How RetireForecast works

Last updated: 2026-07-03.

This page explains how RetireForecast produces its figures: the methods behind the tax, pension, housing and longevity calculations, the assumptions they rest on, and — just as important — what the tool does not model. It is written to be honest about the edges. Every statutory figure the engine uses carries a source and the date it was last checked against that source.

Two companion pages give the numbers behind two of the biggest inputs: the economic assumptions (investment returns, inflation, house and salary growth) and the mortality data (how long people are assumed to live).

The shape of a forecast

Everything is in today's money. The engine works internally in future (nominal) pounds — so that frozen tax thresholds bite against rising incomes, exactly as real fiscal drag does — then converts every figure back to real, today's-money terms for display. Unless a figure is labelled otherwise, it is real and net of tax.
Money is exact. All amounts are held as whole pence, never as decimals that can drift. Tax rates are held exactly (8.75%, 39.35% and so on). Tax-free cash is rounded in your favour, so tax is never understated.
No magic numbers. Each rate and threshold lives in a per-tax-year record with a gov.uk source link and a "verified on" date. The tool currently holds the 2025-26 and 2026-27 tax years, for England, Wales and Northern Ireland. Scotland is deliberately refused rather than quietly given the wrong bands.
One engine, two views. A single year-by-year projection engine produces both:
- a central projection — one path, using the expected (best-estimate) return, inflation and a representative lifespan. This answers "what happens on the main assumptions?"
- a Monte Carlo — thousands of paths with randomly varying returns, inflation, lifespans and care needs. This answers "how likely is it, and what is the range of outcomes?"
Because both run the same engine, the central case and the simulated distribution stay consistent with each other by construction.

Income tax and National Insurance

UK income tax stacks three kinds of income in a fixed order — non-savings (earnings, pensions, the State Pension, rental income), then savings (interest), then dividends — through one shared set of tax bands. The engine reproduces this in a single pass, because each 0% allowance still uses up band space and pushes the income above it into higher bands:

The personal allowance (£12,570) is set against income in the statutory order, and is tapered away by £1 for every £2 of income over £100,000.
The basic (20%), higher (40%) and additional (45%) bands are filled in turn.
The personal savings allowance depends on the top band your income reaches (£1,000 at basic rate, £500 at higher, £0 at additional), and the £5,000 starting-rate band for savings is eroded pound-for-pound by non-savings income.
Dividends get their £500 allowance, then are taxed at the dividend rates for each band (from 2026-27 these rise to 10.75% ordinary / 35.75% upper, with 39.35% additional).

Each person is taxed individually, as in the UK system. The tool reproduces HMRC's published worked examples to the penny.

National Insurance is modelled for employees only: the main rate (8%) between the primary threshold and the upper earnings limit, 2% above it, and category-aware rates for reduced, deferred or nil categories. NI stops at State Pension age and is never charged on pension income, drawdown or the State Pension.

Not modelled: Scottish income-tax bands; employer (secondary) NI; self-employed Class 2/4 NI; marriage allowance, blind person's allowance, or Gift-Aid/pension band extension.

Pension withdrawals and the lump-sum tax shock

This is one of the two headline outputs. When you take money from a defined-contribution pension:

25% is tax-free (up to the lump-sum allowance of £268,275, tracked across all your pensions), and the rest is taxable.
On your first flexible withdrawal, HMRC applies emergency "Month 1" tax: it gives you only one-twelfth of your allowances against that single payment and taxes the rest at the top rates — so a large one-off withdrawal is massively over-taxed up front. The tool shows this over-deduction and which form you use to reclaim it: P55 (pot not emptied), P50Z (pot emptied, no other income) or P53Z (pot emptied, with other income).
The true tax finally due is worked out as the difference your withdrawal makes to a full income-tax calculation — so it correctly accounts for the way pension income can push your savings and dividends into higher bands.
Taking taxable pension money flexibly triggers the money-purchase annual allowance (MPAA), cutting future pension contributions to £10,000 a year. Taking only the tax-free cash does not.

The tool also models the annual allowance (£60,000), its high-income taper (down to £10,000) and carry-forward of unused allowance.

A note on the standalone "tax shock" panel: to keep it self-contained it assumes your other income is your salary if you are still working, and nothing once retired — it does not layer in the State Pension or a defined-benefit pension already in payment. The full year-by-year forecast does include those. The emergency over-deduction the panel leads with does not depend on that other income. The tool models the size of the emergency deduction, not HMRC's PAYE tables to the exact pound.

Defined-benefit pensions are paid from their normal retirement age, escalated for inflation, with optional tax-free commutation (a lump sum in exchange for a permanently reduced pension) and a survivor pension. An annuity can be bought with part of a pot, level or inflation-linked, single or joint life. The rate is worked out for you from published market rates for the age the income starts, whether it escalates and whether it continues to a survivor, and you can type your own quote over it. Buying one out of a pension pot crystallises the money it uses, so a quarter of it comes back to you as a tax-free lump sum and the other three quarters buy the income; an annuity bought with money that is not pension money has no lump sum, and only the interest part of each payment is taxed.

The State Pension

State Pension age is derived from date of birth using the legislated timetable (66, rising to 67 for those born 1960–61, and towards 68 for those born 1977–78).
Entitlement comes from your own forecast weekly amount, or is estimated pro-rata from qualifying years (nothing below 10 years; the full amount at 35). Deferral delays the claim: the pension is not paid during the deferral period (that income is forgone), and the later, uplifted rate — 1% more for every 9 weeks deferred — is drawn for life from the delayed start. So deferring only pays off if you live well past that start; a deferred pension still counts as income for Pension Credit while paused (you cannot gain Pension Credit by deferring).
Over the projection the triple lock is modelled as a proxy: the pension rises each year by the greater of inflation or 2.5%.

Not modelled: pre-1960 birth cohorts (out of the forward-looking scope); any State Pension age rise beyond 68; the earnings-growth leg of the triple lock (the proxy uses inflation or 2.5% only).

Selling or downsizing your home

When a home is sold, the tool works down a sale waterfall:

sale price − mortgage repaid − selling costs − any capital gains tax = net proceeds

Selling costs default to 2% of the price if you do not itemise them. If you then buy somewhere cheaper, it deducts the purchase price, stamp duty and moving costs; any surplus is invested. If the cheaper home still costs more than the sale frees and you have set an interest-only (retirement) mortgage rate, the shortfall is borrowed and carries a lifetime interest cost. If you sell and rent, the whole net proceeds are invested and rent is paid each year, growing at its own inflation rate.

Renting has to be agreed as well as afforded. A landlord's referencing agency asks for gross annual income of at least 30 times the monthly rent, and the standard income reference does not count the money the sale frees. The tool checks each year of a rent plan against that bar and flags the years that fall short. Where it is short, it states the usual way round it and what it costs: a guarantor referenced at 36 times the monthly rent against their own income. It also says that rent in advance is no longer a way round: since 1 May 2026 a landlord in England may ask for at most one month of it. Some referencing firms accept proof of funds from a completed sale instead; the flag says so but does not model it. The deposit is charged as a real cost at the start of a rent plan, capped as the Tenant Fees Act 2019 caps it: five weeks' rent, or six weeks at an annual rent of £50,000 or more. The first month's rent in advance is not charged twice, because the first year of a monthly-in-advance tenancy is twelve payments and the year's rent already charges twelve; the day-one total you would have to produce is stated on the result.

Stamp duty (SDLT) is charged band by band (0% to £125k, 2% to £250k, 5% to £925k, 10% to £1.5m, 12% above). The 3%–5% additional-property surcharge exists in the engine but is not applied in the buy-vs-rent comparison, which assumes you are replacing your main residence.

Capital gains tax on a home is worked out the way HMRC does it — by occupation, not mortgage type. A home you have lived in throughout is fully relieved (£0 tax). If it was ever let, relief is time-apportioned: the years you lived there (plus the final 9 months) are relieved, and the rest of the gain is taxable. Jointly-owned homes split the gain, and each owner gets their own £3,000 allowance and pays their own rate (18% or 24%).

Time you spent away from the home can still count as living there ("deemed occupation"): up to 3 years in total for any reason, up to 4 years in total where a job kept you elsewhere in the UK, and any length of time working abroad. An away period only counts if the home was your main home before it, and you came back to it afterwards, though a job that stopped you coming back is excused that.

Not modelled: Welsh LTT / Scottish LBTT; the SDLT surcharge on the replacement home; lettings relief. A let period other than the final 9 months is treated as chargeable.

Means-tested benefits — the downsizing trap

Freeing home equity into savings can reduce means-tested help, which the tool surfaces so it is not a surprise:

Capital tariff: the first £10,000 of savings is ignored; above that, every £500 (or part) is treated as £1 a week of income.
The £16,000 cliff: Housing Benefit and Council Tax Support stop once savings exceed £16,000. (Pension Credit itself has no upper capital limit.) The forecast flags the first year your assessable capital is above the limit, as a note on your inputs. One exception: a household actually being paid Guarantee Credit is passported to both with no upper capital limit, so it is never told its savings have ended them.
Pension Credit (Guarantee Credit) is calculated every year of the projection: the minimum guarantee for a single person or couple, less your assessable income and capital tariff. Both members must be over State Pension age (a mixed-age couple gets nothing). Assessable capital includes the equity of a let former home — so selling or letting your home can erode the credit dynamically, year by year.
Pension Credit is contingent, not secure. The results page reports it beside the guaranteed income floor, never inside it. It has to be claimed, around a third of eligible pensioner households never claim it, and once claimed it moves with income, with capital, with a change of circumstances and with a review. The claim prompt is also shown to a household that receives nothing, when its assessable income is within 10% of the guarantee: this tool models the Guarantee Credit alone and applies no income disregards, so at that distance its own simplifications can be the whole of the gap, and only the DWP can settle entitlement.

The severe-disability addition applies where the household qualifies — a single disabled pensioner, or a couple where both partners receive a qualifying disability benefit (a non-disabled partner blocks it) — and a carer addition where a partner cares for a disabled partner. Not modelled: Savings Credit (Guarantee Credit only); Housing Benefit / Council Tax Support are never awarded, only their loss above £16,000 is flagged (and not flagged at all for a household on Guarantee Credit, which keeps both); the carer addition is supported by the engine but not yet exposed as a builder input.

Support for Mortgage Interest

A year of Guarantee Credit also brings Support for Mortgage Interest, with no waiting period at pension age, and the forecast models it wherever the household still owns its home. It meets the interest on the first £100,000 of your mortgage at the government's own standard rate of 2.09% a year, and for a pensioner it covers your service charge and ground rent in full. That is often the cheapest secured borrowing anyone in later life can get: it is roughly a third of the rate a lifetime mortgage rolls up at, on the same security.

It is a loan, not a payment. What is met comes off your spending, and the same amount is added to a second charge on your home, which rolls up alongside the mortgage and is repaid when the home is sold or when you die. The forecast shows it as its own balance and takes it off the equity, so the help you get now and the inheritance it costs are both on the page.

Not modelled: whether you would actually claim it (you have to apply and accept the charge; the forecast assumes you do); the separate interest rate the government charges on the loan, which is set from gilt yields (the forecast rolls the charge up at the standard rate instead); and the exclusions on particular service charges, beyond leaving out the part of yours that buys utilities. Nothing is met against a mortgage you pay no cash interest on, such as a rolled-up lifetime mortgage, because there is then no interest for it to meet.

Inheritance tax

Inheritance tax is offered as a toggle, to compare spending your pension down against preserving an estate. When it is on, the forecast values your estate at each death inside the projection and computes the tax due: the nil-rate band (£325,000), the residence nil-rate band (£175,000, tapered away £1 for every £2 of estate over £2,000,000 and capped at the home passing to direct descendants), and 40% on the rest. The April 2027 change that brings unused pensions into the estate applies to any death from that year on.

The result depends on your relationship status (a household input, defaulting to married / civil partnership):

Married or in a civil partnership: on the first death everything passes to the survivor free of tax (the spouse exemption), and both partners' nil-rate bands are available on the second death (the transferable band — modelled as two full bands). The tax lands on the second death, when the estate passes to direct descendants.
Cohabiting (not married): there is no spouse exemption and no shared band. The deceased's share of the estate passing to the surviving partner on the first death is a chargeable transfer, and each death has only one set of bands — so the same estate is taxed more heavily. A cohabiting couple also can't inherit a State Pension or (usually) a scheme's survivor pension; the forecast flags where those may overstate the survivor's income.

The estate at each death is valued from the projection state as liquid savings + home equity (net of mortgage) + unused pension (the pension counting only from April 2027). It is computed in the year's nominal pounds against the frozen bands (so a growing estate against a frozen band is taxed more over time, the real fiscal drag) and then shown in today's money. A first death splits the jointly-owned home 50/50 (immaterial for a married couple, whose first death is exempt).

Not modelled (so this is an illustration of the headline bands, not a full estate computation): lifetime gifts and the 7-year taper; trusts; business or agricultural relief; the reduced 36% charitable rate; non-descendant beneficiaries or split legacies; the income tax a beneficiary later pays on an inherited pension. The headline panel values deaths at the plan's representative (median) ages; a completed Monte Carlo run also shows the spread of IHT across futures (the share leaving any bill, and the median vs high-end), since it varies with longevity and returns.

ISA allowances

An ISA shelters investment returns from tax entirely, and the amount you may pay in is capped at £20,000 per person per tax year. The forecast enforces that cap on any regular saving you enter into an ISA. Anything above the allowance is not lost — it goes into your general investment account instead, where its dividends and interest are taxed, which is what would happen in reality. The cap applies only to money going in: a pot that has grown past the allowance inside an ISA is entirely legitimate and is left alone.

The forecast also uses the allowance, not just enforces it. Each year, whatever is left of your £20,000 after anything you pay in is spent moving money you already hold in a general investment account into an ISA, so from then on its growth and dividends are tax-free. This is what people call "bed and ISA", and it is what a household with a large unwrapped pot would really do; assuming they never did it would show you paying tax you would not pay, and it would do so hardest on a plan that sells a home and invests the proceeds. Because moving the money means selling it, each move is a capital-gains disposal, so the forecast keeps it small enough that the gain stays inside your annual exempt amount and costs nothing. In a year where drawing money to live on has already used that exempt amount, nothing moves: spending comes first. We assume you do this, and say so on your results with the pounds involved. If you would not, the model can be told to leave the money where it is.

One thing this deliberately does not do. Because an ISA is modelled as a single invested balance rather than a cash sleeve plus a stocks sleeve, the April-2027 cut to the cash ISA allowance and the charge on cash held inside a stocks-and-shares ISA have nothing to apply to (both figures are recorded, dated, and ready if that changes).

What paying for advice would cost

Because the forecast charges investment costs at all, it can answer a question people actually ask: what would an adviser's fee do to this plan? It runs the same projection twice — once bearing the charges your assumptions already say it bears, once with an ongoing advice fee added on top — and reports the difference in lifetime charges, in what is left at the end, and in the year the money runs out.

The advised side is your own charge plus the fee, and nothing else. That is deliberate. The ongoing fee is the one component that can be benchmarked (the 2026 UK average is 0.83% a year, from NextWealth's fee benchmarking work); how much dearer an advised fund choice is varies far too much between firms to assume for any particular household, and building the advised total out of a guessed fund uplift would mean inventing the larger half of the number. You can replace the benchmark with a real quote.

This is a cost, not a verdict. Advice costing X% a year has to add more than X% a year of value to be worth paying for, and whether it does is a question this tool cannot answer. The most-cited component of an adviser's value is behavioural — talking someone out of selling in a crash — and none of it is modelled here. Neither is the cost of getting something wrong without one. A one-off piece of advice (commonly £1,500–£4,000, or a percentage of the amount invested) is charged on top and is not modelled.

If one of you died: the protection gap

A plan for two people quietly assumes you both live roughly as long as the tables say. The first death is the sharpest single change in a projection: one State Pension stops, a defined-benefit pension may drop to a survivor's fraction or stop altogether, any salary ends, and any means-tested top-up is recalculated — while spending falls only by the survivor factor. The forecast already shows that cliff in the secure-income floor; the protection panel puts a number on it.

For each of you, the tool pins your death to next year and re-runs the same deterministic projection on the same plan (the housing strategy currently on display, not a stay-put reading of a sell plan). It then searches for the smallest lump sum, arriving in that year, that puts the survivor's plan back where the couple's plan stands today. The bar is deliberately relative: "the money lasts at least as long as it does now". An absolute "must never run short" bar would be unanswerable for a plan that already runs short, and would blame a death for a shortfall it did not cause. The answer is rounded up to the nearest £1,000, because a protection figure rounded down would not quite do the job.

Employer death-in-service cover is modelled as its own input, because most employed people have it, most forget they do, and it can close the whole gap at no cost. It is a lump sum a group life scheme pays if you die while still employed — stated either as a multiple of salary (which keeps pace with pay) or as a fixed sum assured (which does not, so it is worth less the further off the death is). The tax treatment follows the registered-scheme rules: tax-free if you die under 75, up to your remaining lump sum and death benefit allowance (£1,073,100 less any tax-free cash already taken), and taxable as the recipient's income above that or if you die at 75 or over. It falls outside your estate for Inheritance Tax, including under the April-2027 rule that brings unused pensions in. To the survivor it is capital, not income, for means-tested benefits — so a large payout can end a survivor's Pension Credit through the capital tariff, which the forecast shows rather than hides.

The point of modelling it separately is the cliff: the cover ceases when the employment does. The panel therefore also prices the same death a year after retirement, when the policy has gone and the survivor still has decades to fund.

What this is not: it sizes the hole, it does not price or recommend a policy, and cover on someone older or in poor health can be expensive or simply unavailable. A lump sum is only one way to close the same gap — less borrowing, more savings, or a larger survivor's pension do it too. The figures walk the central (expected) path, not an unlucky one.

Care costs

Later-life care is a fat-tailed risk, so it is modelled explicitly:

In the Monte Carlo, each person has roughly a 1-in-4 chance of a care spell, of typically ~2.5 years (capped at 8), placed at the end of life, at self-funder fees (around £1,300 a week residential, £1,600 nursing, from LaingBuisson). Those fees rise faster than general prices: care is largely National-Living-Wage-pinned staff cost, so the fee escalates at CPI + 2% real (editable per scenario) to the year the spell falls — a late-life spell therefore costs more the later it occurs. The share of futures where care costs the household anything, and the typical and high-end bills it bears, are shown as a risk panel.
Each care year is means-tested under the DHSC charging rules (England), assessed on the resident's own means: full fees while their own capital is above £23,250; capital ignored below £14,250 with a tariff (£1 a week per £250) in between; once below the limit they contribute their income less the Personal Expenses Allowance (£31.80 a week, 2026-27) and the local authority pays the balance. The home is disregarded while a partner still lives in it, and assessed once the resident is alone (or the home is let). A partner's own income and savings are never assessable. The £86,000 lifetime care cap is not modelled — it was cancelled in July 2024.

The central (best-estimate) projection stays care-free — care is a risk, not a certainty, so folding an averaged amount into the central line would mislead. Instead the "What you can afford" screen shows, beside each plan's care-free verdict, an "if significant care is needed" stress: the same expected path with one adverse ~~4-year nursing spell (~~£1,800 a week) placed on the last-surviving partner, means-tested and escalated like any care cost. So a plain "lasts for life" is never shown against a silently care-free path, while the ordering still reflects the expected case.

Simplifications of the stress itself: it uses fixed adverse parameters (not yet user-editable) and applies one spell rather than both partners'; the probabilistic Monte Carlo care model is unchanged. What the care model as a whole does not cover is listed once, under What we don't model below.

How long you might live

Lifespans drive everything, so they are sampled from ONS 2024-based cohort life tables (see mortality data for the source and method). Each partner's age at death is drawn from their one-year mortality probabilities; the central projection uses a representative (median) lifespan instead of sampling. Above age 100 a smooth statistical tail runs to a hard cap of 110.

Not modelled: partners are sampled independently — there is no "broken-heart" correlation between a couple's deaths; ages below 50 are outside the grid.

The Monte Carlo simulation

The simulation runs 1,000 paths for a quick preview and 10,000 for a full run, from a recorded random seed — so the same inputs and seed reproduce byte-identical results. For each path, in order, it draws the two lifespans, any care spells, then a year-by-year path of investment returns and inflation, and runs the full projection.

Investment returns are modelled in real (above-inflation) terms across three asset classes — equities, bonds and cash. Each year draws correlated random returns (mean plus volatility, using the asset correlations). Inflation is drawn alongside them, not separately: it carries a year-to-year memory, so a high-inflation year makes the next one more likely, and it is correlated with each asset class's real return, so prices and markets can go against you in the same year. House-price and salary growth are drawn too, each with its own volatility and its own correlation to equities, so a home's value and a wage do not simply travel up a straight line. The invested pot earns the blend of its holdings; the default is a cautious 40% growth / 60% defensive mix. Every return, volatility and correlation figure carries its own source and the date it was last checked, set out in the economic assumptions.

Not modelled: returns are drawn from a normal distribution, so there are no fat tails or regime shifts; partner deaths are independent.

The historical stress test

Alongside the Monte Carlo, a sequence-of-returns backtest replays each past starting year's actual UK real returns and inflation over your plan, to isolate the risk of a bad run of years early in retirement (which hurts far more than the same years later). It reports what share of historical start years your plan would have survived, and names the worst (1929, the 1970s, 2000, 2008). The data is the Jordà–Schularick–Taylor Macrohistory database of UK total returns (1871–2020); start years without enough following data fall back to the expected assumptions, and the backtest uses representative lifespans so only the market path varies.

Putting it together: the year-by-year projection

Each projected year the engine:

Adds up every living person's income by source — part-year earnings in a retirement year, DB pension, State Pension, income streams, planned pension withdrawals, annuities, survivor pensions. The retirement year is split on both sides. Pay stops on the birthday you retire, and the income replacing it starts on its own date rather than on 1 January. The year is split in whole months, at the end of the month the date falls in: pay counts up to and including that month, and the new income counts only the months after it. So a State Pension first paid in November counts one month (December) in that year. The same rule applies to a DB pension at its normal retirement age, an annuity from the birthday it is bought on (or its deferred start), and an income stream from its start age. An income stream whose start age you have already reached in the first projected year is taken as already being paid, and counts the whole year.
Taxes each person individually (income tax in the combined pass above, plus employee NI on earnings). National Insurance stops at State Pension age, so the earnings of the months before that date in the year you reach it are still charged. Unwrapped savings interest and GIA dividends are taxed each year while the asset itself grows at capital only; ISAs are tax-free.
Credits any Pension Credit.
Meets the household's spending — an essential floor plus discretionary spending, with costs that switch off when they should (mortgage costs when the mortgage ends, property costs when the home is sold, commuting when work stops), a survivor spending factor, one-off costs, rent and care.
Handles mortgage-maturity events (refinance, repay from capital, or a forced sale in place), and models the three shapes a mortgage can take: interest-only (the balance stays level and the interest is a spending line), equity-release roll-up (the balance compounds unpaid and is repaid from the estate, capped at the home's value), and repayment / capital & interest (the balance amortises to zero over the term and the instalment stops when it does). Whichever shape it takes, a mortgage payment is treated as what it is — fixed in cash terms, so it costs less in real money each year, and unchanged when one partner dies, unlike ordinary household spending. That is the payment only. Any other line that runs only while the mortgage does (a protection premium, a fee, a voluntary overpayment) stops with it, but until then rises with prices and stays in the tier it was entered in.
Funds any shortfall by the chosen withdrawal strategy, grossing pension withdrawals up for tax; invests any surplus.
Charges capital gains tax on any GIA disposal, and on death passes the estate to the survivor (so money is never stranded and mis-read as "running out").
Grows the balances into the next year, then takes the year's investment charges out of them — the platform/administration fee plus the funds' own ongoing charges, on pensions, ISAs and investment accounts (cash deposits pay none). The growth rates in the assumptions panel are before charges, which is how they are published, so the charge is deducted separately and shown as its own pounds figure beneath the cashflow table rather than folded into a smaller growth line.

Fiscal drag is built in: tax thresholds are frozen to April 2031 (as legislated), then indexed with inflation. Withdrawal strategies range from the default tax-efficient order (non-pension first, pension last) to "fill your tax-free bands first", and adapt when a household is on Pension Credit (drawing capital first to avoid clawing the credit back).

The tool also compares your spending against the PLSA Retirement Living Standards (Minimum / Moderate / Comfortable) as a neutral yardstick — it reports which standard your spending reaches, and never says whether that is "enough".

What we don't model

An honest list of the current limits (each is flagged in the code):

Region: England, Wales and Northern Ireland only. Scotland is refused; Welsh LTT is treated as SDLT.
Emergency tax reflects the size of the over-deduction, not HMRC's PAYE tables to the pound.
SDLT surcharge is not applied to a replacement main residence; no first-time-buyer relief.
Capital gains: no lettings relief; one rate per owner; capital losses are not relieved; the CGT band is judged on non-savings income. Deemed-occupation absences are modelled, but we do not check that no other home was eligible for relief while you were away, and job-related accommodation is not distinguished. Selling personal possessions (art, jewellery, antiques) is charged where you tell us what the item cost, using the per-item exempt amount and the marginal relief just above it; the restricted-loss rule and the exemption for wasting assets (anything with an expected life of 50 years or less, such as a car) are not modelled, so leave the cost blank for those.
Inheritance tax (when the toggle is on) values the estate at each death inside the projection and applies the headline bands (relationship-status aware), but not the fuller estate: no lifetime gifts or 7-year taper, trusts, business/agricultural relief, the 36% charity rate, non-descendant beneficiaries, or the inherited-pension income-tax interaction. The headline panel values deaths at representative ages; a completed Monte Carlo run additionally shows the spread of IHT across futures.
Care: the lifetime care probability is sex-differentiated (women's chance is materially higher, anchored to the Dilnot/PSSRU ~1 in 4 population mean); age-conditioning of the onset rate and a sex split of the duration remain flagged refinements; Monte Carlo only. The in-path means test counts a resident's share of any Pension Credit into their contribution, but it does not re-assess a couple as two single people when one moves into permanent care, assumes the local authority pays at the same fee rate as a self-funder (so no third-party top-up is charged), and skips deferred-payment / 12-week-disregard mechanics (below the annual grid).
Mortgages: no lender or broker fees, arrangement/redemption charges or early-repayment charges are modelled anywhere, so the cost of taking or leaving a deal is understated. On a repayment mortgage there is no input for overpayments (the overpayment input applies only to an equity-release roll-up), and a variable reversion rate is modelled as a single fixed rate for the rest of the term rather than a rate that moves.
Benefits: Guarantee Credit only (no Savings Credit); the severe-disability addition follows the couple-eligibility rule (both partners must be on a qualifying disability benefit); the carer addition is supported by the engine but not yet exposed as a builder input.
Monte Carlo: normal returns (no fat tails); independent partner deaths; the historical data ends in 2020.
Pensions: the lump-sum-allowance cap on a DB commutation lump sum is not enforced; one smooth inflation proxy for DB escalation. Contribution tax relief is modelled for "net pay" schemes (the contribution comes off gross salary, so relief lands at your own tax rate straight away, and National Insurance is unaffected) and for the £3,600 non-earner route (you pay £2,880 from savings and the provider adds £720, until you turn 75); general relief at source is not modelled yet and is refused rather than quietly treated as no relief, and salary sacrifice is not modelled at all. If a pension's relief method is left unset we model no relief, and say so on your results. Contributions are measured against the annual allowance (£60,000 a year, counting your employer's payments too), or the £10,000 MPAA once you have flexibly accessed a pension. Nothing above the allowance is refused: it is paid in, and an annual allowance charge falls on the excess at your marginal rate, shown on your results in the year it falls. In the year you first take money flexibly we measure the whole of that year's payments against the MPAA, because the forecast has no date inside a year; in life only what goes in after the draw is, so the charge we show for that one year can be higher than the real one. We do not carry forward unused allowance from earlier years, and we do not taper the allowance for very high earners.
Investment charges: one household-wide rate covers the platform/administration fee and the funds' ongoing charges together, taken from pensions, ISAs and investment accounts each year (cash deposits bear none). There is no per-account or per-pot rate, so a cheap ISA and an expensive old pension are charged the same; and the rate does not include the cost of financial advice, which an advised household pays on top.
Other: a one-off cost triggers on the first-listed person's age only; a person under 50 is out of the mortality grid.

How we check it

The deterministic engine reproduces HMRC's published worked examples to the penny.
Reconciliation checks assert that totals equal the sum of their parts, and completeness checks that every input that should count actually reaches the result (no income stream silently dropped).
The engine is framework-free and holds no live clock or database, which is what makes the worked-example tests trustworthy. Statutory figures are versioned per tax year with a source and a verified-on date, and freshness commands flag any figure or mortality dataset that has aged.

This is guidance, not advice

RetireForecast is an education and guidance tool. It illustrates the consequences of the figures and assumptions you enter. It does not tell you what to do, does not make a personal recommendation, and is not regulated financial advice. Pension and housing decisions are significant and hard to reverse — for free, impartial guidance see Pension Wise and MoneyHelper, or speak to an FCA-regulated adviser.

Sources

Statutory rates and thresholds are taken from gov.uk (income tax, National Insurance, dividends, stamp duty, capital gains tax, Pension Credit, inheritance tax, care financial assessment), each stamped with the date it was last verified. The modelling assumptions and their sources are set out in full in the companion pages:

Economic assumptions — returns, volatility, inflation, house/rent/salary growth (FCA projection rates; Dimson-Marsh-Staunton / Barclays historical volatilities; OBR/ONS).
Mortality data — ONS 2024-based cohort life tables.
Historical stress-test data — the Jordà–Schularick–Taylor Macrohistory database (UK, 1871–2020).