Every number. Every formula. Every source.

SortedUK's decode-v2 tool reads UK letters (HMRC, DWP, councils, courts, DVLA, NHS) and returns: sender, what it's about, the real deadline, what happens if you do nothing, your next step, and a colour-coded risk badge. Here's how each part is determined.

How we identify the sender

• Pattern matching against known UK government letterheads, postcodes, reference number formats (e.g., HMRC UTR is 10 digits, DWP claim references vary by benefit, court claim numbers follow a county-then-number format) and disclaimer footers ("Information is available in large print, audio and Braille…").
• AI classification using the live decoder (when the API key is set) against a UK-letter taxonomy of ~45 sender types. When the key is not set, falls back to the classic rule-based decoder.
• Cross-verification: if the letter mentions a phone number, we check it against a list of ~80 verified UK government and regulator helplines. Mismatch flags a scam.

How we set the risk badge

• 🟢 No action needed — informational letters, confirmations of awards, reminders with no consequence for the reader, courtesy notices.
• 🟡 Action soon — letters with a real deadline more than 14 days away, requests for evidence with consequences that are reversible.
• 🔴 Act now — court summonses, enforcement notices, MR/appeal windows closing within 14 days, benefit suspension threats, eviction notices, bailiff schedules. Any letter whose escalation path includes losing your home, your driving licence, your benefits or your bank account.

The decoder identifies dates by: (a) ISO and UK-format pattern matching on the letter body, (b) keyword anchoring around "reply by", "by", "no later than", "deadline", "within X days/weeks", and (c) sender-type defaults when a letter has no explicit date (e.g., a Mandatory Reconsideration request implies a 1-month deadline from the disputed decision letter; a PCN gives 28 days to challenge).

Defaults are sourced from the actual UK rule, not estimated:

• Self Assessment online: 31 January after end of tax year · source: HMRC, gov.uk/self-assessment-tax-returns/deadlines
• Self Assessment paper: 31 October after end of tax year
• Mandatory Reconsideration: 1 calendar month from decision letter · source: DWP Decision Makers' Guide
• First-tier Tribunal (SSCS): 1 calendar month from MR Notice · source: HMCTS
• Council PCN appeal: 28 days from issue (14 if discount paid) · source: Traffic Penalty Tribunal
• Private parking POPLA: 28 days from rejection letter · source: POPLA terms
• Companies House confirmation statement: 14 days from review date · source: gov.uk/file-your-confirmation-statement-with-companies-house
• Companies House accounts: 9 months after accounting period end (private Ltd) · source: Companies Act 2006 s.442
• FOS escalation: 6 months from firm's final response · source: FOS rules DISP 2.8.2
• Housing Ombudsman: 12 months from landlord's final response · source: HOS Scheme paragraph 26

The /better-off scan asks 7 questions (situation, household, income, age, employment, home, concerns) and returns six panels: Money you may be owed · Savings you can make · Home opportunities · Business opportunities · Risks · Recommended next steps. Each panel shows an estimated £/yr range based on real 2024/25 and 2025/26 UK scheme values.

How a range is built

Each recommendation has a low and high £/yr value drawn from the actual UK scheme. For Pension Credit: low = £1,000/yr (small top-up case) → high = £4,200/yr (DWP-published average award) → ceiling = £5,400/yr (single-person, full Guarantee Credit minimum income £218.15/wk - £121.10/wk State Pension = £97.05/wk × 52 weeks).

The presented range is intentionally honest about the variance. We do not show a single point estimate because eligibility variables (housing costs, savings, partner income) move the actual outcome.

Total improvement card

The dark navy total at the top sums the midpoints of the qualifying ranges, not the highs. We deliberately avoid over-promising. The figure shown is "Sorted believes you may be ~£X/yr better off if everything applies"; the breakdown shows which schemes contributed.

The Mortgage Ready Score takes income, deposit, target price, monthly debts, employment status and FTB status, then returns a 0–100 score across 4 weighted factors. This is a self-assessment tool, not an FCA-regulated affordability decision.

Verdict bands

• 80–100 ready: most high-street lenders will offer at requested LTV.
• 60–79 close: probable acceptance with some lenders; specialist broker recommended.
• 40–59 stretch: needs work on deposit, debts, or income evidence.
• 0–39 not yet: focus on the largest single factor first.

Stamp Duty in the result panel

SDLT calculated against current 2024/25 rules — see section 5 for full bands.

The SDLT figures shown on /home-buyer, /mortgage-ready-score and /decide use HMRC's current published bands. England & Northern Ireland only; Wales uses LTT, Scotland uses LBTT.

Standard residential (single property, primary residence)

First-time buyer relief

Additional property (second home, BTL)

The 13-question benefits checker screens against ~20 UK benefits using rule-based eligibility logic. Each benefit's gate is reproduced from the official DWP/HMRC/NHS rule. We do not estimate awards beyond what the rule lets us — for example, Universal Credit's exact monthly award depends on housing costs, deductions and earnings tapering we cannot compute without bank-grade data (which requires login + ICO Tier 1).

Examples of gate logic shown to the user

• Pension Credit: age ≥ 66 AND weekly income (single) < £218.15 OR (couple) < £332.95. Source: DWP rates 2025/26.
• Attendance Allowance: age ≥ 66 (State Pension age) AND personal-care or supervision need > 6 months. Source: gov.uk/attendance-allowance/eligibility.
• Carer's Allowance: caring 35+ hours/week for someone on qualifying disability benefit AND carer earnings < £196/wk after deductions. Source: gov.uk/carers-allowance.
• Marriage Allowance: married/civil partner AND one earner < PA (£12,570) AND other earner basic-rate (< £50,270 in England/NI/Wales; < £43,662 in Scotland). Source: gov.uk/marriage-allowance.
• Healthy Start: pregnant or child under 4 AND on qualifying benefit (UC with low income, IS, JSA, ESA). Source: NHS healthystart.nhs.uk.
• Council Tax Reduction: low income or specific status; rules vary by council. We show the typical structure (income test + capital test) but tell users the precise scheme is their local council's, with link to gov.uk/apply-council-tax-reduction.

/biggest-opportunity scores the user against 14 known UK opportunities by combining (a) eligibility match from the 8 quiz answers and (b) £/yr value of the opportunity.

The top-scoring opportunity is shown in the dark hero card. The next four are runners-up. Each card cites the source for the value and the eligibility test. Examples:

• Pension Credit avg award £4,200/yr (DWP National Statistics 2024)
• Attendance Allowance £3,843–£5,741/yr (DWP 2025/26 rates)
• Council Tax Reduction avg £900/yr (Policy in Practice 2025)
• Marriage Allowance £252/yr (HMRC published)
• Disabled Facilities Grant up to £30,000 (one-off, not /yr)

The Discover For Me 10-question wizard returns a single priority action by combining three signals from the answer set:

Four result cards (Money / Risks / Opportunities / Local) are then personalised by the answer set. Each card routes into an existing Sorted hub.

The /outcomes ledger publishes per-category counts, win rates and average £ amounts from opted-in users only. Today this is browser-only (localStorage). Once SortedUK Ltd is registered and ICO Tier 1 is paid, accounts will move this to a verified backend.

Opt-in protocol

• When a user saves a case to /passport and later marks it as resolved, they are asked: "Help the next person — share this outcome anonymously?"
• If yes: only the category, the outcome (won/partial/lost/abandoned), and an optional £ amount are added to the local sorted-outcomes-anonymised-v1 record. No personally identifying data.
• If no: the outcome stays on the user's case in their Passport. Nothing is shared.

What we publish

Per-category numbers are published only once a category has ≥30 verified outcomes. Below that threshold we show "Early data" and refuse to compute a win rate, to avoid headline statistics from too-small samples.

Every claim on SortedUK is sourced. Our verification protocol:

• Tier A — UK official: GOV.UK, HMRC, DWP, NHS, Companies House, VOA, Ofcom, Ofgem, Ofwat, FCA, ICO, FOS, Housing Ombudsman, ACAS, HSE, Ofsted, DVLA, HM Passport Office, HMCTS, ONS, CAA, CMA, ORR, PHSO. Direct paragraph-level citation where possible.
• Tier B — UK regulated: Citizens Advice, MoneyHelper, StepChange, Shelter, Age UK, Carers UK, Turn2us, Trussell Trust, Samaritans. Registered charities or government-funded services.
• Tier C — Sector data: Policy in Practice, British Business Bank, Innovate UK, Energy Ombudsman annual report, Carers UK State of Caring, UK Pet Food, ABI, Cabinet Office SME data.

We do not cite: third-party blogs, comparison sites, paid claims-management firms, generic Google searches, or "industry experts" without an institutional affiliation.

Every recommendation on SortedUK is tagged with a confidence band so you know how much to trust the number.

• High — the number is published by an official UK body, currently in force, and verified within the last 90 days. Examples: published SDLT bands, State Pension rates, HMRC penalties, NLW.
• Medium — the number depends on an assessment (e.g., PIP daily-living score, Council Tax band reassessment), or is a published average across all UK claimants. Examples: Pension Credit average award, FOS uphold rate.
• Low / Building — the number is from a small sample, early SortedUK opt-in data, or a sector estimate with known caveats. Examples: SortedUK's own per-category outcome rate while <30 outcomes; Truely-style estimates we don't have paid data for yet.

• Tax-year-sensitive numbers (PA, NI thresholds, ISA limits, SDLT, dividend tax, CGT allowance): reviewed at start of every new tax year (6 April) and within 7 days of any in-year change announced at Budget.
• Benefit rates: reviewed at start of every new benefit year (1 April for State Pension & means-tested benefits; 1 April for UC).
• Helpline numbers: reviewed monthly via a sitewide grep of tel: links and verified against the listed body's published contact page.
• Verified-source URLs: reviewed quarterly. Any 404 triggers a corrections row.
• FAQ / Schema.org structured data: reviewed when the underlying answer changes, regardless of cadence.

When SortedUK gets something wrong:

• We fix the page within 7 days of being notified or noticing.
• We publish a correction row at /corrections with the date, the page, what was wrong, why, and what changed.
• We never quietly edit a factual claim and pretend it was always that way.
• The sitewide footer shows the running corrections count so it stays honest.

To flag a correction: corrections@sorteduk.uk

SortedUK is deliberately clear about what it is not:

• SortedUK is not FCA-regulated. We do not give personalised investment, mortgage, insurance or pension advice. Where a regulated decision is involved (mortgage offer, investment product, life insurance), we route the user to an FCA-regulated firm or to MoneyHelper.
• SortedUK is not a UK charity. We are not registered with the Charity Commission. Donations are not tax-deductible.
• SortedUK is not a law firm. We do not give legal advice. Where a solicitor is needed (will, LPA dispute, conveyancing, employment tribunal representation), we route to SRA-regulated firms or to legal aid routes.
• SortedUK is not a tax adviser. We do not file Self Assessment, register companies for tax, or represent against HMRC. Where this is needed, we route to chartered accountants (ICAEW / ACCA / CIOT).
• We do not have government endorsement. References to GOV.UK, HMRC, DWP and NHS are as sources, not as partners. None of those bodies has reviewed or approved SortedUK content.

If you see anything on SortedUK that implies otherwise, that's a bug. Please flag it at corrections@sorteduk.uk and we'll fix it within 7 days.