Editorial governance / Version 13 July 2026
NeobankDB is an independent database, comparison, review and transaction-intelligence platform. It is not a bank and does not hold customer funds. This policy explains how a source becomes a candidate, how a candidate becomes an approved fact, where editorial judgement may enter, and which failures stop or reverse publication.
Current public editorial pages use NeobankDB organisation-level authorship. This page does not invent named writers, fact-checkers, credentials, committees or team size. “Researcher”, “reviewer” and “publisher” below describe accountable workflow functions and records, not unsupported staff titles.
Non-negotiable red line
A crawler can retrieve a page. A model can propose a field or a draft. A provider can send a response. An affiliate programme can offer compensation. None of those events decides that a claim is true, complete, current or suitable for publication. Approval requires the exact subject and market, a valid evidence record, source ranking, conflict analysis and a recorded review decision.
If that chain fails, NeobankDB holds or rejects the claim. The interface may show unknown, stale, conflicting, empty, error or unavailable. It must not substitute a plausible fact, reuse a neighbouring market, hide an inconvenient primary source, report a false submission success or silently fall back to candidate data.
Accountable workflow
Passing one gate does not imply that the next has passed, and publication does not end review.
Identify brand, legal entity, product, plan, market and field. If the exact chain is unresolved, research stops before a global claim is drafted.
Prefer controlling regulator, legal, protection, pricing and help records. Discovery sources can open a lead but cannot close a material fact.
Write a candidate explanation from structured evidence without changing its value, scope, status or confidence. A readable sentence is not an approval event.
Validate fields, source relevance, observation date, competing evidence and the decision language. High-risk claims always require this gate.
Render through repository interfaces with visible source, date and state. Public UI never reads raw crawler output or ORM rows directly.
Freshness expiry, provider changes and corrections create new candidates. History remains visible while changed copy is corrected, qualified or withdrawn.
Authorship and review
The current policy names workflow functions because no verified public individual profile is yet part of the publication record.
The research function defines the question, exact market and subject chain; locates the strongest relevant sources; records observation dates and provenance; and separates unresolved values rather than merging them. It may propose a verification state and confidence treatment, but proposal is not approval.
The review function checks that the source actually supports the rendered claim, that market language is exact, that high-risk entity, licence, protection, fee, availability and route fields have not been inferred, and that credible conflicts remain visible. The decision is approve, reject or hold—not rewrite evidence to fit a preferred conclusion.
The publication function confirms that route, metadata, structured data, source display, responsive states, tests and memory records satisfy the page definition of done. It does not bypass review. When required data cannot render safely, the publishing decision is to fail closed rather than make the page look complete.
Current attribution rule: articles and policy pages are credited to the NeobankDB organisation. Named individual attribution requires a real, reviewable profile and accurate role record. No biography, credential, experience claim, portrait, review assignment or employment relationship may be generated to make authorship appear more personal.
Evidence and rendered copy
The distinction keeps factual values auditable while allowing precise explanation.
A material record carries subject, field, market, product or plan where relevant, normalised and display value, source identity or URL, authority tier, observed and effective dates where known, verification state, confidence and reviewer status. Evidence from one market is not silently promoted to a global provider record.
Regulator registers and legal agreements lead the authority ladder for the questions they control. A high-authority page still has to be relevant to the exact field. A scheme page can establish that a scheme exists without proving that a named provider or product participates. A payment-rail page can establish infrastructure without proving a named transaction route.
Editorial copy may explain what the record means, why a condition matters and which trade-off follows for the stated use case. It must retain the subject, market, conditions and uncertainty of the evidence. It cannot transform “may be eligible” into “is covered,” a provider programme into NeobankDB’s own product, or “unknown” into a negative assertion.
Quotations are used sparingly and attributed. Paraphrase must not change legal meaning. Private reviewer notes, raw crawl payloads, credentials and operational prompts stay out of public copy. Public pages expose enough source and date context to audit the claim, not internal secrets or sensitive user data.
Reviews, comparisons and rankings
NeobankDB may make an editorial recommendation for a stated market and job, but it does not force every database collection into a ranking. A comparison can say which option better fits a defined use case only when the factors and limitations are visible. “Best,” “safest,” “cheapest” and “most popular” require their own evidence and cannot be inferred from brand recognition, a commercial relationship or the number of available source records.
A single number cannot conceal missing legal, protection, fee, route, reliability or support evidence. If components are incomplete, NeobankDB omits the score and publishes supported qualitative findings.
Scores and verdicts are market- and use-case-specific. A US business decision cannot become a global personal-account rank, and a transfer route cannot prove a product is the best place to hold funds.
Any published score must expose its components, weights, calculation method, evidence limits and date. Confidence in an evidence record is not a provider rating or a probability of customer success.
NeobankDB does not claim hands-on testing of application success, app quality, support, reliability or transaction completion unless dated test conditions are recorded. Unsupported test claims never receive substitute points.
Commercial independence
Commercial records and editorial records remain separate so the same facts and conclusions survive without an affiliate relationship.
A provider cannot purchase a legal classification, licence statement, protection claim, eligibility answer, score component, comparison winner, route verdict, correction outcome or source priority. Commission amount and conversion likelihood are not organic ranking inputs. Advertising and sponsorship, if introduced, must be visibly labelled and cannot be styled as an evidence-backed editorial conclusion.
An affiliate call to action may appear only when its destination and offer status are verified. The link is disclosed. If no verified relationship exists, the page may use a clearly identified non-affiliate official destination or omit the commercial action. NeobankDB does not invent discounts, bonuses, eligibility, approval likelihood or tracking URLs to fill an empty offer record.
Crawler and AI use
Crawlers and models are useful for discovering official pages, extracting candidate fields, comparing a new snapshot with an active record, flagging conflicts and helping draft a scoped explanation. Their output stays private until it passes the same evidence and review gates as manually entered research.
A generated summary is not a source. Review retains the URL, snapshot or retrieval record, observation date, extraction provenance and proposed transformation so the candidate can be reproduced or rejected.
A candidate can fit the expected TypeScript and runtime schema while naming the wrong entity, market, product or effective date. Source relevance, conflicts and high-risk meaning still require review.
Raw extraction is never queried directly by the public website. No crawler or model can promote its own result, mutate an approved fact, accept a correction, award a score or publish a page.
If the approved repository, correction store or required service is unavailable, the interface says so. It does not use cached candidate output, fabricate a receipt or turn a system error into a claim that a provider feature is unavailable.
Public control register
alwaysbaseline
Describe NeobankDB only as an independent database, comparison, review and transaction-intelligence platform.
The publication researches financial products but is not itself a bank, lender, broker, exchange, wallet or adviser.
Reject copy that crosses the product boundary; do not soften it into a disclaimer around a misleading claim.
Governing rules: Product boundary · Master blueprint · Reviewed 2026-07-13researchmaterial fact
Rank regulator and official primary records above discovery sources, then test whether the source supports the exact scoped claim.
An authoritative source can be irrelevant to a provider feature, while a provider page may support only a narrow product condition.
Hold the claim when authority or relevance is insufficient; preserve the gap explicitly.
Governing rules: Master blueprint source priority · Financial data rules · Reviewed 2026-07-13researchhigh risk
Crawler, extractor and model output may discover, summarise or propose candidate records but cannot approve or publish them.
Well-formed automated output can still join the wrong entity, market, date, value or source.
Reject automated publication and restore the last approved view or an unavailable state.
Governing rules: Publication model · Shared agent rules · Reviewed 2026-07-13researchmaterial fact
Keep the source URL, retrieval metadata, snapshot, extraction method and proposed field separate from approved public copy.
Reviewers need the original evidence and transformation trail to reproduce or reject a candidate.
Hold candidates whose evidence trail cannot be reproduced.
Governing rules: Master blueprint ingestion flow · Source and copy separation · Reviewed 2026-07-13researchmaterial fact
Research and write through Brand to Legal Entity to Product to Plan to Country to Fact to Evidence to Verification Event.
Brand-level wording must not overwrite a market-specific entity, plan, protection or eligibility record.
Hold publication when the exact entity or product chain cannot be resolved.
Governing rules: Financial data rules · Financial data implementation rule · Reviewed 2026-07-13researchhigh risk
Collect only the public evidence and limited contact details needed for editorial review; never request account credentials, card data, private keys or full transaction traces.
Editorial verification does not require access to a reader's financial account or sensitive authentication material.
Do not ingest sensitive material; direct account, fraud, legal and emergency issues to the appropriate institution or authority.
Governing rules: Privacy and security boundary · Correction policy · Reviewed 2026-07-13draftingbaseline
Use researcher, reviewer and publisher to describe accountable workflow functions, not unsupported job titles or claims about team size.
The policy can explain who must perform a control without fabricating an organisation chart.
Rewrite the statement around the required control and recorded decision.
Governing rules: Editorial policy · Publication model · Reviewed 2026-07-13draftingmaterial fact
Frame comparisons by exact market, use case and evidence rather than forcing one global winner or generic best-provider label.
Entity, eligibility, fee, feature, route and protection facts change by market and customer job.
Replace the ranking with a conditional decision framework or omit it.
Governing rules: Review model · Non-negotiable data rules · Reviewed 2026-07-13Corrections, conflicts and freshness
Change signals reopen review without letting an unverified replacement overwrite the last approved record.
A reader or provider can submit the exact page, market, claim, proposed correction, public HTTPS sources and observation date. Valid intake records an unapproved candidate. It does not promise acceptance or publication. A provider’s identity can establish who replied, but not whether the proposed value is correct; controlling regulator, legal or scheme evidence still outranks a promotional assertion.
Account support, fraud, legal notices, transaction disputes and emergencies do not belong in editorial intake. NeobankDB cannot access or fix an account. Those matters should go to the provider or appropriate authority without sending passwords, one-time codes, card details, account credentials or private keys to this publication.
When a freshness window expires, a source changes, or credible evidence conflicts with published copy, the affected fact returns to review. The prior approved value and date remain part of the change trail. Newer evidence does not automatically win; reviewers compare authority, scope and effective date before approving a replacement.
If a material claim is known to be unsafe and a correct scoped replacement cannot be approved, NeobankDB qualifies, holds, marks unavailable or withdraws that claim. A page should never preserve misleading certainty merely to avoid an empty section or protect a commercial conversion path.
Privacy, safety and limitations
Data minimisation protects readers and keeps the publication inside its product boundary.
Useful editorial evidence includes public regulator records, legal agreements, protection scheme materials, fee schedules, help articles, product pages and reproducible public route or product documentation. A limited contact email may be used for correction follow-up. Submitted material is still validated, scoped and reviewed before it can affect the publication.
Do not send account numbers, passwords, one-time codes, card data, identity documents, full transaction histories, wallet recovery phrases or private keys. NeobankDB does not connect to bank accounts, execute transfers, hold customer balances, approve applications, investigate fraud or provide personalised financial, legal or tax advice.
Audit the policy in context
The editorial policy defines the control. The methodology explains the research model, the source directory exposes approved source records, the evidence guide defines decision states, the conflict ledger preserves active disagreements, the correction policy accepts scoped challenges, and the change ledger records published page checkpoints. These public views do not expose raw crawls, candidate facts, private reviewer notes or credentials.