodology
Every CTD story begins with a traceable source record. We preserve source context, test whether observations are independent, connect material claims to evidence, disclose uncertainty, and complete required editorial review before publication.
Method at a glance
From a possible signal to accountable reporting
-
01
Discover signals
Find possible developments and preserve their source context.
-
02
Corroborate and verify
Test source independence, evidence quality, and claim support.
-
03
Review and publish
Apply editorial checks, disclose limits, and approve a version.
-
04
Monitor, correct, and learn
Track later evidence without allowing performance to rewrite facts.
Discover
Discover signals
Discovery starts a reporting file. It does not establish truth.
CTD monitors sources it is authorized and technically able to use for meaningful changes in consumer costs, behavior, services, demand, and markets. Coverage may include official records and statistics, company disclosures, research, structured feeds, public web pages, and separately governed consumer submissions.
Source context is preserved
For a captured observation, CTD records the source, canonical URL, publication or observation time, capture time, relevant context, and a cryptographic integrity record. Where permitted, the source material itself is retained under its approved use and retention rules.
Access and authority are evaluated
A source profile records authority, ownership, independence, permitted uses, quotation rights, freshness, and other governance limits. Appearing in the system does not make a source’s claims true or authorize every possible reuse.
Verify
Corroborate and verify
Repetition is separated from independent confirmation
CTD groups related observations, compares canonical URLs and semantic similarity, evaluates publisher ownership, and records likely syndication. Multiple websites repeating the same press release, dataset, or original report are not treated as multiple independent confirmations.
Evidence review asks
- Can the original and most authoritative available source be identified?
- Are the observations genuinely independent in origin and ownership?
- Are dates, geography, population, denominator, and comparison period appropriate?
- Is the claim observed, calculated, inferred, predicted, disputed, or alleged?
- Does meaningful contradictory evidence remain unresolved?
- Does the language express more certainty than the evidence supports?
Claim support is explicit
Candidate claims are linked to validated evidence records. The required support increases with risk, and unresolved contradiction blocks approval. High-risk claims require an accountable human actor; automated approval is rejected.
Approved claims and their evidence are assembled into a versioned, integrity-checked reporting package before downstream publication work can begin.
Publish
Review and publish
Publication is an approval decision, not a generation step
The article and supporting publication materials are prepared from the released reporting package. Material claims must remain connected to supporting evidence, and the headline, article, charts, and summaries must not express greater certainty than the evidence allows.
Facts and interpretation stay distinct
Direct evidence, calculations, analysis, inference, allegations, and predictions should be identified in language appropriate to each. Relevant uncertainty, caveats, dates, and geographic scope remain visible.
Risk determines the workflow
Story classification assigns the required research, data, editorial, or senior-editor checks. Monitoring signals and unresolved duplicates are not publishable story classes.
Only the approved version publishes
Public release requires a released reporting package, approved content and experience versions, and the recorded publication approval. The published version is immutable and later versions supersede rather than rewrite it.
Monitor
Monitor, correct, and learn
What happens after publication is governed too
CTD may monitor new evidence, later developments, reader correction reports, source changes, engagement, readability, distribution performance, and whether readers reach the intended information. These signals can improve presentation, accessibility, distribution, and future coverage.
Measurement may inform
- Follow-up reporting and freshness reviews
- Presentation and accessibility improvements
- Distribution choices and future research questions
- Governed experiments that meet minimum sample requirements
Measurement may not
- Rewrite historical evidence or make an unsupported claim true
- Suppress a valid correction or contradictory evidence
- Let traffic or conversions control factual conclusions
- Turn noise into a trend or bypass evidence and sample requirements
Material errors, later updates, source substitutions, and withdrawals are handled through the Corrections Policy, with retained versions and downstream impacts where applicable.
What qualifies as a consumer trend
The word “trend” carries an evidence burden
CTD uses increasingly strong language as evidence develops. Its current r21 candidate rule looks for at least three independent signals observed over at least 30 days, together with evidence-strength and trend-strength thresholds. Those thresholds create a floor; editorial review still determines the supportable scope and wording.
A documented occurrence
An event, value, complaint, behavior, or condition tied to a source, place, and time.
A possible developing change
One or more observations suggesting a development that may need further monitoring.
A direction is taking shape
Several sufficiently related and independent observations align, but scope or duration remains limited.
A supported directional change
A sustained or meaningfully repeated change supported by evidence appropriate to its claimed scope.
Factors considered
Independence, number and quality of observations, time span, geography, represented population, magnitude, direction, sample size, denominator, seasonality, contradictory evidence, and alternative explanations.
What is not enough by itself
One viral post, one consumer report, copied articles, a company promotion, a brief search spike, one local event, or a large percentage change from a very small baseline.
Automation and artificial intelligence
Automation can assist the process. CTD remains responsible for the result.
May assist
Bounded newsroom work
Source monitoring, capture, organization, duplicate detection, classification, similarity analysis, prioritization, claim-to-evidence checks, draft preparation, formatting, quality control, and monitoring for later changes.
May not
Non-negotiable limits
Invent facts, quotations, sources, or evidence; conceal limitations; turn inference into fact; treat copies as independent confirmation; override required approval; remove material correction history; or let commercial value control conclusions.
Limitations and uncertainty
A governed method reduces avoidable error. It cannot eliminate uncertainty.
CTD describes limitations when they materially affect how a reader should interpret a finding.
CTD does not monitor every possible source, and source availability or permissions can change.
Public data may contain errors, omissions, revisions, or publication delays.
Early signals can weaken, reverse, or prove narrower than first observed.
Local or online behavior may not represent a national or full consumer population.
Correlation and timing alone do not establish causation.
Privacy, licensing, legal, security, or source-protection duties may limit what evidence can be shown publicly.