Methodology & Ratings
How we choose what to check, how we verify it, how we rate it, how AI is used and checked, and the standards we hold ourselves to. This page governs every fact check and every figure we publish.
What has actually been checked?
Publication and assurance are separate. A published page is not automatically independently reviewed. A check only counts when its evidence, date, reviewer role and version are recorded.
- Source checked
- The stated figure or finding was compared with its cited source.
- Calculation checked
- The arithmetic was reproduced from the recorded inputs.
- Independently reviewed
- A reviewer separate from the original analyst examined the analysis and its limits.
- Approved for publication
- The responsible approver recorded a publication decision for this version.
The site uses documented agent review. We do not claim independent human assurance or AQuA certification. Existing records are being brought into this register; an absent entry means no check is recorded here, not that the underlying work never happened.
Recorded checks and their limits
- What does existing research establish about benefit-fraud media coverage?
Historical context alongside the financial comparison; not a current media-coverage ranking.
Review priority: high. A political comparison could imply an unsupported finding about media priorities.
- Source checked: 2026-10-11 · Coding-agent source review
- Calculation checked: not recorded for this version
- Independently reviewed: not recorded for this version
- Approved for publication: not recorded for this version
- Coverage study concerns 2008–2017; survey responses concern 2016.
- No current like-for-like tax-gap comparison or audience-reach count was established.
Version: 2026-10-11-citizen
How we decide what review is needed
Every output needs a clear question, sources, scope and a check that it answers the question. Reproduced statistics need checks of units, dates and population. Calculators, factual verdicts and policy recommendations need more: assumptions, reproducible calculations, tests of alternative assumptions and a separate review proportionate to their consequences. Approval and unresolved limitations must be recorded. Material changes require a new version and fresh checks.
Household spending: what “essentials” includes
This page combines food and non-alcoholic drinks (category 1) with housing (net), fuel and power (category 4). It divides their weighted mean spending by weighted mean total expenditure, including other expenditure (category 13). This is not the mean of individual household shares and is not a full minimum budget.
Mortgage payments, Council Tax and Northern Ireland rates are not in category 4. The total includes mortgage interest and Council Tax but not mortgage capital. Transport, childcare and other necessary costs are outside this limited numerator. Amounts are nominal pounds for the stated year; no CPIH or other deflator is applied. Income grouping is equivalised; expenditure remains per household.
Published rounded sample sizes accompany each group. No respondent records have been used. Small differences may be sampling noise; no significance test is claimed. A history back to 1968 is withheld until categories, weighting and survey breaks can be reconciled.
Source tables and calculationsAnalytical assurance
How far can these assessments take us?
These are provisional appraisals for public scrutiny. They help identify promising options and the work needed before a decision. They are not completed business cases.
- all ten criteria scored
- 0 / 51
- whole-life costs recorded
- 0 / 51
- scope verified in the record
- 0 / 51
The 8 October analytical follow-up specifies a proposed intervention, outcome test, costing approach and safeguards for 51 of these 51 candidates. Open “What the evidence supports” on an option to read it. These proposals do not replace the attributed scores or establish a funded business case.
Independent analytical assurance and final sign-off are not yet recorded. Automated checks verify calculations, references and consistent rendering; they cannot establish that an intervention will work or replace independent judgement.
Our assurance framework is based on the lifecycle in the 2025 AQuA Book: define the intended use and risks, document methods and assumptions, verify implementation, validate fitness for purpose, communicate uncertainty and obtain independent assurance and approval. Full compliance is not claimed.
1 · Question and scope
Specify the population, place, baseline and change being assessed. Keep a problem’s scale separate from the effect of a proposed intervention.
2 · Evidence and uncertainty
Trace claims to primary sources and reproducible calculations. Keep observations, forecasts, illustrations and causal evaluations distinct. Record breaks, uncertainty and alternative explanations.
3 · Options and delivery
Compare the increment beyond funded activity. Document costs, harms, distribution, capacity, powers, safeguards and a measurable outcome. Test how judgements change under different weights.
4 · Assurance and review
Record analyst checks separately from independent assurance. An approver must examine the evidence and residual risks before sign-off; changes require a versioned review.
Ten criteria: what each score means
Each criterion is an editorial judgement from 0 to 3, with a reason. Unknown is not zero. An incomplete record has completion bounds, not a total or rank; these bounds are not confidence intervals. Funding and delivery gates remain separate from the score.
| Criterion | 0 | 1 | 2 | 3 |
|---|---|---|---|---|
| Scale and severity | No material problem | Limited harm/narrow reach | Serious harm to a defined group or widespread moderate harm | Severe harm across a broad population |
| Increment beyond baseline | No increment | Small or poorly supported increment | Specific material change to existing delivery | Substantial clearly distinct increment; not novelty alone |
| Confidence in the proposed action | No effect evaluation | Indirect, mixed, modelled or weak causal support | Relevant evaluated evidence with limitations | Consistent strong evidence for the specified intervention/outcome |
| Expected incremental benefit | Negligible | Small | Material for the target population | Large for the target population; no multi-domain bonus |
| Residual harms | Unacceptable serious harm | Material harm/unresolved mitigation | Manageable harm with specified safeguards | Low assessed residual harm |
| Incremental affordability | No credible route for a large commitment | Major unfunded commitment | Costed affordable route subject to allocation | Demonstrably small affordable incremental public cost |
| Delivery capacity | Unavailable within scope | Binding scarce resource | Stretched but plausibly obtainable | Little scarce capacity required |
| Distribution | Strongly regressive | Material disadvantage to vulnerable groups | Mixed or broadly neutral with safeguards | Benefits concentrated on disadvantaged people with assessed burdens |
| Meaningful first outcome | Ten years or more | More than three but less than ten years | More than one and up to three years | Within one year |
| Readiness | Undefined enabling steps | Legislation/external agreement or major enabling work | Named commissioning/allocation/coordination step | Existing owner and route can act operationally |
Assessment register: missing evidence and next work
Read from the current shared records. The reviewer and date describe that assessment, not independent assurance of it. A gap must be resolved with evidence or a specified delivery decision, not filled with a guessed score.
Health-linked work support
Original assessment author (name not supplied) · 2026-10-06
Unscored: Incremental affordability.
Costing work: Cost integration, staff caseloads, clinical time, data and follow-up against commissioned provision.
Delivery work: Specify cohort and voluntary offer; secure clinical referrals, trained employment specialists, employer links and linked-data evaluation.
Evidence, scores and monitoring →Skills in the building trades
Original assessment author (name not supplied) · 2026-10-06
Unscored: Expected incremental benefit; Incremental affordability.
Costing work: Additional training, trainee support, employer supervision and trainer recruitment; levy eligibility is not an allocation.
Delivery work: Identify trades and locations; secure paid placements, trainers, facilities and employer demand; measure completion and retention.
Evidence, scores and monitoring →Steady public investment through a published pipeline
Original assessment author (name not supplied) · 2026-10-06
Unscored: Expected incremental benefit; Incremental affordability.
Costing work: Project lifetime costs, optimism bias, maintenance, financing and the coordination team. The headline strategy is an overlapping envelope.
Delivery work: Distinguish funded projects from aspirations; sequence resources; name owners and independent assurance; protect maintenance.
Evidence, scores and monitoring →Deliver what is already funded
Original assessment author (name not supplied) · 2026-10-06
Unscored: Expected incremental benefit; Incremental affordability.
Costing work: Commercial staff, data systems, legal work and transition costs; separately verify cashable savings.
Delivery work: Name contracts, renewal dates and weaknesses; recruit commercial expertise; publish baselines, performance and remedies.
Evidence, scores and monitoring →Homes built, not just permitted
Original assessment author (name not supplied) · 2026-10-06
Unscored: Expected incremental benefit; Incremental affordability.
Costing work: Site-level land, infrastructure, finance, construction, affordable-housing subsidy and maintenance.
Delivery work: Select sites; resolve land, planning, utilities, transport and viability; commission tenure mix and delivery agreements.
Evidence, scores and monitoring →Grid connections and industrial power
Original assessment author (name not supplied) · 2026-10-06
Unscored: Expected incremental benefit; Incremental affordability.
Costing work: Capital and operating network costs, financing, bill incidence, relief duration and funding.
Delivery work: Separate queue management, network reinforcement and industrial levies; identify projects, kit, engineers and regulator decisions.
Evidence, scores and monitoring →Turn research into products and adoption
Original assessment author (name not supplied) · 2026-10-06
Unscored: Increment beyond baseline; Expected incremental benefit; Incremental affordability.
Costing work: Administration, grants, procurement risk and private match; estimate displacement and additional private investment.
Delivery work: Separate spinout terms, procurement trials, grants and firm adoption; define market failure, beneficiaries and comparator.
Evidence, scores and monitoring →Cut EU trade friction with a scoped food-and-farm deal
Original assessment author (name not supplied) · 2026-10-06
Unscored: Expected incremental benefit; Incremental affordability.
Costing work: Public implementation, business compliance, adjustment and distributional costs under alternative agreements.
Delivery work: Define covered goods, alignment, oversight and border changes; secure agreement and business transition.
Evidence, scores and monitoring →City-region growth powers
Original assessment author (name not supplied) · 2026-10-06
Unscored: Increment beyond baseline; Expected incremental benefit; Incremental affordability.
Costing work: Administrative transition, funded responsibilities, investment and evaluation; distinguish transfers from new resources.
Delivery work: Specify powers, stable budgets and accountability; integrate transport, housing and skills; assess local capacity.
Evidence, scores and monitoring →Pension money into UK assets
Original assessment author (name not supplied) · 2026-10-06
Unscored: Expected incremental benefit; Incremental affordability.
Costing work: Administration, fees, guarantees if any and savers’ risk-adjusted opportunity cost.
Delivery work: Define a change beyond existing commitments; test investable demand, fiduciary safeguards, liquidity and net additional capital.
Evidence, scores and monitoring →Alcohol minimum unit price in England
Original assessment author (name not supplied) · 2026-10-06
Unscored: Incremental affordability.
Costing work: Enforcement, treatment support and evaluation plus household burdens and retailer effects; low fiscal cost is not full costing.
Delivery work: Choose price/indexation; consult; legislate; arrange enforcement and support for dependent drinkers; predefine evaluation.
Evidence, scores and monitoring →Older people safe at home
Original assessment author (name not supplied) · 2026-10-06
Unscored: Expected incremental benefit; Incremental affordability.
Costing work: Assessment, adaptations, recurring care, follow-up and displaced workforce; do not assume fewer falls equal cash savings.
Delivery work: Separate budgets and cohorts; arrange risk assessment, therapists, installers and care staff; compare reablement outcomes.
Evidence, scores and monitoring →Controls on what feeds amplify
Original assessment author (name not supplied) · 2026-10-06
Unscored: Expected incremental benefit; Incremental affordability.
Costing work: Platform engineering, independent research, auditing, enforcement and compliance; a cheap pilot is not a costed regulatory system.
Delivery work: Specify ranking intervention and trial agreement; independent researcher access, consent/privacy safeguards and meaningful outcomes; separate later regulation.
Evidence, scores and monitoring →Family hubs built on Sure Start features
Original assessment author (name not supplied) · 2026-10-06
Unscored: Incremental affordability.
Costing work: Premises, staffing, outreach, recurring services and evaluation beyond existing allocations.
Delivery work: Specify service intensity and target reach; staff health/family services; measure uptake, fidelity and long-term outcomes.
Evidence, scores and monitoring →Safer streets, fewer scams
Original assessment author (name not supplied) · 2026-10-06
Unscored: Expected incremental benefit; Incremental affordability.
Costing work: Data integration, operations, investigation, evaluation and victim support.
Delivery work: Choose scam pathway, bank/platform partners, lawful data sharing, comparator and compensation safeguards; separate policing action.
Evidence, scores and monitoring →School mental health
Original assessment author (name not supplied) · 2026-10-06
Unscored: Expected incremental benefit; Incremental affordability.
Costing work: Practitioner time, supervision, specialist referrals, school time and follow-up.
Delivery work: Specify intervention and referral threshold; secure qualified staff and escalation to specialist care; evaluate outcomes.
Evidence, scores and monitoring →Trust in political money
Original assessment author (name not supplied) · 2026-10-06
Unscored: Increment beyond baseline; Expected incremental benefit; Incremental affordability.
Costing work: Regulator staffing, IT, compliance, litigation and transition.
Delivery work: Separate disclosure, enforcement and donation limits; define regulator powers, legal safeguards and consultation.
Evidence, scores and monitoring →A costed social-care capacity plan
Original assessment author (name not supplied) · 2026-10-06
Unscored: Increment beyond baseline; Expected incremental benefit; Incremental affordability.
Costing work: Full workforce/pay bill, provider viability, capital, commissioning, transition and long-term liabilities.
Delivery work: Model unmet need; specify pay, recruitment, training and commissioning; support unpaid carers; agree recurring funding.
Evidence, scores and monitoring →SEND
Original assessment author (name not supplied) · 2026-10-06
Unscored: Increment beyond baseline; Expected incremental benefit; Incremental affordability.
Costing work: Specialists, training, placements, transport, capital and recurring provision; separate historic deficits from future costs.
Delivery work: Define specialist and mainstream support; protect entitlements; plan staff pipeline and commissioning; track delivered provision.
Evidence, scores and monitoring →Waiting lists
Original assessment author (name not supplied) · 2026-10-06
Unscored: Increment beyond baseline; Expected incremental benefit; Incremental affordability.
Costing work: Capital, staff, diagnostics, maintenance, displaced activity and recurring running costs.
Delivery work: Map diagnostic, theatre, staffing and discharge constraints; select a specialty/locality; demonstrate addition beyond funded plans.
Evidence, scores and monitoring →Collect the tax already due
Original assessment author (name not supplied) · 2026-10-06
Unscored: Expected incremental benefit; Incremental affordability.
Costing work: Recruitment/training, systems, litigation, taxpayer compliance and collection costs; gross yield is not net cash.
Delivery work: Reconcile existing forecast baseline; define cases, staff, legal powers and collection horizon; audit net additional receipts.
Evidence, scores and monitoring →Settled homes for families in temporary accommodation
Original assessment author (name not supplied) · 2026-10-06
Unscored: Incremental affordability.
Costing work: Capital, finance, refurbishment, rent, maintenance, support and actual avoided temporary-accommodation costs.
Delivery work: Define families and locations; compare acquisition, lease and new build; inspect standards; secure lettings and tenancy support.
Evidence, scores and monitoring →Close capital-tax reliefs rather than raise rates
Original assessment author (name not supplied) · 2026-10-06
Unscored: Increment beyond baseline; Expected incremental benefit; Incremental affordability.
Costing work: Net receipts after behaviour, administration and transition; investment and compliance effects.
Delivery work: Name reliefs, thresholds, interactions and transition; obtain distributional/behavioural costing and administrative assessment.
Evidence, scores and monitoring →Warm homes, inspected before payment
Original assessment author (name not supplied) · 2026-10-06
Unscored: Incremental affordability.
Costing work: Works, assessment, inspection, maintenance, remediation and administration against existing grants.
Delivery work: Target poor-quality homes; survey suitability/ventilation; use qualified installers and independent inspection; establish redress.
Evidence, scores and monitoring →Claim what people are owed
Original assessment author (name not supplied) · 2026-10-06
Unscored: Expected incremental benefit; Incremental affordability.
Costing work: Outreach and administration plus additional entitlement paid; show fiscal transfers separately from social resource costs.
Delivery work: Specify eligible cohort, trusted outreach, application support and comparison group; arrange benefits-processing capacity.
Evidence, scores and monitoring →A child-poverty target and essentials test
Original assessment author (name not supplied) · 2026-10-06
Unscored: Increment beyond baseline; Expected incremental benefit; Incremental affordability.
Costing work: Separate low-cost monitoring from recurring policy measures; model household distribution and behavioural effects.
Delivery work: Define poverty/essentials measures, then compare specific benefit, housing-cost and service changes; assign accountability.
Evidence, scores and monitoring →Property and land tax reform
Original assessment author (name not supplied) · 2026-10-06
Unscored: Expected incremental benefit; Incremental affordability.
Costing work: Valuation, administration, transition relief, net receipts and behavioural response by jurisdiction.
Delivery work: Separate council-tax revaluation, land/property tax and transaction-tax changes; model losers and deferral options.
Evidence, scores and monitoring →A tax on the very top fortunes: three designs
Original assessment author (name not supplied) · 2026-10-06
Unscored: Increment beyond baseline; Expected incremental benefit; Incremental affordability.
Costing work: Valuation, enforcement, liquidity arrangements, net revenue and behavioural scenarios for each design.
Delivery work: Separate annual wealth tax, minimum effective tax and one-off levy; define base, residence, valuation and anti-avoidance.
Evidence, scores and monitoring →Utilities: stop overpaying for finance, make failure cost shareholders
Original assessment author (name not supplied) · 2026-10-06
Unscored: Increment beyond baseline; Expected incremental benefit; Incremental affordability.
Costing work: Acquisition/compensation, debt, financing, capital maintenance, contingent liabilities and consumer bills under each option.
Delivery work: Separate sectors and instruments; assess legal/regulatory powers, financing, compensation, investment and service quality.
Evidence, scores and monitoring →Work that pays
Original assessment author (name not supplied) · 2026-10-06
Unscored: Increment beyond baseline; Expected incremental benefit; Incremental affordability.
Costing work: Employer costs and incidence, enforcement, support programme costs and fiscal effects by option.
Delivery work: Separate wage-floor/enforcement, progression and targeted youth support; specify increment and monitor hours and access.
Evidence, scores and monitoring →Childcare places that parents can actually use
Original assessment author (name not supplied) · 2026-10-06
Unscored: Expected incremental benefit; Incremental affordability.
Costing work: Verify unmet demand, existing awards and provider viability; cost places, workforce, premises, operating subsidy and parental fees. Track access and sustained parental work. Proposed owners: councils, DfE and providers in England; separate devolved arrangements.
Delivery work: Verify unmet demand, existing awards and provider viability; cost places, workforce, premises, operating subsidy and parental fees. Track access and sustained parental work. Proposed owners: councils, DfE and providers in England; separate devolved arrangements.
Evidence, scores and monitoring →Transport access to jobs
Original assessment author (name not supplied) · 2026-10-06
Unscored: Expected incremental benefit; Incremental affordability.
Costing work: Name route/mode, baseline users, service hours and comparator; cost operation, capital, subsidy and maintenance. Assess accessibility and displacement, not gross local job counts. Proposed owners: transport authorities/operators and relevant national bodies.
Delivery work: Name route/mode, baseline users, service hours and comparator; cost operation, capital, subsidy and maintenance. Assess accessibility and displacement, not gross local job counts. Proposed owners: transport authorities/operators and relevant national bodies.
Evidence, scores and monitoring →Employer-linked shortage training beyond construction
Original assessment author (name not supplied) · 2026-10-06
Unscored: Expected incremental benefit; Incremental affordability.
Costing work: Specify sector, cohort, placement and recognised completion; cost tuition, trainee support, employer time, trainers and evaluation. Track earnings and sustained employment. Proposed owners: skills bodies, colleges and employers.
Delivery work: Specify sector, cohort, placement and recognised completion; cost tuition, trainee support, employer time, trainers and evaluation. Track earnings and sustained employment. Proposed owners: skills bodies, colleges and employers.
Evidence, scores and monitoring →Existing-firm management and technology adoption
Original assessment author (name not supplied) · 2026-10-06
Unscored: Expected incremental benefit; Incremental affordability.
Costing work: Define firm eligibility and existing offer; cost advisers, adoption support and evaluation, including firm contributions. Measure additional productivity, survival and displacement. Proposed owners: business department, local business-support bodies and delivery partners.
Delivery work: Define firm eligibility and existing offer; cost advisers, adoption support and evaluation, including firm contributions. Measure additional productivity, survival and displacement. Proposed owners: business department, local business-support bodies and delivery partners.
Evidence, scores and monitoring →Access to effective adult depression treatment
Original assessment author (name not supplied) · 2026-10-06
Unscored: Incremental affordability.
Costing work: Verify unmet clinical need and commissioned care; cost therapists, supervision, triage, follow-up and escalation; monitor symptoms, functioning, deterioration and access. Proposed owners: NHS commissioners and providers; jurisdiction-specific.
Delivery work: Verify unmet clinical need and commissioned care; cost therapists, supervision, triage, follow-up and escalation; monitor symptoms, functioning, deterioration and access. Proposed owners: NHS commissioners and providers; jurisdiction-specific.
Evidence, scores and monitoring →Targeted loneliness support
Original assessment author (name not supplied) · 2026-10-06
Unscored: Expected incremental benefit; Incremental affordability.
Costing work: Specify cohort and intervention; cost facilitators, venues, accessible transport and sustained participation. Measure loneliness, exclusion and wellbeing. Proposed owners: councils, public health and voluntary-sector partners.
Delivery work: Specify cohort and intervention; cost facilitators, venues, accessible transport and sustained participation. Measure loneliness, exclusion and wellbeing. Proposed owners: councils, public health and voluntary-sector partners.
Evidence, scores and monitoring →Targeted access to smoking treatment
Original assessment author (name not supplied) · 2026-10-06
Unscored: Incremental affordability.
Costing work: Verify service gap and current funding; cost adviser time, treatment, referral and follow-up. Measure sustained quits and inequalities, with adverse effects monitored. Proposed owners: public-health commissioners and health providers.
Delivery work: Verify service gap and current funding; cost adviser time, treatment, referral and follow-up. Measure sustained quits and inequalities, with adverse effects monitored. Proposed owners: public-health commissioners and health providers.
Evidence, scores and monitoring →Close early-language support delivery gaps
Original assessment author (name not supplied) · 2026-10-06
Unscored: Incremental affordability.
Costing work: Audit reach, eligibility and delivery fidelity; cost trained staff, release time, screening, materials and follow-up. Protect SEND provision. Measure language outcomes and access. Proposed owners: schools, education bodies and councils as appropriate.
Delivery work: Audit reach, eligibility and delivery fidelity; cost trained staff, release time, screening, materials and follow-up. Protect SEND provision. Measure language outcomes and access. Proposed owners: schools, education bodies and councils as appropriate.
Evidence, scores and monitoring →Timely workplace adjustments for disabled workers
Original assessment author (name not supplied) · 2026-10-06
Unscored: Expected incremental benefit; Incremental affordability.
Costing work: Audit delays and current legal/service baseline; cost assessors, adjustments, employer contributions and administration. Measure retention, access and worker experience. Proposed owners: DWP, employers and support providers.
Delivery work: Audit delays and current legal/service baseline; cost assessors, adjustments, employer contributions and administration. Measure retention, access and worker experience. Proposed owners: DWP, employers and support providers.
Evidence, scores and monitoring →Prevent avoidable loss of rented homes
Original assessment author (name not supplied) · 2026-10-06
Unscored: Expected incremental benefit; Incremental affordability.
Costing work: Define arrears/risk trigger, legal advice and assistance design; cost staff, grants and sustained tenancies. Assess additional prevention against existing duties. Proposed owners: councils, housing/advice providers and landlords.
Delivery work: Define arrears/risk trigger, legal advice and assistance design; cost staff, grants and sustained tenancies. Assess additional prevention against existing duties. Proposed owners: councils, housing/advice providers and landlords.
Evidence, scores and monitoring →Meter water-stressed households with an effective social tariff
Original assessment author (name not supplied) · 2026-10-06
Unscored: Incremental affordability.
Costing work: Incremental lifetime costs and funding not established. Require a funded protection design and additional rollout baseline; do not automatically generalise one company estimate.
Delivery work: Compare the extra cohort with the committed rollout; cost meters and tariff protection before proceeding. Require a funded protection design and additional rollout baseline; do not automatically generalise one company estimate.
Evidence, scores and monitoring →Fix NHS equipment and estate bottlenecks that cancel care
Original assessment author (name not supplied) · 2026-10-06
Unscored: Expected incremental benefit; Incremental affordability.
Costing work: Incremental lifetime costs and funding not established. Scheme costs and staff capacity must be established before approving expansion.
Delivery work: Prioritise safety-critical work separately; appraise operational fixes using downtime and completed-care data. Scheme costs and staff capacity must be established before approving expansion.
Evidence, scores and monitoring →Protect flood-defence maintenance and appraise schemes by risk
Original assessment author (name not supplied) · 2026-10-06
Unscored: Expected incremental benefit; Incremental affordability.
Costing work: Incremental lifetime costs and funding not established. Use engineering and risk appraisal; low trial evidence is not grounds to leave essential protection unmanaged.
Delivery work: List safety/maintenance gaps, cost them and appraise new schemes using expected damage and distribution. Use engineering and risk appraisal; low trial evidence is not grounds to leave essential protection unmanaged.
Evidence, scores and monitoring →Rehearse continuity for critical cloud-dependent services
Original assessment author (name not supplied) · 2026-10-06
Unscored: Expected incremental benefit; Incremental affordability.
Costing work: Incremental lifetime costs and funding not established. No invented national AI-outage loss; retain service-specific risk thresholds.
Delivery work: Run a costed service-level exercise before buying blanket duplication. No invented national AI-outage loss; retain service-specific risk thresholds.
Evidence, scores and monitoring →Make council and SEND funding risks explicit before promises expand
Original assessment author (name not supplied) · 2026-10-06
Unscored: Expected incremental benefit; Incremental affordability.
Costing work: Incremental lifetime costs and funding not established. Do not treat asset sales or borrowing permission as new recurring grant.
Delivery work: Publish a costed continuity plan and alternatives for deficits without hiding service reductions. Do not treat asset sales or borrowing permission as new recurring grant.
Evidence, scores and monitoring →Make heat protection operational in vulnerable settings
Original assessment author (name not supplied) · 2026-10-06
Unscored: Expected incremental benefit; Incremental affordability.
Costing work: Incremental lifetime costs and funding not established. Use risk-based standards; a mild summer is not proof the intervention worked.
Delivery work: Cost a local risk-led plan before the next warm season, prioritising people unable to protect themselves. Use risk-based standards; a mild summer is not proof the intervention worked.
Evidence, scores and monitoring →Protect defence sustainment and retention through risk appraisal
Original assessment author (name not supplied) · 2026-10-06
Unscored: Increment beyond baseline; Expected incremental benefit; Incremental affordability.
Costing work: Incremental lifetime costs and funding not established. Low public evidence does not imply defence is unimportant or exempt from scrutiny.
Delivery work: Require a costed sustainment assessment alongside procurement commitments; publish only safe aggregate assurance. Low public evidence does not imply defence is unimportant or exempt from scrutiny.
Evidence, scores and monitoring →Track asylum cases through appeals and safe move-on
Original assessment author (name not supplied) · 2026-10-06
Unscored: Expected incremental benefit; Incremental affordability.
Costing work: Full incremental costing and allocation remain unverified. Specify the process change and unit costs; no banked hotel saving without alternative accommodation.
Delivery work: Audit an end-to-end cohort and test one bottleneck with tribunal/council participation. Specify the process change and unit costs; no banked hotel saving without alternative accommodation.
Evidence, scores and monitoring →Phase pension-saving reform with low-income protection
Original assessment author (name not supplied) · 2026-10-06
Unscored: Expected incremental benefit; Incremental affordability.
Costing work: Full incremental costing and allocation remain unverified. No use of pension assets as free government money.
Delivery work: Update cohort modelling and consult on who pays before committing. No use of pension assets as free government money.
Evidence, scores and monitoring →Test targeted violence prevention on harm and community trust
Original assessment author (name not supplied) · 2026-10-06
Unscored: Increment beyond baseline; Incremental affordability.
Costing work: Full incremental costing and allocation remain unverified. National funding alone neither establishes nor refutes the additionality of a specific local improvement.
Delivery work: Identify a real delivery gap and compare hotspot cohorts; preserve proportionality safeguards. National funding alone neither establishes nor refutes the additionality of a specific local improvement.
Evidence, scores and monitoring →Manage court throughput with prisons and probation capacity
Original assessment author (name not supplied) · 2026-10-06
Unscored: Expected incremental benefit; Incremental affordability.
Costing work: Full incremental costing and allocation remain unverified. Legal-policy changes need their own appraisal; do not rank constitutional rights by this total.
Delivery work: Produce a costed court/prison/probation flow plan; do not infer that fewer jury trials is the answer. Legal-policy changes need their own appraisal; do not rank constitutional rights by this total.
Evidence, scores and monitoring →
1.How we find claims
We monitor public posts on Facebook and Instagram for political claims circulating in Portsmouth, Hampshire and South East England. This monitoring uses Meta’s Content Library API, accessed for research purposes, alongside claims submitted to us directly by readers.
Our monitoring is limited to publicly available content: public Facebook Pages, public Facebook Groups and public Instagram accounts. We do not access private profiles, private groups, private messages or any content that is not publicly visible.
We search around locally relevant topics and keywords, including:
- Portsmouth City Council and Hampshire County Council
- Southampton, Fareham, Gosport, Havant and Basingstoke local authorities
- Asylum accommodation and asylum support costs
- Council tax and local government reorganisation (LGR)
- Local crime figures, care services and council spending
We prioritise claims that meet two tests: they have achieved high engagement with local audiences, and they contain a checkable statistical element — a cost in pounds, a percentage, a count, a comparison. We select claims on reach and checkability, never on the politics of the person making them.
2.How we verify
Every fact check follows the same four-step process:
Step 1 — Archive the original claim
Record the original wording, permalink and date. An archived copy or screenshot should be linked when available; do not imply one exists when none has been captured.
Step 2 — Trace the claim to its primary source
We identify where the underlying figure actually comes from and obtain it directly: Freedom of Information responses, council meeting minutes and accounts, Home Office statistics, ONS datasets, Ombudsman reports, or Hampshire Constabulary crime statistics. We do not rely on a media report of a figure when the underlying document is obtainable.
Step 3 — Seek a right of reply
Where a fact check concerns a specific organisation or public figure — a council, an MP, a public body — the record should state whether a response was requested and reflect any substantive reply. Missing correspondence must not be presented as a completed right-of-reply process.
Step 4 — Write and publish the verdict
We set out the claim, the evidence, and our rating, linking every source so readers can verify our work themselves. The review method, reviewer and limitations should be recorded with the verdict. The site-wide evidence programme uses documented agent review authorised by the project owner; it must not be described as independent human assurance. Missing or failed evidence keeps an assessment provisional.
3.Our rating system
Fact checks use the labels below from the same registry as the claim pages. Six describe the evidence; Under review, Superseded and Corrected describe review or change history. Draft, Published and Withdrawn are separate publication states. A draft label is never a final verdict.
Supported
The claim is accurate and fairly presented. The figures match the primary source, and no relevant context has been omitted.
Mostly supported
The central claim holds up, but a specific figure, date or detail is off, or one part of a multi-part claim does not — the core is right, a detail is not.
Missing context
The claim is technically accurate on its own terms but omits information a reader would need to avoid a false impression — the same standard this site’s editorial-context work applies to headlines generally.
Misleading
The claim contains an element of truth but is presented in a way likely to leave readers with a false impression.
Incorrect
The central factual assertion of the claim is contradicted by the best available primary evidence.
Unverifiable
The claim cannot currently be verified or refuted from reliable primary sources.
Superseded
The claim was accurate when made but has since been overtaken by newer data, a policy change, or a revised official figure.
Under review
This claim has not yet been finalised — it is being checked against primary sources and its verdict may change.
Corrected
The original verdict or a figure on this claim was wrong and has been corrected — see the correction history for what changed and why.
Delivery: progress against the promise
Delivery ratings answer whether the pledge is on course on its own measure and deadline. They do not rate the truth of a quotation or the quality of a source. Opposition proposals are not graded as government delivery.
- On track
- The measure the pledge names is moving to the target on time, or the promised action has been completed.
- Mixed
- Part of the pledge has been met and part has not, or the evidence points both ways.
- Off track
- On the pledge’s own measure, the latest result is not on course to meet the target.
- Too early
- The deadline is far enough away that a verdict now would be premature.
- Cannot be assessed
- A matching measure or sufficiently defined assessment test is unavailable in this review; missing evidence is not failure.
Implementation stages are a separate question
The current-government tracker records a policy’s stage, such as consultation, funding or legal commencement. In force does not establish a measured outcome; an announcement cannot automatically be mapped to On track.
- In force
- In force and taking effect.
- Announced & funded
- Money and a delivery route have been committed. The delivery board shows whether a date has been set.
- Dated & costed
- A date, an estimated cost and a mechanism have been stated; funding or legal authority awaits a decision.
- Needs legislation
- Cannot happen until Parliament passes something.
- Proposal / consultation
- Proposal published; final form not settled.
- Stated intention
- A commitment without a complete dated and costed delivery plan; individual steps may have begun.
- Delayed or revised
- Timetable slipped or the policy has been rewritten since announcement.
- Dropped
- No longer being pursued.
Historical Brexit promises retain their recorded outcome qualifications, including Partly supported, Not yet measurable and Not falsifiable. Those are not silently converted into a stronger factual verdict. Read the original tests and limitations.
4.Sourcing standards
We trace statistical figures to primary sources where available and identify secondary reporting, research and attributed opinion explicitly. Readers should be able to click through to the same document we used and check our reasoning end to end.
Sources we routinely rely on include:
- Office for National Statistics (ONS) datasets and bulletins
- Home Office statistics, including asylum support and accommodation data
- Local authority statements of accounts, budgets and committee papers (Portsmouth City Council, Hampshire County Council and neighbouring authorities)
- Hampshire Constabulary and police.uk crime statistics
- Local Government Association publications and Local Government & Social Care Ombudsman reports
- Freedom of Information responses, published in full
Where sources conflict, we say so, explain which we consider more reliable and why, and rate the claim accordingly. Where the only available evidence is secondary, we identify it as such.
5.Corrections policy
We hold ourselves to the same standard we apply to others. If we get something wrong, we correct it promptly and transparently:
- Substantive errors are corrected as soon as they are established, normally within 48 hours.
- Every corrected article carries a dated changelog stating what was changed and why. We do not silently edit published verdicts.
- If a correction changes a rating, we say so prominently at the top of the article.
To request a correction, email connect@uk-statpack.co.uk with the article link and the evidence you believe we have missed. We acknowledge every correction request and respond with our decision and reasoning. Recorded corrections and page histories are listed on our corrections page.
6.How UK Facts is built
This site is built and researched with extensive use of AI. That is worth stating plainly rather than leaving readers to guess, and worth being specific about what it does and does not mean for the numbers on this site.
- AI is used extensively in building and researching this site — drafting page structure, pulling figures from official datasets, and proposing first-draft analysis.
- Multiple agents and models are often used to challenge and compare each other’s outputs, rather than one model’s answer being taken as final.
- Model output is never treated as evidence. A model saying a number is true does not make it true. Figures should trace to a named publication; where a figure comes from secondary analysis or our own calculation, the page should say so.
- Claims and figures are meant to be checked against their sources before publication, not accepted on a model’s say-so. Not every check is complete or recorded yet; a page’s own source and review notes say what has been checked.
- Calculations are reproducible: where this site has computed something itself (rather than quoting a published figure), the inputs and method are stated so a reader could redo the arithmetic.
- The project owner has authorised documented agent review for the site-wide evidence programme. Source verification, analytical checks and visual inspection remain required. Agent review and passing automated checks are not independent human approval or AQuA certification.
- AI can still hallucinate — state things confidently that are not true. This process reduces that risk; it does not eliminate it.
- UK Facts does not claim 100% accuracy. Corrections and uncertainty are treated as a normal part of the system, not a failure of it — see the corrections policy above.
The trust model
Six checks stand between a model’s draft and a published page:
- 1Primary-source verification
- 2Reproducible calculations
- 3Visible provenance
- 4Cross-checking
- 5Documented review and its limitations
- 6Public corrections
None of this eliminates error. It reduces it, and it means an error found has a published correction, not a quiet edit. If you find one, use the corrections process above.
7.Ethics & personal data
Our research involves analysing public social media content, and we handle that data carefully:
- We do not publish personal data of private individuals. Our fact checks address claims, not the people sharing them; private individuals who post or share a claim are not named or identified in our articles.
- We do not use any data we collect for advertising, ad targeting, profiling or political campaigning of any kind, and we do not sell or share it with third parties.
- See the About page privacy notice for implemented analytics storage and retention. Evidence records and source archives follow their stated review purpose; this page does not promise an unverified deletion schedule or server location.
- We honour deletion requests: if content we hold is deleted at source or its author asks us to remove it from our research records, we do so, subject only to what is necessary to support a published fact check.
Questions about personal data can be sent to connect@uk-statpack.co.uk.
Danny Shipp and James Harris run and self-fund UK Facts. They are responsible for what it publishes and for correcting mistakes. For questions, corrections and complaints, use the contact details on the About page. We use AI-assisted tools to develop the site and prepare material. AI output can be wrong, and automated or agent review is not independent human verification. Review dates, source records, limitations and correction histories describe the work performed; they do not imply a personal sign-off.
8.Funding & non-partisanship
UK Facts is a self-funded, not-for-profit public-information project. It is free to use, with nothing to buy. Funding must not determine the claims examined or the conclusions reached.
We are non-partisan by policy and by practice. We fact-check claims made by and about all political parties, applying identical evidentiary standards regardless of who benefits from the outcome. Contributors must disclose conflicts of interest, and no one holding an active role in a political party or campaign works on our fact checks. Our commitment to these standards is described further on our About page.
9.Evidence classes
The factual-verdict system in section 3 above is a verdict about whether a claim is true. It is a different question from what KIND of evidence a figure is — an official estimate can turn out correct or misleading; a forecast can turn out accurate or wrong. Pages built on this site's statistical registries label every figure with one of the evidence classes below, independent of any verdict about it, so a reader can see how solid the underlying evidence is before judging what it shows.
| Class | What it means |
|---|---|
| Accredited official statistic | Measured official statistical release; link to release and dataset. |
| Official estimate | Official estimate, including provisional, model-based or survey result; show qualifier. |
| Government-commissioned research | A government department commissioned a named third party to research this using its own methodology; not a measured official statistic and not UK Facts’ own analysis — name the commissioning department AND the researcher. |
| Regulator finding | An independent regulator’s own investigation, market-definition or decision — state the regulator, the case, and the decision’s legal status (e.g. final decision vs provisional finding). |
| Public-body operational record | A public body’s own administrative/operational record (facility specs, access rules, eligibility) — not a statistic, forecast or announcement; state the publisher and the record’s last-updated date, since these change without a formal release cycle. |
| Government announcement | Stated intention or announcement; not equivalent to delivery. |
| Forecast or scenario | Assumption-dependent future estimate; show assumptions and range. |
| UK Facts calculation from official data | UK Facts calculation; store formula, inputs and reproducible code. |
| Costed policy | Cost and funding published with enough detail to inspect. |
| Uncosted proposal | No credible complete costing; use purple status and state this plainly. |
| Independent analysis | Named organisation and method; do not present as official fact. |
| Attributed opinion | Clearly attributed viewpoint, never presented as a statistic. |
The three classes in the middle of the table above — government-commissioned research, regulator finding, and public-body operational record — were added 2026-09-25 for the AI capability and resilience feature, whose source register includes government-commissioned studies, a competition-regulator finding and a public body's own facility/access records: distinct evidence types the original 9-value scheme had no clean slot for. They follow the same rule as every other class here: name the actual publisher and, where the class requires it, the underlying researcher, the regulator and case, or the record's last-updated date.
Public-concern surveys
Question: What do you think are important issues facing the UK today? Respondents could choose more than one answer; the order of options was randomised.
Fieldwork ran from 5 to 30 August 2026 among adults aged 16 and over in Great Britain. This question had 3,550 respondents, within a wave of 3,560. People answered online or by telephone, after taking part in an earlier TLFS or OPN survey.
Survey weights represent the population; since April 2026 calibrated to 2024 mid-year estimates projected using 2022-based population projections. The 95% intervals are ONS estimates. Sampling and nonresponse uncertainty remain; multiple choices do not add to 100%.
This question began in October 2022, so it cannot supply a 1996 comparison. Collection frequency and weighting have changed. Concern is not a measure of service performance or policy success.
ONS release, 18 September 2026Original table 13ONS survey method
Ipsos Issues Index: a separate perspective
Question: What do you see as the main/other important issues facing Britain today? Ipsos combines main and other unprompted mentions. September 2026 fieldwork ran 2–8 September: 1,003 adults aged 18+ in Great Britain, interviewed by telephone and weighted to the population profile.
Sampling and non-sampling error; no overall confidence intervals supplied for these rows. None invented. Different questions, age bases, modes and dates mean ONS and Ipsos percentages cannot be averaged or treated as a trend. Topic pages keep their original response labels.
Ipsos release, 29 September 2026 · Original Table 4 and method
Last updated: 11 October 2026