Editorial standard · version 1.0
Research methodology, evidence labels, and corrections
This is the operating policy behind Akoum.me authority pages, datasets, diagrams, and evidence-led articles. It is designed to make the boundary between source, measurement, inference, scenario, and opinion visible to the reader.
Published and last reviewed
Selection rule
Source hierarchy
A source tier describes what a source can establish. A company page is primary evidence of what the company announced; it is not independent evidence that a benchmark generalizes or a future promise will be fulfilled.
- Tier 1Primary and official
Official documentation, model cards, laws, regulators, court records, official statistics, and first-party policy or product announcements.
Use: Required for volatile product facts, prices, dates, legal claims, and an organization’s own definition or commitment.
- Tier 2Peer-reviewed or transparent data
Peer-reviewed research, preprints with inspectable methods, and datasets with provenance, fields, dates, and limitations.
Use: Preferred for measured findings and comparisons.
- Tier 3Independent evaluation
Independent evaluators, research institutions, and policy analysis with visible methodology and conflicts.
Use: Used to test, contextualize, or challenge first-party claims.
- Tier 4Reputable reporting
Professional journalism with named sources, documents, and a corrections process.
Use: Used for context and discovery; replaced by the underlying primary source when available.
- Tier 5Unverified monitoring
Social posts, rumors, screenshots, anonymous claims, and unattributed summaries.
Use: May create a research lead. It cannot establish a published factual claim.
Evidence labels
Every material claim on an authority page receives one of these labels. The label describes the claim, not the prestige of the organization being cited.
- Verified product fact
- A dated fact verified on an official product, documentation, pricing, release-note, legal, or regulatory surface.
- Measured finding
- A result produced by an identified method, dataset, task distribution, unit, and denominator.
- Attributed definition
- A named person or institution’s stated definition, commitment, goal, or framing. Attribution is part of the claim.
- Estimate or projection
- A numerical or directional forecast with an identified author, assumptions, horizon, and uncertainty.
- Inference
- A reasoned conclusion drawn from cited observations. It is not directly stated by the source.
- Scenario
- A possible pathway or outcome used for analysis without an assigned certainty.
- Opinion or proposal
- The author’s normative judgment, design choice, framework, or recommendation.
- Unknown or disputed
- Evidence is missing, contradictory, definition-dependent, or insufficient to support a confident conclusion.
Review, dates, and freshness
- 01Record the canonical source URL, publisher, publication date when stated, access date, tier, and evidence class.
- 02Write the smallest claim the source actually supports. Preserve attribution when the source is a company, laboratory, policymaker, or commentator describing its own position.
- 03Record metric, unit, denominator, geography, method, uncertainty, and counterevidence when a measured claim is involved.
- 04Set a review cadence according to volatility. Product, model, price, regulation, and hiring claims are reviewed at least every 30 days; evergreen definitions can use a longer documented cadence.
- 05Keep original publication, revision, review, and next-review dates distinct. A rewrite never receives a fictional new origin date.
- 06Retain a dated changelog when a correction, definition change, source replacement, or material conclusion changes.
Corrections policy
Factual errors are corrected in place. A material correction adds a dated changelog note describing what changed and why. Older wording is not silently preserved when it could continue to mislead search engines or readers.
New evidence can change a status from supported to contested or unresolved. Conflicting high-quality sources are shown as counterevidence rather than averaged into false certainty.
Corrections should identify the exact page, claim, and stronger source. Contact Muhamad on LinkedIn (opens in a new tab).
AI-assistance disclosure
AI tools may assist with research discovery, extraction, drafting, code, data normalization, diagram production, and consistency checks. AI output is not treated as a source.
Published factual claims must point to inspectable external evidence. The site owner remains accountable for editorial choices, framing, corrections, and the decision to publish. Private reasoning, credentials, personal data, and unverified model output do not enter the public evidence ledger.
Reuse, downloads, and citation
Research assets provide visible HTML and, where applicable, CSV, JSON, SVG, and PNG. Cite the named asset, version, reviewed date, author, and canonical URL. Always follow the underlying source when you need to verify an external claim.