How projections are calculated
Each year, the model grows your assets by their assumed growth rate, applies your income and expenses, calculates Income Tax and National Insurance, then carries the resulting balance forward to the next year. Contributions are applied after growth (end-of-year basis). All figures are shown in nominal terms unless stated otherwise.
Tax is modelled as the year's total tax liability (Income Tax, National Insurance and, where investment sales are modelled, capital gains tax) charged in the year it arises. Hundred Summers does not model the timing of collection: there is no separate withholding step, and no following-year refund or self-assessment true-up. This mirrors how PAYE broadly works for UK employees, and keeps every chart (including the cashflow map) showing the same single tax figure. Tools built around other tax systems sometimes chart withholding and refunds as separate flows instead; if you owe or reclaim tax across year boundaries in real life, the annual totals here will still be right, but the timing may differ slightly.
The stress test runs a Monte Carlo simulation - many randomised return paths - to estimate how often your plan would succeed under varying market conditions. The success rate is the proportion of simulated paths in which your money lasts to your plan-end age. It is an illustration of sequencing risk, not a guarantee.
How the stress test varies the base assumptions. Your ordinary projection uses a single fixed growth rate for each asset. The Monte Carlo stress test keeps each asset's expected return exactly where you set it (each pension pot's random yearly return is centred on its own risk-profile or growth-rate figure, each property on its own growth rate, and savings accounts on theirs), but scatters every year's draw by a volatility (standard deviation) reflecting how bumpy that asset really is: 16% for equities, 6% for bonds and 8% for property, with each pension pot's volatility scaled by its equity share (a cautious pot swings less than an adventurous one). Because each year compounds into the next, the order in which good and bad years fall matters - an early loss in retirement hurts more than the same loss later. Running the whole plan hundreds of times captures this sequence-of-returns risk.
The market model behind the draws. Each simulated year is one correlated market event, not a set of independent coin flips. Equities, bonds, property and inflation are drawn together (equity–bond correlation 0.1, equity–property 0.5, bond–inflation -0.3), so the simulation can produce years like 1974 or 2022: shares and bonds down with inflation up, together. Every holding shares the same market year by its asset composition: an equity ISA falls in the same year as an equity-heavy pension (the wrappers differ; the market doesn't), and cash accounts take no market shock at all. Return draws are Student-t distributed (8 degrees of freedom) rather than normal, so deep-loss years arrive at roughly their historical frequency instead of being ruled out by thin tails. Inflation is persistent: a high year tends to be followed by another (autocorrelation 0.7, long-run spread 1.5%), and the simulated inflation path scales your living costs against the plan's expected inflation, so a multi-year cost-of-living squeeze is a scenario the test genuinely runs. The State Pension tracks the simulated inflation too (the triple lock guarantees at least CPI); DB indexation and annuity escalation keep your plan's own assumptions, a stated simplification. Tax and NI thresholds also keep your assumed indexation path rather than following each simulated inflation path: in a high-inflation simulation your spending rises but the tax bands don't keep pace, so those paths carry extra fiscal drag; mildly conservative, and arguably how governments actually behave in inflationary episodes. This model is validated against ~a century of market history in the test suite: its structure reproduces the record's inflation persistence, asset co-movement and tail weight, and in a classic retirement-drawdown backtest it sits deliberately on the conservative side of the realised record; real markets historically mean-reverted over long horizons, and this model does not assume that rescue will repeat. The structure changes which plans fail and when more than it changes the headline rate: measured against an independent-normal model on the built-in profiles, the success rate moved by only a few points in either direction.
The growth rate means the same thing in both views. Your growth rate is a compound rate: the central projection applies it every year. A random yearly return whose plain average is that rate compounds to less, because losses and gains do not cancel (a 20% fall needs a 25% rise to recover); left uncorrected, the median simulated path on a 50-year plan finished near a third of the projection, so the success rate judged the plan against a quietly worse central case. Each holding's simulated return is therefore centred half its own variance above its growth rate (≈1.3% a year for an all-equity holding, 0.18% for bonds, nothing for cash), the standard lognormal correction, so a holding's median simulated path reproduces the projection and the fan spreads around the plan rather than below it. The scatter of the paths is unchanged; only the centre moves.
Charges come off before the simulation varies anything. Each holding's rate is first taken after its charges: with growth rates entered before charges (the default), its charge is subtracted; entered after charges, the rate is used as it stands. The simulated returns are then centred on that after-charges rate, so the two settings give the same stress test for the same after-charges figure: entering 4% after 1.7% charges, or 5.7% before them, is one plan. This plan's rates are entered before charges (Plan assumptions, above).
Same plan, same answer: the simulation is deterministic. The random market paths are drawn from a fixed seed, so every visit to the stress test replays the SAME set of simulated market histories. The success rate therefore never wobbles between runs: it moves only when your plan actually changes, and a change that cannot affect the cashflow test (renaming a profile, adding a debt-free property that is never drawn on) moves nothing. What varies between plans is how your money fares along those same market years, which is exactly the comparison a stress test should make. It also means the success rate, the fan chart and the per-year spending likelihood are counted from the same simulations and can never disagree with each other.
The margin of error on the success rate. A fixed seed makes the figure repeatable, not exact. It is still counted from a sample of market paths, and a different sample of the same size would give a slightly different answer. So every success rate is shown with its 95% margin of error: the range the model's own figure would likely fall in if it could run every possible path. With 500 paths a plan showing 92% has a margin of error of 89–95%. The range is the Wilson score interval, which stays honest at the extremes (500 successes out of 500 gives 100% with a range of 99–100%, not a false certainty). It is rounded outwards to whole percentages, so it is never shown narrower than it is. It measures sampling error and nothing else. Whether the assumptions behind the paths are right (returns, volatility, inflation, how long you live) is a separate and usually larger uncertainty, which the Outcome range below addresses. Where two plans are compared (What-If, saved scenarios), they run through the same market paths, so the difference between them is more reliable than either margin of error suggests.
Every simulated path runs through the full plan engine. Each simulated market path is fed through exactly the same projection as your main plan: the same income tax and National Insurance, pension rules and tax-free cash, capital gains tax, savings-interest and dividend tax, ISA allowances, drawdown strategy (including fill-and-recycle), events, care costs and property decisions. Only the market years change. So the stress test and the projection cannot disagree about how your plan works. With market swings switched off, every simulated path reproduces the projection to the penny; the automated tests check that, check that any single path replayed on its own gives exactly the same result, and check that the plan's books close to within £1 in every year of every path. Living costs and the State Pension follow the simulated inflation; DB pension indexation and annuity escalation follow your plan's own assumptions, as noted above. The same is true of every other simulation in the app: the couples stress test, the historical backtest, the 85%-confidence option in "What would it take?" and the saved-scenario comparison all run their paths through this engine. (Until September 2026 they used a faster, simplified copy of the plan's rules that left out capital gains tax, savings-interest and dividend tax and the fill-and-recycle step.) Each path is a full run of your plan, so the simulations run in the background after an edit. The confidence answers stop each trial as soon as its verdict is decided (once enough simulated markets have succeeded or failed to settle it), which gives exactly the answer a full run would, faster.
Two different ways of being wrong: the stress test vs the Outcome range. The master chart (Report → Chart) offers both, and they answer different questions. The stress test asks: even if my assumptions are right on average, how much do random market sequences scatter the outcome? It is probabilistic, and it applies to net worth only, because that is what the simulation tracks, which is why it appears as the Net Worth series' "stress test" state. The Outcome range asks: what if my assumptions themselves are wrong? It contains no randomness at all; the plan is re-projected through two alternative worlds, a pessimistic one (investment growth 1.5 percentage points lower, property 1 point lower, inflation and care inflation 1 point higher, every year) and an optimistic one with the same shifts reversed, and every drawn dataset is shaded across the envelope of the three worlds (the central plan is included because some figures, such as drawdown tax, do not move smoothly between worlds). No probabilities attach to the Outcome range; it is a scenario illustration. A plan that holds up at the pessimistic edge of the Outcome range and at the stress test's 10th percentile is robust to both ways of being wrong.
The couples stress test (on the Couples tab) runs every partner's plan through the same simulated market years (a shared sequence of good and bad years, as a real household would experience), and counts a simulation as a failure only when the household's combined unmet spending exceeds the partners' combined remaining pots and savings. It is slightly optimistic in assuming money can move freely between partners; year-by-year transfers are not modelled.
How best and worst cases are calculated. The tool does not hand-pick an optimistic or pessimistic set of assumptions. It runs the full plan hundreds of times - each a different random sequence of years - then, at every age, sorts all the resulting net-worth figures and reads off percentiles. The worst-case band is the 10th percentile (only 1 in 10 runs did worse), the median is the 50th percentile, and the best-case band is the 90th percentile (only 1 in 10 did better). The shaded 10th-90th range holds the middle 80% of outcomes, and the inner 25th-75th band the middle half. The spread you see comes entirely from varying the base return and inflation assumptions year by year - not from changing your inputs.
Tax assumptions (2026/27 England rates)
Personal Allowance
£12,570 (tapered above £100,000, gone at £125,140)
Basic rate (20%)
£12,571 - £50,270
Higher rate (40%)
£50,271 - £125,140
Additional rate (45%)
Above £125,140
Employee NI
8% between £12,570 and £50,270, 2% above
Pension annual allowance
£60,000 (editable in Plan assumptions above), tapered by £1 for every £2 of adjusted income over £260,000 (down to a £10,000 floor); only applies if threshold income exceeds £200,000. Contributions above the allowance receive no tax relief in the model: in effect the annual allowance charge, which claws back the excess relief at your marginal rate; scheme-pays elections and the exact charge calculation are not modelled separately. Carry-forward: unused allowance from the previous three tax years (entered under Your Facts → Pensions) tops up the current year, consumed oldest-first; the projection then rolls its own three-year window forward. MPAA: from the year after the plan's first taxed pension withdrawal (UFPLS taxable element or crystallised drawdown income; a PCLS lump sum or annuity purchase alone does not trigger it), the allowance is capped at the £10,000 Money Purchase Annual Allowance. This bites when a role windowed past retirement keeps contributing while the pension is drawn; carry-forward stops applying and accruing from that point (the alternative-allowance split is not modelled). DB schemes: each pension can carry a survivor percentage (used by the Couples survivor analysis), a bridging income paid until State Pension age, and a commutation: a tax-free lump sum at the start age (consuming Lump Sum Allowance) in exchange for income reduced by lump ÷ commutation factor, capped at the permitted maximum of 25% of the benefits' value (20 × the pension left + the lump), which is £85,714 on a £20,000 pension at 12:1; a larger lump entered is paid only up to it. Employee NI is charged on earnings until State Pension age (not your own retirement age); income tax and student-loan repayments follow earnings wherever they fall. Interest on a student loan is not added to its balance, so the balance falls faster than a statement shows and repayments may stop sooner (the optimistic direction; the Liabilities tab says so beside it). Self-employed earnings (Income tab flag) pay Class 4 NI on profits (6% between £12,570 and £50,270, 2% above) instead of employee Class 1; the flag covers all earned income. Child benefit (children + youngest birth year on the Income tab) arrives as tax-free income until the youngest turns 18, reduced by the High Income Child Benefit Charge between £60,000 and £80,000 of income; modelled through working years. The marriage allowance (recipient side) applies as the ~£252/yr credit while no higher-rate tax is due.
Capital Gains Tax
18% (within basic-rate band) / 24% (above), after the £3,000 annual exempt amount. Gains sit on top of taxable income, so the 18% band left is £37,700 less your income above the personal allowance: an unused personal allowance cannot be set against gains. Applied to gains realised when drawing from General Investment Accounts. ISAs, cash and pensions are exempt.
Retirement withdrawals
Each retirement year's withdrawals are taken at the START of the year, before that year's investment growth: from pensions and from savings alike. The money you draw to spend in a year therefore earns no growth that year. In reality people usually draw monthly, so the average pound withdrawn would earn about half a year's growth; taking it at the start is the cautious side of that, and it is the same rule for every source, so no drawdown order is flattered by a timing difference. (Until 28 September 2026 savings were drawn after the year's growth and pensions before it; moving savings to the start lowered plans that draw heavily on savings: the demo plan's net worth at 92 by about 8.6%.)
Fill & recycle drawdown
An optional drawdown strategy (Pension Income tab): band-fill behaviour, plus in any retirement year whose taxable income sits below the higher-rate threshold, extra pension is drawn up to the top of the basic-rate band beyond spending needs, and the after-tax surplus is swept into savings; ISA first, within the annual ISA allowance, overflowing to a general investment account. The aim is to use cheap tax years (typically before State Pension and DB income begin), so later withdrawals avoid higher rates. The projection charges the extra tax in the year of the draw, loses the pension's tax-free compounding on the amount moved, and taxes subsequent GIA growth, so the outcome is shown honestly and is not always a win. This recycles taxable drawdown, not tax-free cash, so HMRC's PCLS-recycling rules are not engaged. Off by default.
Spending while working (What If)
The What If tab's spending-while-working lever adds to or takes from the single annual living-expenses figure for your working years, in today's money, rising with inflation; Apply writes the new figure to the Expenses tab. It is not available while your Expenses are itemised: an itemised budget is changed by adding or editing an expense line, never scaled. A separate extra-spending figure the lever set briefly on 13 October 2026 still counts in any plan that carries it, and the Expenses tab offers to move it into your expenses.
Salary sacrifice
When switched on (Tax tab), the sacrificed amount reduces the salary on which income tax and employee NI are computed and flows into your pension as an employer contribution throughout your working years. It counts toward the annual allowance but is not double-relieved against taxable income. Sacrifice replaces your own employee contributions £-for-£ up to their total, the conversion real arrangements make, so pensions receive max(your contributions, sacrifice) plus employer money, never both in full; sacrifice beyond your own contributions is additional money in. Relative to an ordinary employee contribution the incremental benefit is the NI, since income-tax relief applies to both routes. No employer NI pass-on is assumed. From April 2029 (2029/30 in the projection) only the first £2,000 a year of sacrificed salary stays free of employee NI, as announced in the November 2025 Budget (HM Treasury guidance, 26 November 2025; HMRC policy paper, 4 December 2025): the rest is added back to the pay NI is charged on, while income tax relief is unchanged. The £2,000 is held fixed, as no indexation has been announced. Your employer's own NI, and whether they pass any saving on, are not modelled.
ISA allowance
£20,000 a year (editable in Plan assumptions above), enforced across every route into ISA accounts: contributions, surplus, windfalls, transfers, sale proceeds and reinvested lump sums share one annual cap. Anything above it is diverted into a General Investment Account, where later gains attract CGT. ISA-to-ISA transfers do not use up the allowance. From 6 April 2027 cash ISAs have their own limit inside it: £12,000 a year for anyone 64 or under at the end of the tax year, the full allowance at 65 and over (ISA (Amendment) Regulations 2026). The projection tests the age reached in the year, which for a January–April birthday can apply the limit one year longer than the law does: the cautious side. Cash a cash ISA cannot take stays as cash, in a taxable cash account, rather than being invested for you; stocks and shares ISAs and Lifetime ISAs are unaffected, so the other £8,000 can still go into one. The 22% charge on interest from cash held inside a stocks and shares ISA is not modelled: each account holds one asset class, so a cash-class ISA is treated as a cash ISA.
Regular savings contributions
A contribution you set on a savings or investment account is paid only from the year's surplus (what is left after spending, tax, National Insurance and pension contributions), and only while you are working. If the surplus is smaller than the contributions you have set, the accounts are paid in the order they are listed in Your Facts → Assets & Debt: the first in full, the next with whatever remains, and so on, so a later account can receive part or none of its contribution. Nothing is drawn from savings to make a contribution up, so an unpaid contribution never shows as unfunded spending: it simply does not happen, and the projection already reflects that. The Savings & Investments tab shows what each account was actually paid over the plan against what was set, and names the first year the surplus falls short. Pension contributions are different: they are treated as spending, so a year that cannot afford them draws on savings.
Use my ISA allowance first
An optional setting (Savings tab, and a What-If lever). Off, each year's leftover surplus banks in your cash hub. On, whatever is left of that year's £20,000 allowance is filled from the surplus first, and only the remainder goes to cash: the allowance cannot be carried forward, so a year that passes unused is gone for good. It fills the allowance your own contributions left: only the headroom is sent, so nothing spills into a General Investment Account that would otherwise have stayed as cash. The money goes to your largest invested ISA, or to a cash ISA if that is all you hold (from April 2027, if you are under 65, only up to the £12,000 Cash ISA limit: the rest stays in cash); if you hold no ISA at all the projection adds one, labelled so you can see it, growing at the rate of your own invested money. A Lifetime ISA is never chosen: it takes £4,000 a year, stops at 50, and money out before 60 carries a 25% charge, though its own contributions still consume the shared allowance. This is a modelling switch, not a recommendation: whether sheltering spare cash suits you depends on what the cash is for.
Move existing savings into the ISA allowance
A second optional setting (Savings tab, and a What-If lever). The switch above can only shelter money a year has left over, which is no help to a plan running no surplus: the commonest retired shape, and often the one with the most sitting unwrapped. It runs last in the year, after every draw on your savings has been made: the retirement drawdown, any working-year shortfall and any bond-gain bill, so it can only shelter what the year did not need. With it on, whatever allowance the surplus leaves is then filled from savings you already hold: cash first, because moving it costs nothing (interest is taxed as it arises, so there is no gain to crystallise), then unwrapped investments, which is the move usually called a Bed & ISA. It fills the same ISA the switch above would choose, so where that is a cash ISA, an under-65 moves at most £12,000 a year from April 2027. Holdings are sold cheapest-gain-first, so the tax bill is the smallest that fills the allowance. Your current account is sold from too: in this projection it is the cash hub where every year's leftover surplus lands, so on a plan that saves hard it is not a transactional buffer but the largest unwrapped holding you have; excluding it would leave the allowance expiring beside half a million pounds of taxable cash. It is safe to sell because this runs after the year's spending has already been funded, and because an ISA is instant-access too, so the money stays just as reachable; among equally cheap money the hub is drained last. Investment bonds, by contrast, are never sold, whose withdrawals are chargeable events taxed as income under the 5% rule rather than CGT disposals; modelling those as a capital disposal would understate the tax. Selling an unwrapped investment IS a disposal: the gain is assessed with any other gains that year, sharing the £3,000 annual exempt amount quoted above, and the resulting CGT is met out of the sale proceeds, so slightly more is sold than is subscribed, and the setting can never push a plan into unfunded spending. Because opening balances are entered without a cost basis, holdings start with no embedded gain and the tax grows as the projection does. This is a modelling switch, not a recommendation, and it is genuinely not always worth doing: it brings a CGT bill forward to stop future dividend and interest tax, and which way that lands depends on your numbers; compare the net worth at plan end with it on and off. The stress test mirrors the move but, carrying no income bands, charges the disposal at the 24% higher rate and does not model the dividend and interest tax the move avoids, so it shows the cost without the benefit; read the trade-off in the projection.
Savings & dividend tax
Interest on non-ISA cash accounts (Premium Bonds are their own account type, whose prizes are tax-free), and on bond/cash holdings in a GIA is taxed at 20/40/45% above the Personal Savings Allowance (£1,000 basic-rate / £500 higher-rate / nil additional-rate), rising to 22/42/47% from April 2027 per the Autumn Budget 2025. Dividends on GIA equity (2% yield assumed; mixed funds 1% dividends + 1.5% interest, capped by the account's growth rate) are taxed at 10.75/35.75/39.35% (the April 2026 rates) above the £500 dividend allowance. UK-wide rates apply even under the Scottish regime, as in law. The income itself stays reinvested inside each account's growth rate: only the tax is charged to your cashflow. The £5,000 starting rate for savings is applied: interest just above your personal allowance is taxed at 0%, and every £1 of other income above the allowance reduces it by £1, so it is gone once other income reaches £17,570 (the £5,000 is held at its cash value, like the savings allowance). The order follows HMRC: the personal allowance first, then the starting rate, then the Personal Savings Allowance; income the personal allowance or starting rate already covers does not use up the savings or dividend allowance.
Financial independence and crossover points
The Goals tab's timeline shows each financial independence milestone at the age the projection reaches it (Lean, Regular and Fat FI at 25 times a spending level, the 4% rule; Coast FI; Barista FI, tested with the full projection), each with its own explanation, and three crossover points, each likewise its own entry at the age it arrives, read from the projection's own years, for pensions and savings and investments only (a home is not drawn on). Your money out-earns you: the first year their growth after charges is at least your pay. Compounding takes over: the first year that growth is at least what goes in (your and your employer's pension contributions and what you save). Self-sustaining: the first year the growth after inflation, on that year's opening balance, covers the larger of your living spending and what the plan actually takes out of your pensions and savings that year (gross, so the income tax on pension withdrawals counts, and net of any surplus swept back in); on these assumptions what is invested then doesn't run down, and the goal names the first later year it stops holding, if any (care costs, a large one-off). Figures in today's money. The minimum pension age is not part of these tests; the plan's own funding test is the stricter one.
Search the combinations
On the Pension Income tab, on request: every combination of four drawdown orders (savings first, pensions first, fill the basic-rate band, fill and recycle), two ways of taking the tax-free quarter (a quarter of each withdrawal, or a lump sum at retirement or the plan's tax-free cash age) and three annuity shares (none, half, all; with all of it in an annuity there is no drawdown order to choose), up to 18 in all, each run through the full projection with everything else in the plan unchanged. Each shows the most a year the plan could spend in retirement with every year covered (the central projection's spending solver), whether the planned spending is covered, what is left for heirs (net worth at the end of the plan less inheritance tax and, when the plan ends at 75 or over, 20% income tax on pensions left, as inherited pensions are then the beneficiaries' taxable income and their rate is unknown) and lifetime tax, all in today's money. The rows can be sorted by the most you could spend, the most left for heirs or the least lifetime tax, and the best figure in each column is marked; heirs and tax are compared only among combinations that cover the planned spending, since one that runs short spends, and so pays tax on, less. Each column is one measure: sorting or marking does not choose a combination. The search does not value an annuity's lifelong guarantee or drawdown's flexibility, which only the stress test shows; each row opens in What If, where it can be stress-tested. The AI Copilot can run the same search.
Band-fill drawdown (optional)
When selected on the Pension Income tab, each retirement year draws taxable pension income only up to the top of the basic-rate band (given that year's other taxable income), meeting the rest from savings/ISAs tax-free; pension above the band is used only if savings run out.
Premium Bonds
An account type of its own (NS&I). Prizes are tax-free, so a Premium Bonds account adds nothing to the savings tax. Up to £50,000 can be held per person: anything that would take the holding over the limit (a contribution, leftover surplus, or prizes reinvested once it is full) is kept as ordinary taxable cash instead, as it would be paid out in reality. The account's growth rate stands in for the prize fund rate, which is an average across all bondholders; your own prizes will vary. It is never used as the cash hub that collects each year's surplus. The type decides the tax, not the name: an account set as taxable cash is taxed as cash even if it is called Premium Bonds, and the Assets tab offers to switch it.
Tax wrappers
ISA/LISA growth and withdrawals are tax-free. Lifetime ISA: contributions capped at £4,000/yr (using part of the £20,000 overall ISA allowance), and only until 50; the 25% government bonus is added to what goes in; the balance is locked until 60, so the drawdown order skips it before then. Onshore and offshore bonds roll growth up inside the wrapper and are taxed on chargeable-event gains (see Investment bonds, below). EIS/VCT holdings: 30% income-tax relief on new subscriptions (assumed actually invested, capped at the tax otherwise due), tax-free dividends and no CGT on disposal; the EIS three-year holding rule, loss relief and business-relief IHT treatment are not modelled; enter qualifying business relief on the IHT tab.
Risk questionnaires
Alongside attitude to risk and capacity for loss, the Risk tab carries a Knowledge & Experience self-assessment (six questions, banded Basic / Informed / Experienced) and a Sustainable Investing preference questionnaire (five questions, banded from no particular preference to strong preference). Both are educational documentation, not regulated appropriateness or suitability tests, and change no projection numbers; scores and bands appear on the Excel Risk Profile sheet. Each questionnaire also flags internally inconsistent answers - two responses at least three points apart on the 1-5 scale once at least half the questions are answered - as a prompt to re-read, not a verdict.
Income floor
The Pension Income tab's floor card compares the essential share of retirement spending (items not marked nice-to-have; the whole budget when none are marked) against guaranteed lifetime income - DB, State Pension and annuities - net of the tax due on that income alone, both measured in the first year everything is in payment. The purchase that would close a gap is quoted at the market annuity rate for the purchase age with the user's own annuity options, grossed up for basic-rate tax (a stated simplification).
Stress battery
The Stress Test tab's named battery re-runs the plan through the full deterministic engine under specific shocks: a 25% equity-weighted market fall today, an unexpected £20,000 bill next year (an ordinary expense event, funded from savings or flagged unfunded), redundancy next year (twelve months with no earnings, then the same salary again; no income protection, which does not cover redundancy), all growth 2pp lower, inflation 2pp higher, stopping work three years early, living to 100, and paid care from five years before the Later Life age. Each is judged twice - the full budget, and essentials only (nice-to-have items given up first). Deterministic what-ifs on the plan's own assumptions; the Monte Carlo covers random market paths.
Per-year spending likelihood
The Stress Test tab reports, for every plan year, the share of simulated market paths in which that year's spending could be funded: the full budget and the essential budget (retirement items not marked nice-to-have) separately. The pass mark shown is essentials met in at least 95% of paths in every year. Counted from the same fixed-seed simulation as the success rate, so the two can never disagree.
Historical backtest
The Stress Test tab also replays the plan through every rolling window of annual market history from 1928 to 2023 (S&P 500 and US 10-year Treasury total returns from Aswath Damodaran's long-run dataset, US CPI from the BLS): a US series used as a developed-market proxy, because UK data of that length is licensed; the stylised facts that matter (crashes, persistent inflation, co-movement) hold in both markets. Each window contributes its year-by-year DEVIATIONS from the record's own averages, layered onto the plan's own growth and inflation assumptions (history's shape, not its level), so the backtest tests sequence risk while the user's assumptions stay authoritative; because assumed returns typically sit below the record's realised average, the backtest can fail more windows than raw history did. In replay mode earned income, other income, DB pensions AND the tax and NI thresholds track the cumulative inflation surprise exactly as spending does (wages, DB indexation and tax bands broadly followed prices through the great inflations; without this the inflation replay dominates every result), while level annuities stay level. The record has no property series, so property carries only the inflation surprise. It is the single path history took, not a distribution: the Monte Carlo remains the primary stress test. Two readings are built on it. "The most you could spend that history never broke" searches for the highest yearly base retirement spending (today's money; Freedom-years extra and care stay on top as planned) at which every complete window survives, and names the start year that sets the limit; being set by the single worst sequence, it sits below the central projection's figure. "Market episodes from history" replays seven remembered stretches (1929, 1937, 1966, 1973, 2000, 2008, 2022) from the plan's first year of retirement, or this year once retired, because a bad sequence does most damage early in drawdown; before the episode and after the record ends each year runs on the plan's own assumptions exactly, with no volatility adjustment, so the no-shock card reproduces the main projection to the pound. Episode net worth is shown in today's money at that episode's own inflation and includes the home.
Life goals
Goals that are not purchases (Your Facts → Life goals) are recorded, not modelled: a goal adds no income, spending or tax to the projection. Each is placed at the age you choose and in the stage of life that age falls in, on the same boundaries the plan uses for retirement spending (Freedom until the end of your active years, 75 by default; Contentment until 83 by default; Later life after). A goal you give a cost becomes an ordinary expense event, and its funded or at-risk verdict is that event's own. The reflection questions are adapted from George Kinder's life-planning approach; the answers are not stored, only the goals you choose to add.
Goal-seek solvers
The Goals tab's “What would it take?” answers are solved on the full engine by binary search: most sustainable spending, earliest affordable retirement age, and, on each row of the When could I retire? table that runs short, the annual redirection that would make it last (spending less until then sweeps into savings automatically; the pension route adds the freed amount to the largest pension as a contribution, so tax relief usually makes it the smaller number). The AI Copilot can also judge the earliest age and the most a year at ≥85% simulation confidence: the fixed-seed Monte Carlo must succeed in at least 85% of simulated market paths, which is stricter than the single central projection. On the Goals tab, the When could I retire? table gives each age's stress-test success rate instead. Asset headroom is the largest fraction by which pots and savings could fall today with every year still funded (property excluded, not drawable wealth in this model).
When could I retire? (age table)
On the Goals tab, the same plan re-run at every retirement age up to five years either side of the planned one (never today or earlier, never at the plan's horizon; not offered once the planned age has passed). Only the retirement age changes: the planned spending, pensions, savings and every assumption stay as they are. Each row shows whether the planned spending lasts on the central projection, the stress-test success rate at the plan's own number of simulated markets with its 95% margin of error (so the planned age's row matches the Stress Test tab exactly), and the most a year the plan could spend in retirement, in today's money, while still funding every year on the central projection. Each row is a full stress test, so the table runs in the background and fills in row by row. It shows how the date moves the plan; it does not recommend one.
Today and the first projected year
The projection starts from your balances today, and its first projected year is next calendar year: the rest of this year's salary, spending and other ongoing flows are not added on top of today's balances. One-off items dated this year are different (a windfall, a one-off expense, a final-salary lump sum paid at your current age, a downsizing or equity release timed now, and this year's instance of anything that repeats less than yearly), and they are counted in the first projected year, so nothing you date this year is lost. A yearly item starting this year begins in the first projected year, like your salary. (Until September 2026 one-offs dated this year were computed and then discarded with today's snapshot, so they had no effect at all.)
The planning guide's starting figures
The planning guide sets up a plan from a short conversation, typed or spoken; it understands each answer on your device and reads every figure back. What it does not ask, it fills with stated starting figures, all shown on its checking list and changeable under Your Facts: spending while working is 55% of earnings (as in the setup form); one pension pot holds the total you give, on a moderate risk profile; savings outside an ISA are mixed investments; a mortgage is a repayment mortgage at an assumed 4.5%. Where you say you're not sure, it uses, and says it uses: 5% from you and 3% from your employer into a workplace pension (the usual minimum), the full new State Pension, retirement spending at Pensions UK's moderate Retirement Living Standard for one person, savings counted outside an ISA, and a 10% deposit (or, when you're moving, your current home's equity today). A home you plan to buy is priced in today's money and inflated to the year you buy it at the plan's inflation; when you're moving, your current home is sold that year and its mortgage repaid from the proceeds before the purchase. Run on a plan that already holds figures, the guide starts from them, asks only what is missing, and changes only what you tell it; the starting figures above apply only to what it fills in for the first time.
Expense priorities & recurring events
Financial events can repeat every N years (a car every 5, school fees each year of a course). Retirement expense items can be marked essential or nice-to-have; with the opt-in priority-tiers toggle (Expenses tab), a year the money cannot fully fund first gives up nice-to-have items: the reported shortfall then measures unfunded ESSENTIAL spending. Off (the default), the whole budget is one must-fund figure, exactly as before.
Pension access age
Money in a defined-contribution pension cannot be drawn, taken as a lump sum, exchanged for an annuity or used to pay an event before the normal minimum pension age: 55 today, 57 from 6 April 2028 (Finance Act 2022), applied by the tax year the draw would fall in. This holds whatever retirement age the plan chooses: an earlier retirement is funded from ISAs, other savings and any other income until then, and a year those cannot cover is shown as unfunded, never as a pension draw. The protected pension ages a few older schemes carry are not modelled. When the plan retires before the access age, the Pension Income tab shows the bridge: the savings when work stops, what is drawn before the pension can be, and what is left, or the age the savings run short. The earliest-retirement-age search starts from today's age, not from 55, for the same reason: an earlier age works only when savings carry the years before the pension (until October 2026 the search started at 55).
Pension tax-free cash
Set on the Pension Income tab as a list of withdrawals ('Tax-free cash withdrawals'), and nowhere else. Each takes cash from one pension at an age you choose: the most it can pay tax-free (25% of the part not yet used for tax-free cash, or a protected share or amount if the scheme gives one, the only way above 25%), a share of the pension's projected value at that age (not today's value; the tab shows that value), or an amount in the pounds of that year. A withdrawal can be taken from the minimum pension age whether or not you have retired. 'The most' crystallises the whole pension; a share or an amount crystallises only four times its tax-free cash, so the rest stays uncrystallised and keeps a quarter of each later withdrawal tax-free (partial crystallisation). Within a pension, later withdrawals come from the crystallised part first, which is fully taxable (the cautious order), and new contributions are uncrystallised. Whatever a share or an amount asks above what can be paid tax-free (above 25% of the uncrystallised part, or once the Lump Sum Allowance is used up) is still paid from the pension, and taxed as income at the year's marginal rates; it also triggers the £10,000 Money Purchase Annual Allowance. Withdrawals use the £268,275 Lump Sum Allowance in order, by age and then as listed, and every tax-free part, protected or not, is capped by what remains of it (the cautious reading, as some older protections allow more); the tab shows a running total of the allowance each uses. A pension with no withdrawal pays a quarter of each withdrawal from it tax-free (UFPLS-style). All the cash is deposited into the account you choose; with no account chosen (or one since deleted) it goes into your current account, or your first instant-access cash account, so it always stays in your net worth. Spending or gifting it is a Financial Event. A plan saved before the list reads as the list it amounts to; one still on the older plan-wide lump sum takes it from its pensions together, in proportion, exactly as before, until the list is first edited.
Pensions in the estate (April 2027)
From 6 April 2027 unused pension funds count into the taxable estate (Finance Act 2025). Hundred Summers applies this automatically: the per-year IHT estimate includes remaining pension pots for every projection year from 2027 onward, while a death assumed before then leaves them outside the estate.
Property income rates (April 2027)
From April 2027 property income has its own income tax rates, 2 percentage points above the main rates (22% / 42% / 47%): Autumn Budget 2025. Hundred Summers charges this as a 2% surcharge on the taxed slice of income lines marked as property/rental (the top slice of non-savings income under the reform's ordering rules). Scottish rates on property income are devolved and not changed by this reform, so the surcharge applies outside Scotland only.
IHT nil-rate band
£325,000 (editable in Plan assumptions above), frozen since 2009. Unused NRB transfers to a surviving spouse (entered on the Inheritance tab). PETs within 7 years of death erode the NRB before other assets: after the £3,000 annual gift exemption per gift year (carry-forward of one prior year's unused exemption is not modelled). Taper relief follows the statutory rule: it reduces only tax charged on the slice of gifting that exceeds the available NRB (gifts consumed in date order); a gift fully inside the band gets no relief however old it is. Business & Agricultural Property Relief: for deaths from 6 April 2026 the 100% rate applies to the first £2.5M of combined qualifying property, 50% beyond. Not modelled: a late spouse's unused allowance can transfer (up to £5M combined), which leaves the estimate on the high side. AIM-listed shares are relieved at only 50% and do not use the allowance; everything entered as BPR is treated as full qualifying property, so AIM shares entered there are over-relieved: left as ordinary investments they are taxed in full.
IHT residence nil-rate band
£175,000 (editable in Plan assumptions above) when the main home passes to direct descendants, capped at the property's value and tapered £1 for £2 above a £2m estate, the estate valued BEFORE exemptions and reliefs, so a charity legacy or business relief does not keep it under £2m. Unused RNRB also transfers to a surviving spouse. A charity legacy of at least 10% of the baseline amount (the estate after the nil-rate band, with the legacy added back: HMRC's test) cuts the rate on the rest from 40% to 36%.
All projections are computed in nominal (future) money: salaries, expenses and tax thresholds inflate through time. The toggle in the header can display everything in today's money instead: each year's figures are deflated at the tax-threshold inflation assumption (2.5% unless you change it), which makes distant years directly comparable with today but changes no underlying calculation.
Thresholds are assumed frozen at current levels for the life of the projection unless you change them. We classify each savings/investment account as cash, ISA or a General Investment Account (GIA) - based on its name, or you can set it explicitly - because only GIA gains are subject to CGT. Scottish income tax is fully modelled: set your region to Scotland on the Personal tab and the six Scottish bands (19% starter, 20% basic, 21% intermediate, 42% higher, 45% advanced, 48% top) are applied to non-savings income, while NI and CGT remain UK-wide as in reality. Tax rules change and your own position may differ.
Key growth & inflation assumptions
Plan-wide rates (inflation, care inflation, State Pension growth, tax threshold indexation, property growth, and DB pension indexation) are set in the Plan assumptions section above and apply to all projections. Savings and investment accounts each have their own growth rate (editable in Your Facts → Assets & Debt). Pension pot growth is set by the risk profile you choose for each pot. All rates are nominal (before inflation). Defaults are conservative starting points, not forecasts.
Higher assumed growth rates produce more optimistic projections. Because small changes compound over decades, we encourage you to test lower growth and higher inflation to understand the range of possible outcomes.
Asset class default growth rates
Cash 2.0% | Bonds 4.0% | Equity 6.0% | Mixed/Multi-asset 5.0%: these are only starting points used to pre-fill the growth rate when you tag a savings/investment account with an asset class. They are broad, long-run indicative figures, not forecasts, and you can override them for any account.
Pension risk profiles
Each DC pension pot can be given a risk profile aligned to the five Attitude-to-Risk bands, which sets its growth rate automatically: Cautious (~20% equity, 3.5% growth) | Cautious-Moderate (~40%, 4.3%) | Moderate (~60%, 5.0%) | Moderately Adventurous (~80%, 5.6%) | Adventurous (~100%, 6.5%). Growth rates are a blend of long-run equity (~6%), bond (~4%) and cash (~2%) assumptions weighted by the profile's equity allocation. These are broad indicative figures, not forecasts. A pot's own expected growth can be set beside its profile (Growth, on the Pensions tab): that changes the return only, and the profile still sets the equity share the stress test's ups and downs follow, so expected return and risk are set independently, as they are for savings accounts. The chosen equity allocation is also what the Investment Portfolio & Risk assessment uses to weight each pension's market-risk exposure.
Portfolio risk assessment
The Investment Portfolio & Risk figure is a value-weighted equity-equivalent exposure across your savings/investment accounts (by asset class) and your DC pensions (by their risk profile's equity allocation). Property, defined-benefit pensions and the state pension are excluded. It is an educational comparison against your Attitude to Risk, not a personal recommendation.
Suggested allocation
Alongside the assessment, the ATR view shows the broad industry-practice asset mix for your (capacity-capped) risk profile, 20/55/25 equities/bonds/cash for Cautious through 100/0/0 for Adventurous, next to where your portfolio is now, with the amount that would move each class to the suggested share. For the comparison, mixed funds are counted 55% equities / 45% bonds (mirroring the risk assessment's weighting), and each pension pot by its chosen risk profile's equity share. A shape, not a model portfolio or fund selection: Hundred Summers holds no fund data, and reshaping is best done inside existing tax wrappers.
Loss-of-earnings what-if
The Risk tab's Protection section re-runs the whole plan with earnings stopped from a chosen age: permanently, or for a chosen number of years with work resuming at the salary it stopped on (progression paused during the break, growth restarting on return; future roles that start after the recovery are unaffected). With cover included, critical-illness lump sums arrive in the shock year and income-protection benefits pay tax-free after each policy's deferral period (a part-year deferral scales the first benefit year), ending at the policy's own end or the return to work, whichever is first. Tax, NI and pension contributions all stop with the pay, so the knock-ons are real.
Which account pays, and reserves
Within each class the plan draws savings most accessible first (instant access, then notice, then locked), then in the order you list them, and draws every pension in proportion to its size, unless you set your own order for either on the Pension Income tab, when each account is used up before the next. An account marked as a reserve is left out of the year's planned withdrawal and used only once every other savings account and pension can pay no more, before any spending would go unpaid; a pension drawn from reserve is taxed like any pension withdrawal, and a reserve is never moved into an ISA by the optional 'use existing savings' sweep. The tax-free lump sum and any annuity purchase still come from all pensions in proportion. The stress test runs the full plan on every path, so it uses the same settings. This plan: Within each, savings most accessible first, then as listed; pensions in proportion to their size.
Savings vs pensions drawdown split
How each retirement year's withdrawal is divided between savings and pensions. This plan: In order: savings/ISA, then pensions, draining each before the next. Set it with the blend slider on the Pension Income tab; pension withdrawals are taxable income, savings withdrawals are not, so the split changes your tax bill.
Working-year shortfalls
A working year where total outgoings (spending, tax, pension contributions and debt repayments) exceed income is funded by drawing the gap from your savings accounts, in the same order as retirement withdrawals (see Which account pays): shown as a savings drawdown in the charts, exactly as in retirement. Pensions are never tapped for this (they are usually inaccessible before 55). If savings run out, the remaining gap is recorded as that year's unfunded spending: it fails the plan (the Goals tab's funding goals and event goal cards and the stress test all read this same verdict), because the plan cannot borrow and cannot touch pensions early, however large the pots. This matters for decisions like phased retirement: cutting your salary before your costs fall genuinely spends your savings.
Liquidity-based drawdown order
With no order of your own set, savings accounts are drawn most-liquid-first: instant access, then notice/short-term, then illiquid/locked, regardless of the order you entered them in. Within the same liquidity tier, your original entry order is preserved. This reflects how withdrawals realistically happen, but is still a simplification: it doesn't account for product-specific penalties or fixed-term maturity dates. Set your own order on the Pension Income tab to change it.
State pension basis & deferral
This plan: 0 qualifying years + 25 future working years + 10 gap years, capped at 35. The figure can be entered directly, or derived from your NI record on the Tax tab (35 qualifying years for the full new State Pension, minimum 10, pro-rata between, including working years still ahead and any gap years you plan to buy). Deferring adds a permanent 5.8% for each full year deferred and shifts the start age accordingly.
State pension indexation
The resolved state pension amount is treated as today's figure and indexed at the stated growth rate from today, becoming payable at the resolved start age (State Pension age plus any deferral). This keeps it consistent with the rest of the model, where salary, expenses and rental income all grow in nominal terms from today.
Income tax & NI threshold indexation
Tax and NI thresholds are not assumed static. They stay at today's frozen levels (the £12,570 Personal Allowance, £50,270 higher-rate threshold and matching NI limits, unchanged since April 2021) through the current freeze to 5 April 2031 (extended from 2028 by the Autumn Budget 2025), then indexed each year at your chosen rate (2.5%, editable in Plan assumptions above). During the freeze, rising pay pushes more income into higher bands (fiscal drag); once thresholds index in line with pay, your effective tax rate roughly stabilises rather than drifting upward forever. The £100,000 personal-allowance taper point stays frozen throughout, as it has since 2010.
Tax rates over time
Thresholds move over time (see above). The rates themselves are the current 2026/27 rates, plus the changes already legislated for later years, and are then held for the rest of the projection: income tax at 20% / 40% / 45% (the six Scottish rates where they apply), the 60% effective band in the £100,000–£125,140 personal-allowance taper, employee NI at 8% / 2%, dividends at 10.75% / 35.75% / 39.35% from April 2026, savings income at 22% / 42% / 47% and a 2-point property-income surcharge from April 2027, CGT at 18% / 24%, the 25% tax-free pension lump sum, and IHT at 40%. Every one of these is checked against its official page each morning (see the sources below). A future Chancellor may change any rate, add or remove bands, or alter reliefs: changes no model can predict. Because thresholds move but rates do not, your effective (average) tax rate still changes year to year as income moves through the bands. Treat the tax figures as illustrative under today's rules, not a forecast of future tax law.
Liability amortisation
Liabilities with a monthly payment amortise each year: interest accrues at the stated rate, then twelve months of payments are deducted, floored at zero once cleared. Repayments are assumed to already be part of your stated living expenses, so they are not double-counted. Liabilities without a stated payment simply compound at their interest rate: a conservative treatment of unmanaged debt.
External reference data
Where the model uses figures from outside your own inputs, they come from the following public sources. We try to keep these current, but rates, thresholds and tables are reviewed periodically (typically each tax year), and may lag the latest published figures: always check the primary source for the most current numbers.
Income Tax bands & Personal Allowance
HM Revenue & Customs, Income Tax rates and Personal Allowances, 2026/27 tax year; Scottish income tax, 2026/27. gov.uk/income-tax-rates, gov.uk/scottish-income-tax
National Insurance rates
HM Revenue & Customs, National Insurance rates and categories, 2026/27 tax year. gov.uk/national-insurance-rates-letters
Pension annual allowance & taper
HM Revenue & Customs, Pensions tax manual (PTM); GOV.UK pension allowances guidance, 2026/27.
Lump Sum Allowance (PCLS cap, £268,275)
HM Revenue & Customs, abolition of the Lifetime Allowance and introduction of the Lump Sum Allowance, effective 6 April 2024.
Capital Gains Tax rates & annual exempt amount
HM Revenue & Customs, Capital Gains Tax rates, 2026/27 tax year (18% / 24%; annual exempt amount £3,000 since April 2024). gov.uk/capital-gains-tax/rates
Inheritance Tax nil-rate bands
HM Revenue & Customs, Inheritance Tax thresholds and rates, frozen at £325,000 (nil-rate band) and £175,000 (residence nil-rate band) until April 2030. gov.uk/inheritance-tax
Default investment charges
Department for Work and Pensions, Pension Charges Survey 2020 (published January 2021): the average ongoing charge across members of automatic-enrolment workplace pension schemes is 0.48%, and 0.53% in other workplace schemes; portfolio transaction costs average about 0.07% on top. The 0.5% default for new plans sits at that average. Retail platforms plus funds and adviser-led portfolios often cost more; set your own figure, or give a single pot or account its own rate. gov.uk/government/publications/pension-charges-survey-2020-charges-in-defined-contribution-pension-schemes
Business & agricultural relief cap
HM Revenue & Customs, Business Relief for Inheritance Tax: for deaths on or after 6 April 2026, 100% relief is capped at £2.5 million of combined qualifying business and agricultural property, with 50% relief above it and unused allowance transferable between spouses. gov.uk/business-relief-inheritance-tax/what-qualifies-for-business-relief
State Pension amount & qualifying years
Department for Work & Pensions, new State Pension rates 2026/27 (full rate £241.30/week); 35 qualifying National Insurance years for the full amount. gov.uk/new-state-pension
Voluntary NI contribution cost
HM Revenue & Customs / DWP, Class 3 voluntary National Insurance contribution rates, 2026/27.
Annuity rates
Indicative market-average rates compiled from published UK annuity quote comparisons (2025/26), for a healthy non-smoker. Not from a single provider: real quotes vary and should be obtained via the Open Market Option. moneyhelper.org.uk/annuities
Baseline life expectancy
Office for National Statistics life tables, life expectancy at age 65 by sex at birth (84.5 male, 87.2 female). The health-profile estimate starts from the baseline for your recorded sex; if you'd rather not say, the longer female baseline is used: the safer planning assumption. ons.gov.uk - search 'National life tables'
Healthy life expectancy
Office for National Statistics, Healthy life expectancy, UK: 2022 to 2024 (released 19 February 2026). For each age band and sex, ONS publishes the further years people can expect to spend in good general health, as they report it in the Annual Population Survey, and the share of remaining life that is. The Longevity tab applies that share at your age and sex to your own remaining years, so a health profile that shortens or lengthens your life expectancy moves your healthy years with it; with sex unanswered it uses the later of the two. Beside it, it shows the plain ONS average for your age. This replaced a rule of thumb (healthy to 87% of life expectancy) that placed a typical 40-year-old's healthy years about ten years later than ONS does. It is shown for information and is not used in the projection; the Longevity tab offers it as an end age for your active retirement years, which changes your planned spending only if you choose it. ONS calls the release official statistics in development because survey samples are falling.
Health what-ifs
The What If tab's Health and your plan panel runs four health-led turns through your full plan, each built from parts the plan already has: ill health stopping work (the same calculation as Protection's What if you couldn't work, including any cover you've recorded; the plan lives on savings and that cover until your planned retirement age), going part-time from an age (a phased-retirement step), needing care from an age (your Later Life phase starts then, at your care type, or a residential home if none is set), and spending more while you're able (an extra amount each year in today's money until an age you choose, by default your healthy life expectancy). Each shows when spending first goes short, how much couldn't be paid, and what's left at the end in today's money. None changes your plan.
Retirement spending benchmarks
Pensions and Lifetime Savings Association (PLSA), Retirement Living Standards, used as a reference point only where mentioned in guidance text - not used directly in your projection unless you enter these figures yourself. retirementlivingstandards.org.uk
Long-term care costs by care type
Indicative UK annual costs used for the Later Life phase and the care scenario: in-home care £65,000, residential care home £57,000, nursing home £87,000. These are midpoints of published ranges from Laing & Buisson Care Cost Benchmarks, Age UK, and MoneyHelper (2024/25). Actual fees vary widely by region and provider - always obtain local quotes. laingbuisson.com - ageuk.org.uk - moneyhelper.org.uk
How spending rises
Living spending, working and retired (with the Freedom-years extra), is entered in today's money and rises with the plan's inflation every year, unless you set periods on Your Facts → Expenses that hold it in pounds between two ages or raise it at a rate of your own. A period covers the changes from its first age to its last, so spending held from 60 to 70 is the same in pounds at every age from 60 to 70 and rises again from 71. Care costs keep their own care inflation. In the stress test, only inflation-linked years follow each path's inflation; a period held in pounds stays held whatever inflation does. This plan: Rises with inflation (2.5% a year) every year.
Care cost inflation
Care fees have historically risen faster than general (CPI) inflation. The model defaults to 4% a year for care costs versus around 2.5% for other retirement spending, based on Laing & Buisson and ONS care-price trends. This rate is editable in the Longevity tab.
None of these sources endorse this tool, and we are not affiliated with HMRC, DWP, ONS, the PLSA, Laing & Buisson, Age UK, or any annuity or care provider. Where a figure has likely changed since this was last reviewed, your own entries always take precedence over any built-in default.
How we check the arithmetic
A projection is only useful if its arithmetic can be trusted, so the engine audits itself and the test suite holds it to account. These checks run on every built-in profile, on real anonymised plans, on every stress test in the battery, on every What-If lever in both positions and on a set of randomly generated plans, and any failure blocks a release. Every check is stated in the form an accountant would recognise.
The balance sheet closes
For every projected year, closing net worth minus opening net worth must equal investment and property growth, plus money that entered the plan from outside (earnings, other income, pensions in payment, employer contributions, the Lifetime ISA bonus, windfalls), minus money that genuinely left it (living costs, every tax and National Insurance, premiums, event expenses, annuity purchases, moving costs, debt interest), plus any spending the plan could not fund. Transfers between your own accounts, pots and properties net to zero, so a transfer that debits one side and credits nothing shows up as an unexplained remainder. The remainder is reported on every row and pinned to zero. Money cannot silently appear or disappear.
Every account closes on its own books
The same identity is run for each pension pot, savings account, property and loan individually: its closing balance must equal its opening balance plus every flow booked against it. Money credited to the wrong account, or a sale that lands in the wrong place, balances in total but not here, and is caught.
The cash statement balances
Each year's cash must net to nothing: income and withdrawals on one side; spending, taxes, contributions, debt repayments and what was swept into savings on the other; with any shortfall shown as unfunded spending. A bill that was charged but never paid for, or money drawn that nothing spent, cannot pass.
Allowances are honoured
Tax-free pension cash across upfront lump sums, the tax-free quarter of each withdrawal and defined-benefit commutations never exceeds the £268,275 Lump Sum Allowance over a lifetime, and each year's use is reconciled against the allowance consumed; ISA subscriptions never exceed the annual allowance; Lifetime ISA subscriptions respect the £4,000 cap and stop at 50; capital gains tax is never negative.
Nothing impossible happens
No balance ever goes negative; no pension is touched before its access age; no Lifetime ISA falls before 60; no tax, contribution or withdrawal is negative. Row zero equals the facts as you entered them, to the pound, and extending the plan's horizon changes no earlier year.
Investment bonds (onshore and offshore)
Each premium earns a 5% tax-deferred allowance a year for 20 policy years from when it was paid, cumulative, so allowances unused before the plan starts are carried in (enter when the bond started and any premiums paid later). Taking money across every segment is tax-deferred until withdrawals to date exceed allowances to date; the excess, less any excess already taxed, is a chargeable gain, top-sliced over the complete years since the previous excess (or since the bond began). Cashing in whole segments is a full surrender of those segments: the gain is their share of value plus earlier withdrawals, less premiums and gains already taxed, top-sliced over the years since the bond began; segments are cashed in whole, and what they pay beyond the amount needed moves to your cash. The draw that empties a bond is its full surrender, taxed on the real gain over its life. You choose, for each bond with more than one segment, whether the plan withdraws across all segments, uses the 5% allowance and then whole segments, or cashes in whole segments only. Offshore gains are taxed as income at full rates; onshore gains have the life fund's 20% tax treated as paid, so only tax above the basic rate is due. Simplifications, each the cautious way: a loss on surrender (deficiency relief) is not credited; top-slicing relief does not reinstate the personal allowance and the savings allowances are not applied to the gain; offshore gains are not top-sliced or time-apportioned; policy years are taken as the plan's years, and premiums, withdrawals and gains are spread evenly across segments. A gain from a draw made to pay the year's own tax bill is taxed the following year.
Taxes are paid, not merely charged
Taxes that arise from the year's own withdrawals (capital gains on selling investments, chargeable gains on investment bonds) are drawn for in the same year, grossed up for the tax that drawing them causes. Whatever no account can pay is shown as unfunded spending, never hidden in an overstated balance.
Validation pack
Standard cases for every fixed-rule area (income tax (England, Wales, Northern Ireland and Scotland), National Insurance, savings and dividend tax, CGT, pension allowances and tax-free cash, the State Pension, annuities and inheritance tax), each worked by hand from the official rule and set beside the engine's own figure, published at hundredsummers.co.uk/validation.html with CSV and JSON downloads. The engine figures are computed when the site is built, and the build fails if any case differs by more than £1.
Cross-checked against an adviser tool
One real plan (anonymised) was rebuilt input-for-input from the adviser's Voyant report of June 2026 and projected here; the report's own printed figures sit beside it in the test suite. Measured on 20 September 2026 and re-measured on 28 September: net worth within 1% of Voyant for the first fifteen years and about 6% lower at age 89 (5% before savings withdrawals moved to the start of each year; see “Retirement withdrawals”); pension values within 2% until drawdown; property identical; cumulative tax over the working years agreed to the thousand; both tools funded every year and both simulations passed the plan (Voyant 100%, here 100%). The residual is tax treatment in retirement (we tax the interest on a large cash balance, apply the personal-allowance taper and keep thresholds frozen to April 2031, where the adviser tool escalated every threshold from 2026), and we take each year's spending out before that year's growth, where the adviser tool appears to let it grow first; so we finish lower, never higher. The comparison is pinned: a change that drifts the projection away from the adviser tool fails the build and must be explained. It is one plan, so it shows agreement on that plan's shape, not on every case.
Outputs are pinned
Full projections for reference plans are stored as golden masters and compared on every change; a moved figure must be explained line by line before it is accepted. Every capability the site claims is tied to a test, and every published figure the engine relies on is checked against its official source each morning.
What this does not prove
These checks establish that the arithmetic is internally consistent and that the rules are applied as documented. They cannot make the assumptions right (growth, inflation and longevity are estimates, stated on this page), and they do not replace regulated advice. Every simulation in the app (the stress test, couples stress test, historical backtest, confidence solver and scenario comparison) runs each path through this same engine, and each path's books are checked to close.
Your data
Your plan lives on your own device, in this browser, so you can return to it. It is never uploaded to us. Because it lives in the browser, clearing your browser data, or switching to a different device or browser, means it won't automatically be there.
Automatic backup keeps a copy without any manual saving: set it up once in Settings, choose where the file lives, and Hundred Summers rewrites it every time your plan changes, and loads a newer version automatically at startup. Pick a folder that syncs with your cloud storage (Google Drive, OneDrive, iCloud or Dropbox), and the same plan follows you between your phone and your laptop. A "Change location" button lets you move it at any time. The file is protected automatically with AES-256 encryption using a key tied to your account: nothing to type, and any device you sign into can read it. (A file you previously protected with a passphrase keeps that passphrase: each device asks for it once, and there is no recovery without it.) Browser support for choosing a live file location is currently limited to Chromium-based browsers (Chrome, Edge); elsewhere, use the manual save below. After a browser restart you may be asked to reconnect the file: a single tap.
Alongside it, Export snapshot (in the profile menu) creates a dated copy of all your profiles: an archive of the plan as it stood that day, useful before big changes, at year-end, or as protection if the live backup file is ever damaged. It's also the way to share a copy without linking anyone to your live file, and the primary save method on devices without auto-backup support. On most browsers you can choose exactly where it goes and what to call it. You can save it wherever you like, and if you wish, you can save it into a cloud-synced folder such as Google Drive, OneDrive, iCloud or Dropbox, so it's backed up and available on your other devices. That is entirely at your discretion; Hundred Summers never sends the file anywhere itself. Keep in mind the file contains detailed personal financial information, so store it somewhere you trust, and remember the same if you share an exported report.
When saving a snapshot, you can optionally protect the file with a passphrase: it is then encrypted with AES-256 (the key derived from your passphrase) before it leaves the app, so the file is unreadable without it, including to us. There is no recovery if you forget the passphrase, so choose one you'll remember. Left blank, the snapshot saves as plain JSON so it can be restored anywhere, including by an adviser you send it to, or on a device where you can't sign in.
Use Import / restore to bring any saved file back in: an older snapshot, your auto-backup file on a device that can't connect to it directly, or after clearing your browser. Encrypted backups are detected automatically: account-protected files unlock silently when you're signed in with the account that made them, and passphrase-protected files ask for the passphrase. Restoring while auto-backup is connected also writes the restored plan back to your live file.
The single, optional exception to on-device processing is the Premium AI Copilot: when you ask it a question, the plan summary needed to answer (never your name) is processed securely by our AI provider, with your consent, and is never used to train AI models. The Copilot does no arithmetic of its own on options: when it compares them it asks for each to be calculated, your device runs it through this same engine (the What If tab's levers, on your full plan) and only the resulting figures go back, in today's money. Each calculated option shows as a card you can open in What If to see the same numbers. It can also search for the earliest retirement age and the most you could spend a year (on the central projection, at 85% confidence, or against every historical start year), and run the What If tab's health what-ifs. For those it picks the ages itself as hypotheses: it is never sent your health answers or anything estimated from them. If you never use the Copilot, nothing ever leaves your device. Full details are in the Privacy Policy.