Methodology

The working behind EnoughDay.

Tax rules live in per-tax-year config files. This tool uses the 2026/27 config set, checked against GOV.UK, HMRC and legislation.gov.uk on 2026-07-03.

Engine sources

Each config value carries a source URL and checked date. Future enacted changes are dated config values, not hidden code branches.

Simulation method

The deterministic projection is the base layer. The preview Monte Carlo view runs thousands of simulated futures from a seeded sampler, so repeated runs with the same inputs produce the same result.

The default sampler draws parametric lognormal annual returns: each year’s return comes from a fitted lognormal distribution — a bell-curve model of yearly returns — which is transparent and fast but understates the fat tails and clustered crashes real markets can show. It is not a resampling of historical UK return sequences: a block-bootstrap sampler exists, but the commercial historical dataset license is pending, so the default preview does not replay licensed historical data.

One tool on this site does replay history, and it is a separate, deterministic calculator — not part of this simulation. The free UK retirement history stress-test calculator replays published Bank of England annual series — long-run gilt and consol yields, which are in the public domain — across every historical start year on a fixed cash-and-gilts mix. The planner’s 5,000-future simulation does not use that dataset: it is parametric lognormal and does not replay historical sequences. The two answer different questions from different data.

National Insurance is annualized for planning. Actual payroll can differ by pay period, timing and payroll setup; the annualized model keeps the retirement projection stable and inspectable.

The single deterministic path compounds at the mean return with no variance, so over a long drawdown it is systematically higher than a typical simulated outcome (volatility drag and sequence-of-returns risk). The paid view therefore leads with the Monte Carlo success probability — the share of simulated futures in which the money lasts — alongside the median and the p10–p90 range of terminal wealth. The paid year table follows one representative simulated path: the path whose final wealth is the median across the runs. When more than half of paths deplete, that median terminal wealth is zero, and the representative path is the one whose depletion age is the median of the depleting paths, so the year-by-year detail reconciles with the median line rather than the optimistic mean path. The free view keeps the single deterministic central path and labels it as ignoring volatility and sequence-of-returns risk.

How we cross-check

Verification is layered: golden test vectors derived from legislation and HMRC worked examples, property tests, and a cross-check harness.

The cross-check harness runs a 50-household synthetic panel — every band edge one pound below, at, and above; the allowance taper; Scottish bands; pension contributions; marriage allowance; state-pension-age NI — through the engine and through an independent public calculator, and investigates every divergence over £1.

Each divergence ends in one of two ways: a fixed engine bug (test vector first), or a documented difference in convention listed under known simplifications below. Nothing is left unexplained; the harness fails otherwise.

Known simplifications

These limits are deliberate and stay visible as the engine expands.

Annualized NI rather than per-pay-period NI.
Simulated returns are drawn from a parametric lognormal model, not from historical return sequences.
Average-cost CGT rather than share matching.
No student loans in the MVP model.
No High Income Child Benefit Charge in the MVP model.
No Welsh rates divergence while Welsh rates match rUK.
No Gift Aid modelling in the MVP model.
Statutory self-assessment arithmetic, not PAYE payroll conventions: the personal allowance is the statutory GBP12,570, not a tax-code figure such as 1257L's GBP12,579, and the allowance taper over GBP100,000 is computed exactly rather than in whole-pound steps. Payroll calculators can differ by a few pounds a year for this reason.
Relief-at-source pension contributions follow the statute: taxable pay is unchanged and the basic-rate band extends by the gross contribution. Some payroll calculators instead deduct the contribution from taxable pay.
Marriage allowance is applied as the statutory tax reducer in 2026/27, not as an uplift to the recipient's allowance.
Household spending is two-phase: the spending-now figure applies until the primary person's retirement age, the retirement figure from then on. Both are entered in today's money and rise with simulated inflation.
Salaries default to a constant nominal level. Users can opt into CPI-linked salary growth; stated pension and wrapper contributions still stay as entered unless edited.
During working years, take-home pay funds spending first, then the stated pension and wrapper contributions, and any pay still left over is swept into savings: it fills any remaining ISA allowance first, then goes to a general investment account. Personal (relief-at-source) pension contributions cost 80 percent of the gross figure, with basic-rate relief added by the provider; employer pension contributions cost the household nothing. When take-home pay cannot cover the full stated contributions, they are scaled down, the least tax-advantaged wrapper first, rather than funded from money the household does not have. This runs at the household level and only while at least one person is still working. Once everyone has retired no pay is swept, so decumulation is driven purely by drawdown.
A single portfolio-level asset mix applies across every wrapper. An optional post-retirement mix switches once the whole household has retired, as a one-off step change rather than a continuous year-by-year glidepath.
Defined-benefit pensions and annuities are entered in today's money and taxed as pension income with no National Insurance, from the stated start age onward. Level income stays flat in cash terms; CPI-linked income rises with the simulated inflation path; fixed-percentage income grows at the stated rate each year measured from today.
Defined-benefit pensions and annuities are modelled for the named person only, with no survivor benefits, no mortality model, and no transfer values: the income simply stops at the projection horizon.
Couples are modelled as two people in one household, each with their own salary, pots, contributions, State Pension age (derived from their own date of birth) and defined-benefit or annuity income. Take-home pay, the contribution funding order and the surplus sweep are shared at household level and attributed to each working person by their net income; each person keeps their own ISA and pension allowances. The income-tax region is shared across the household.
The survivor what-if is a deterministic illustration, not a projection default. It models one partner dying and the other continuing alone: the survivor keeps their own assets and inherits the deceased's ISA (via the spousal Additional Permitted Subscription, modelled as an ISA balance transfer), the deceased's pension (spouse transfers are exempt from inheritance tax under both the current and the post-2027 regime; the survivor is modelled as drawing it tax-free if death is before age 75 and as taxable pension income if at or after 75) and the deceased's general investment account (a spouse transfer, made on a no-gain/no-loss basis with the original cost carried over, no capital gains tax on transfer). No survivor defined-benefit or annuity benefits and no joint-life annuities are modelled, the Additional Permitted Subscription is simplified to a balance transfer, and the transfer uses the household's current entered balances rather than balances projected to the death age; the death age sets only the pension tax treatment.
A primary home is modelled as a non-drawdown asset: its value grows each year, either tracking simulated inflation or at a fixed real rate above inflation, but it is never drawn on to fund spending and never counts toward the drawdown wealth that drives the success rate. It is shown separately and its net equity joins the estate for inheritance tax. Because a property is modelled, the residence nil-rate band is capped at the home's net equity when that is lower than the amount the chosen bands option allows, following the rule that the residence nil-rate band is the lower of the available band and the value of the home passing to direct descendants; plans with no property keep the band set purely by the chosen option. An optional mortgage amortises with simple annual interest: the fixed monthly payment is a non-discretionary outflow added to spending each year until the balance reaches zero, and a payment that would not cover the interest is not allowed. An optional downsize at a chosen age sells a share of the home; the released value, after transaction costs and repaying any remaining mortgage, moves into a general investment account with its cost basis set to the proceeds, so no capital gains tax arises on selling a main home. The remaining home keeps growing. When the released share does not raise enough to clear the outstanding mortgage, the remaining balance continues against the retained home rather than the whole home being sold to clear it. Equity release and lifetime mortgages, rental income, second homes, and renting after a downsize are not modelled; the home passes to the surviving spouse exempt from inheritance tax.