T3.04.2Claim-level single source of truthdesignresearch

Multi-channel content needs a single source of truth

Aliases: single source of truth · structured claim · claim registry · content syndication

What it is

A claim-level single source of truth gives verifiable facts—price, eligibility, capability limits, policy status, and procedural prerequisites—an authoritative record, then lets UI, help, notifications, support, and marketing present them for their own contexts. What is single is the fact or claim identity, value, conditions, evidence, and version, not one paragraph reused verbatim in every channel.

The same fact can be compressed beside a control, expanded in help, and reduced to the event-relevant portion in a notification. Those presentations are consistent when conditions and substance survive. Copying an identical sentence everywhere while hard-coding its number in each channel is still not a single source.

Why it happens

Channels impose different constraints on length, voice, order, interaction, and accessibility context, while their underlying facts change together. If each channel stores both fact and phrasing, every update must discover every copy. Separating claim from presentation lets the factual record hold structured values, applicability, and evidence; a channel template or editorial workflow consumes that record and expresses it for the local task. A change can enumerate consumers, and a channel difference can be identified as presentation or a governed override.

The source of truth need not be one CMS or database. Authorities may be federated by fact type: prices from billing, available capabilities from entitlement services, and policy status from governance records. The requirement is one declared authority for each claim, referenced through a stable identifier, version, and interface. Procedures may also need ordered, conditional nodes rather than atomization into isolated sentences that lose context.

Studying it

Sample tasks in which people compare channels, such as price, eligibility, feature limits, and procedural prerequisites. Normalize each expression to claim, value, conditions, effective interval, and source; identify contradictions, omitted qualifiers, stale consumers, and untraceable hard-coding. Do not use string identity as consistency: paraphrases may preserve facts, while identical wording may omit context required in one channel.

Run a change drill for one claim and trace consumer discovery, propagation latency, retry, approval, and rollback. Ask participants to resolve the same task using UI, help, and support answers; measure conclusion agreement, comprehension, and outcome. Hold channel-writing quality constant when comparing architectures. A multilingual audit should verify that locale presentations consume the same claim version rather than comparing translation surface alone.

Where it stops holding

Not all content can or should be generated. Narrative, examples, contextual explanation, and brand expression can be channel-authored, although falsifiable numbers, conditions, and product promises inside them still refer to authoritative claims. Text required verbatim by law or contract can be a version-controlled authoritative asset rather than an editable paraphrase. Experiments may vary framing, sequence, and examples, but not silently alter factual conditions.

Legitimate variants by region, role, plan, or time should be modeled as conditional claims or explicit overrides, not copied and allowed to fork invisibly. A terminology system governs what a concept is called; a claim source governs what is asserted about it. Lifecycle processes decide when consumers need review, while the claim architecture identifies the version they depend on.

Applying it

  • Define a claim schema with claim_id, subject, predicate, value, unit, qualifiers, audience or role, region, effective interval, evidence, authority, owner, version, status, and consumers. Do not bury applicability solely in free-text notes.
  • Keep presentation at the channel layer: language, length, voice, order, example, accessible label, and template version. Resolve claims at render or review time; a human-authored rendition still retains the claim reference and version instead of becoming an unsourced fact.
  • Specify fallback when an authority is unavailable: use a last-verified version with applicability time, block consequential publication, or omit the claim. Never guess silently. Record a channel override's reason, scope, approver, and expiry condition.
  • Use contract tests for values and qualifiers from source to consumer, and content QA for comprehensibility. On claim change, enumerate consumers and wait for critical acknowledgements before release. Monitor version forks, hard-coded facts, propagation failures, and user-reported contradictions.

Related

  • Same group: T3.04.1 Content strategy sets production, review, and retirement rules · T3.04.3 An ungoverned content library keeps entropying
  • Adjacent: T1.04.1 Use exactly one term per concept across the product · T3.03.2 Screenshots and version numbers rot first
  • Search terms: single source of truth · claim registry · content syndication

Cards in the same group

Quick Actions

Share

Share this page

ios_share

https://hci.top/en/handbook/T3.04.2