Review material. Document status: draft. This is exploratory research, not an authoritative service or released data product.

Product requirements for A Life in the UK

Document purpose

This document defines what A Life in the UK must do for its users. It organises the existing approved corpus, safety, semantic and publication contracts into one product view. It does not replace those contracts or authorise a new release.

The product baseline is main at 736d7dc4dbb4e44082f6b7786dd88afd55954792. The public population-complete preview and its retrieval records passed exact-byte and real-browser verification on 13 August 2026. The corpus is not release grade: 291 of 293 service families still require specialist review and none has named specialist acceptance.

When this document and a more specific governed contract disagree, the contract controls. The main controlling sources are:

Product summary

A Life in the UK is an independent, citizen-centred discovery and learning product about public services, rights, responsibilities and life events in the United Kingdom. It helps a person move from a topic or situation to:

  1. an understandable service-family overview;
  2. its place within a wider life-course domain and enclosing process;
  3. ordinary and exception journeys;
  4. explicit national, devolved and applicable local variants;
  5. the authorities, providers, evidence, rules, outcomes and redress involved;
  6. inspectable relationships and their provenance; and
  7. current official source links for further action.

The product is a map, not the territory. It supports discovery and education; it does not perform a government transaction or make a personal decision.

The problem

Information about one life event is often distributed across departments, devolved administrations, councils, health bodies, regulators and private dependencies. Publisher websites usually describe the part owned by one organisation. A person may still need to work out:

Search engines can find pages but do not reliably expose that structure or its evidence. Large graphs can expose structure but become difficult to interpret without hierarchy, grouping and a clear selected route.

Product vision

Anyone exploring a topic within the governed life-course vocabulary should be able to find the relevant family, understand its context, inspect its journey and evidence, and continue to an authoritative source without needing to know which public body owns the subject or how an ontology works.

Intended users and needs

People exploring a public-service situation

They need plain-language orientation and a safe handoff to the current official service. They may not know the service name, responsible organisation or jurisdiction.

Learners and educators

They need familiar examples that explain the difference between data, information, knowledge, taxonomy, ontology and provenance. They need to move between narrative and graph views without losing context.

Researchers, service designers and policy professionals

They need to compare routes, authorities, dependencies, exceptions and gaps across domains and jurisdictions. They need stable identities and evidence for material relationships.

Specialist reviewers

Legal, clinical, safeguarding and service-owner reviewers need to see the exact claim, source, observation date, jurisdiction, assertion status and review state before accepting or rejecting it.

Developers and AI-assisted research users

They need bounded, machine-readable records with stable identifiers, explicit ordering and provenance. They need the product to make unsupported answers and missing evidence detectable rather than rewarding a plausible guess.

Maintainers and publishers

They need authored source, deterministic generation, locked local validation and an exact-byte manual publication path. They must be able to update a claim without hand-editing a generated projection.

Product principles

  1. Start with the person's situation. Organisational structure is useful evidence, not the primary navigation model.
  2. Show context before complexity. Domain, enclosing process, selected family and selected episode should remain clear as a person opens detail.
  3. Keep variants explicit. Similar labels, a GOV.UK route or a shared provider never prove UK-wide equivalence.
  4. Keep ordinary and exception journeys separate. Failure, enforcement, complaint and appeal routes must not be hidden inside an idealised path.
  5. Make trust inspectable. Material claims and graph edges retain authority, evidence, rights, observation time, derivation and review status.
  6. Hand off before advising. Summaries orient the user and active links send them to the current authoritative service.
  7. Reveal uncertainty. Unknown, not applicable, not published by a source and awaiting review are different states.
  8. Do not collect personal data. Review situations and personas are synthetic. The public product does not need a user account or case details.
  9. Remain static and reproducible. The main experience must not depend on a hosted search service, analytics, a remote model or a runtime source fetch.
  10. Teach through use. A user should learn what an ontology contributes by following an everyday journey, not by reading definitions first.

Goals

G1. Reliable discovery

Find every governed family, domain, process and supporting concept through titles, aliases and plain-language topic terms. When there is no supported match, preserve useful domain browsing and do not invent a result.

G2. Process orientation

Explain where the selected family sits in an enclosing process and life-course domain. Show what is explicitly authored as preceding, current, following or exception behaviour without turning mere relatedness into sequence.

G3. Journey understanding

Present at least one ordinary path and one exception or failure path for every family, including applicable jurisdiction, provider, requirements, evidence, rules, channel, cost, time, output, outcome and redress states.

G4. Trustworthy source handoff

Give every applicable route variant at least one verified authoritative link or a sourced non-applicability decision. Make the source identity, observation date, rights basis and limitations visible before the user acts.

G5. Navigable semantic relationships

Provide a graph in which people can move from domain to process to family to episode, then inspect supporting actors, jurisdictions, evidence, rules, dependencies, outcomes, redress and sources.

G6. Accessible learning

Use the same corpus to teach ontology and public-service modelling to a beginner, while retaining sufficient evidence and precision for expert review.

G7. Safe reuse

Provide human-readable and machine-readable records that preserve identity, ordering, provenance, rights and review limits for downstream tools.

G8. Reproducible maintenance and publication

Build and check all projections locally through locked uv commands. Publish only a frozen, explicitly authorised candidate and verify its exact public identity and representative journeys in a real browser.

Non-goals

The product will not:

Governed product baseline

Measure Current governed baseline
Life-course domains 24
Enclosing processes 48
Service families 293
Ordinary and exception episodes 586
Ordered journey steps 881
Typed official-source links 879
Governed directed relationships 15,810
Competency questions 104
Families requiring specialist review 291
Families with named specialist acceptance 0

Population-complete means reconciled against these declared denominators. It does not mean that the real-world subject is exhausted.

Product surfaces

Start and learning route

The landing page and beginner learning plan must explain what the product is, what it contains, how to try it and what it does not claim.

Explore OKF

The self-contained review interface must provide in-browser search, explicit jurisdiction emphasis, ordinary and exception episodes, ordered steps, official-source links and provenance without runtime data calls.

OKF Explorer

The full Explorer must provide Reader, Narrative, Graph, Resources and other compatible views over the large-corpus descriptor. It must preserve deep links and selected context as the user moves between views.

Compact human and AI retrieval records

The catalogue and one complete HTML record per family must expose stable identity, description, jurisdiction routes, episode order, sources and review state in small, directly retrievable documents. The full projection remains the audit artefact.

Repository source and evidence

Authored Markdown and structured files under source/, ontology/, profiles/ and schemas/ remain inspectable. Generated data, manifests and validation reports must point back to their governed inputs.

Required product behaviour

Epic A: orientation and discovery

PRD-DISC-001 — First-use orientation

As a first-time visitor, I want to understand the product before searching so that I do not mistake it for an official or personalised service.

Acceptance criteria:

PRD-DISC-002 — Search by governed topic

As a person who knows a situation but not a service name, I want to search in ordinary words so that I can find the relevant family.

Acceptance criteria:

PRD-DISC-003 — Empty and unmatched searches

As a person whose words do not match the corpus, I want a useful recovery path so that a failed query is not a dead end.

Acceptance criteria:

PRD-DISC-004 — Browse without searching

As a learner or reviewer, I want to browse from domain to process to family so that I can understand the coverage model without guessing keywords.

Acceptance criteria:

Epic B: understand one family in context

PRD-FAM-001 — Plain-language family overview

As a user who selects a family, I want a concise explanation of the situation and interaction boundary so that I know whether I am in the right place.

Acceptance criteria:

PRD-FAM-002 — Enclosing process narrative

As a person trying to understand a process, I want the selected family placed within its wider context so that I can see what may come before and after.

Acceptance criteria:

PRD-FAM-003 — Ordinary and exception journeys

As a person reviewing a family, I want ordinary and exception paths separated so that failure, enforcement or redress is not hidden.

Acceptance criteria:

PRD-FAM-004 — Jurisdiction and local variation

As a user in a UK nation or locality, I want variants shown explicitly so that I do not follow a route that applies elsewhere.

Acceptance criteria:

Epic C: navigate the graph

PRD-GRAPH-001 — Hierarchical graph entry

As a user opening a graph, I want the hierarchy to remain visible so that I understand where I am.

Acceptance criteria:

PRD-GRAPH-002 — Supporting relationships

As a reviewer, I want to inspect the concepts supporting an episode so that I can understand its authority, requirements and outcomes.

Acceptance criteria:

PRD-GRAPH-003 — Dense graph handling

As a mouse, keyboard or touch user, I want dense relationships grouped and selectable so that labels and hit areas do not obscure each other.

Acceptance criteria:

Epic D: evidence, trust and official sources

PRD-SRC-001 — Source summary and handoff

As a user, I want to know what supports the overview and reach the official source so that I can verify current information before acting.

Acceptance criteria:

PRD-SRC-002 — Provenance and rights

As a reviewer, I want to inspect where a statement came from so that I can judge its authority and permitted use.

Acceptance criteria:

PRD-SRC-003 — Freshness and review state

As a person considering whether to rely on a route, I want its freshness and review state disclosed so that I know when to check further.

Acceptance criteria:

Epic E: learning and explanation

PRD-LEARN-001 — Beginner route

As a learner, I want a guided route from a familiar situation to an ontology so that I can understand the concept without prior semantic-web knowledge.

Acceptance criteria:

PRD-LEARN-002 — Narrative and graph together

As a learner, I want narrative and graph views to explain the same governed relationships so that I can move between prose and structure.

Acceptance criteria:

Epic F: machine and AI-assisted use

PRD-MACH-001 — Bounded machine retrieval

As a developer or AI-assisted researcher, I want small complete records and a full audit projection so that I can retrieve safely without losing evidence.

Acceptance criteria:

PRD-MACH-002 — Safe answer boundary

As a reviewer of an AI-assisted answer, I want material statements traceable to records and sources so that unsupported fluency is visible.

Acceptance criteria:

Epic G: inclusive and resilient use

PRD-USE-001 — Operable interface

As a user with different access needs or input methods, I want the interface to remain understandable and operable.

Acceptance criteria:

PRD-USE-002 — Static and managed-device resilience

As a user on a managed laptop, I want a low-dependency experience so that I can review the corpus within common network restrictions.

Acceptance criteria:

Epic H: maintenance and release governance

PRD-GOV-001 — Reproducible change

As a maintainer, I want one governed authoring path so that generated surfaces cannot drift from their evidence.

Acceptance criteria:

PRD-GOV-002 — Controlled publication

As the product owner, I want publication separated from implementation so that untested bytes cannot become public implicitly.

Acceptance criteria:

End-to-end acceptance journeys

The following visible journeys are required across the product surfaces.

Local-service failure

Search for a missed bin, find Report a missed rubbish collection, identify the local-authority responsibility boundary, compare ordinary and failure episodes, inspect complaint or redress relationships, and open the official postcode route.

Four-nation health discovery

Search for finding an NHS dentist, identify the family and its nation-specific routes, preserve the manual link-only health boundary, and open a current official source without generating clinical advice.

Enforcement exception

Search for a speeding notice, distinguish Great Britain and Northern Ireland routes, inspect the notice, deadline, evidence, penalty and court or challenge branches, and avoid making a legal decision for the user.

Death and bereavement

Search after a death, distinguish registration, Tell Us Once or manual notification, estate authority and administration, compare national variants, and show private dependencies without recommending a provider.

Education and evidence

Search for applying for a school place, recover the stable family identity, ordinary and exception steps, jurisdiction routes, sources and review state in both the human interface and compact retrieval record.

Unmatched question

Enter a conversational question outside the governed vocabulary. Receive no invented answer, retain the 24-domain browse route, revise the query and return to the previous context with browser Back.

Source unavailable

Open a family whose official page blocks inline browser access. Retain the family summary and evidence, explain why inline display is unavailable, and offer a labelled official-source handoff without claiming the source was verified in that session.

Dense hierarchy

Open a high-count format or domain group, reveal grouped children one level at a time, see which level is open, select one node unambiguously, then use Back to undo each opening action.

Product success measures

Discovery

Comprehension

Trust and handoff

Reliability and governance

Completion and release gates

Population-complete

All 293 governed families satisfy the discovery, process, journey, source and relationship contract. Specialist-review warnings may remain. This gate is met.

Product-experience complete

The current population is reliably discoverable and understandable through search, domain browsing, Narrative, Graph and Resources. The open search fallback and graph-interpretation requirements in this PRD must pass their visible acceptance journeys before this gate is met.

Release grade

Every applicable legal, clinical and high-impact operational claim has current source re-observation and named reviewer acceptance. This gate is not met.

Publication-ready

A candidate for the intended release class is frozen, explicitly authorised, deployed without rebuilding and verified at its exact public URLs. The current population-complete preview has passed this gate for its stated non-release- grade scope; any later change requires a new candidate and verification.

Current assessment

Product outcome Status at this baseline Evidence or remaining need
Governed population and process mapping Met 293 families, 24 domains and 48 processes reconcile locally.
Ordinary and exception journeys Met 586 authored episodes and 881 ordered steps are present.
Source-linked discovery Met for the frozen preview 879 typed links and representative search-to-source journeys passed.
Exact-title and governed-alias search Met Local search acceptance and representative browser queries passed.
Conversational-query recovery Partly met Several ordinary-language questions returned no result.
Unmatched-query domain browsing Not met in the full Explorer A zero-result query currently reduces all facets to zero.
Family narrative and provenance Met on representative journeys Narrative, graph edge evidence and source handoff passed.
Dense graph and facet interpretation Partly met Hierarchy, grouping, target size and control-density improvements remain.
Compact AI retrieval Met for deployed records Catalogue and 293 family records passed identity and browser verification.
Specialist-reviewed release grade Not met 291 families require review; none has named acceptance.

Dependencies

Risks and mitigations

Risk Product effect Required mitigation
Official pages change or disappear A route may become stale or unusable. Show observation dates, run reviewed metadata-only link audits and block a family when its primary link fails.
Similar services differ by nation or locality A user may follow an inapplicable route. Model variants explicitly and never infer applicability from labels or domains.
Graph density hides meaning A user cannot tell what is selected or how they arrived. Open grouped hierarchy progressively, preserve a visible route and make Back undo each opening.
Static search misses natural phrasing A user assumes the corpus has no relevant topic. Add bounded governed variants, explain no-match states and retain domain browsing without a generative answer.
A polished summary is mistaken for advice A user acts on unreviewed material. Keep independent, non-release-grade and official-source warnings visible at decision points.
Machine consumers invent sequence or authority Downstream answers appear supported when they are not. Preserve stable assertions, explicit order, authority planes and abstention-oriented evaluation.
Publication drifts from reviewed bytes Public evidence no longer proves the live site. Freeze manifests, deploy manually and verify exact URLs after each authorised change.
Specialist review does not scale to 293 families Release grade remains indefinitely incomplete. Prioritise high-impact claims, record reviewer roles and accept partial reviewed subsets without relabelling the whole corpus.

Priorities after this PRD

  1. Close the bounded discovery gaps: conversational-query handling and the 24-domain fallback for unmatched searches.
  2. Reduce graph and facet interpretation cost while preserving provenance, hierarchy and browser history.
  3. Run inclusive usability review across the representative national, devolved, local, health, legal and private-dependency journeys.
  4. Define and staff the specialist-review programme for legal, clinical and high-impact operational claims.
  5. Re-observe sources and freeze a new candidate only when an explicitly scoped release or publication change is requested.

Decisions still required for a release-grade product

These decisions do not block the current population-complete educational preview. They block only the relevant stronger product or release claims.