How Landfall works out your numbers

This page is written for anyone who wants to look under the bonnet: a careful planner, an adviser, or someone in the industry deciding whether to trust us.

A retirement projection can run for thirty years or more. A stale threshold in the first year is not a rounding error by year thirty. It is a number someone may organise their life around, which is why we work this way.

What Landfall is, and what it is not

Landfall is a modelling tool. You describe your household, your pots and the life you want to fund, and it projects what happens year by year under UK tax law.

It does not give financial advice. It never recommends a product, a fund or a course of action. Where a decision depends on your own circumstances, such as buying an annuity, funding care or transferring a defined benefit pension, Landfall says that the decision needs a regulated adviser.

The projection is the same on every device and in both the free app and Landfall +. Nothing in the numbers is held back.

What the model covers

Where something is left out and it changes a figure you are reading, the app says so beside that figure.

Where the figures come from

Every rate, threshold, allowance, date and age limit Landfall uses is a dated record with its source attached. A number without a source counts as a defect, even if it happens to be correct.

When sources disagree, we rank them in a fixed order:

1. Legislation. The Acts and regulations themselves, read on legislation.gov.uk. Where they disagree with anything else, the law wins.
2. Official guidance. GOV.UK, HMRC’s internal manuals, gov.scot, GOV.WALES and nidirect, for the figures that put the law into practice.
3. Professional and parliamentary commentary. House of Commons Library briefings and the Low Incomes Tax Reform Group, for example, used to cross-check and to find edge cases.
4. Everything else. Accountancy-firm summaries, comparison sites and news articles tell us that something has changed. They are never the source of a value.

Nobody’s memory is a source. That includes ours, and the AI tools we use to help build Landfall. They look up the current source and cite it, or they give no figure at all. A figure an AI tool suggests is checked against the cited primary source before it enters the model.

Each record holds the same things. Here is the personal allowance as Landfall holds it today:

We never edit an old record. When a figure changes, its record is closed and a new one starts, so Landfall can reproduce the rules that applied on any past date.

Dates, nations, and the difference between announced and law

Every figure has a start date. It gains an end date when it changes. Your plan runs in UK tax years, and each year uses the rules in force for that year.

Every figure applies to a particular nation. Scotland sets its own income tax rates and bands, and Wales sets its own rates. Northern Ireland follows England for income tax, but its State Pension and care rules are run by its own departments.

Landfall records all four nations separately, even where today’s figures are the same. They match because of current policy, not because the law requires it, and either could change at the next budget.

Landfall projects law, not announcements. Every figure is marked as one of four things:

So a plan can be knowingly out of step with a measure that has been announced and later becomes law. We accept that. The alternative is a plan built on promises that may change before they pass.

How we check our working

We test Landfall’s calculations against people who are not us. We use four kinds of check, and a release does not ship until the first three pass.

Worked examples from the authorities. Where HMRC, the Scottish Government, legislation.gov.uk or the Low Incomes Tax Reform Group has published a worked example, we turn it into a test that Landfall must reproduce exactly. Each test carries a link to the page it came from.

Every boundary. Each threshold is tested £1 below, exactly at, and £1 above, in every nation. That covers the points where tax changes most sharply, such as the taper of the personal allowance between £100,000 and £125,140.

An independent cross-check. Every night we run thousands of synthetic cases through both Landfall and PolicyEngine UK, an open-source model built from legislation by a separate team. The cases cover income tax in all four nations, National Insurance and capital gains tax. Any difference of more than £1 a person a year has to be explained, and we investigate it rather than average the two answers.

The run on 23 October 2026 checked 6,619 cases and found no unexplained differences. The explained ones are written down with the law that decides each one. Most are places where we believe PolicyEngine UK is out of date.

The check works in our direction too. In October 2026 it found an error in our own State Pension age dates, which we corrected the same day from the Pensions Act itself.

Whole-plan comparisons. We build our published sample households in other UK retirement planners and compare the results year by year. These tools make different modelling choices, so we do not expect identical figures. Every difference has to be one we can explain.

How we keep up when the rules change

Machines notice changes, and a person decides what they mean. Nothing changes a figure in Landfall without a reviewed change by a named owner.

• We watch the sources automatically. Feeds from GOV.UK, legislation.gov.uk, gov.scot and GOV.WALES are monitored, and every page our figures cite is re-read and flagged if it has changed. A flag opens a review. It never changes a value by itself.
• Every flag is reviewed with evidence. Each one is checked against the source and the legislation. The outcome, with the figures compared and the dates they were read, is written down whether or not anything needs to change.
• Budget day follows a fixed routine. Everything the Chancellor announces is recorded on the day as proposed. Each measure is checked again when the Finance Bill is published, and reaches your plan only once it becomes law.
• No release during a fiscal event. We do not ship an update between a Budget, or a Scottish or Welsh budget, and the end of our review of it.

Assumptions, and how much they matter

Some inputs to a thirty-year projection cannot be looked up, because nobody knows them yet. They include investment growth, inflation, future State Pension rises and what happens to tax thresholds once the legislated freeze ends after 2030/31.

Landfall treats each one as a labelled assumption with a stated basis and a review date. You can see and change every one in your plan.

What the freeze adds to the tax on a salary, year by year until it ends, is worked out on the Budget calculator, from the same records.

What your pots and your State Pension could pay you a year for life, after tax, is worked out on the pension calculator, by the same engine the apps run.

Where a recognised public benchmark exists, we compare our defaults with it. The Financial Reporting Council’s AS TM1 standard sets the assumptions behind the projections on your own pension statements. It is a basis for statutory illustrations, not a forecast of returns, so we use it as a yardstick rather than an answer.

Every time we change the software, we test how far each assumption moves the result for our sample households, so we can see which conclusions are fragile.

For a sample couple, inflation and investment growth move the result more than any of the tax assumptions tested. A household with large taxable investments, high pension withdrawals or an estate above the inheritance tax thresholds may find the opposite, which is why every assumption is yours to change.

The stress test

The stress test, also called a Monte Carlo simulation, runs your plan between 250 and 2,000 times, each time with a different random sequence of investment returns. It then reports how many of those runs paid for everything your plan sets out to spend.

How the returns are drawn

• Each year’s return is drawn at random, from a lognormal distribution. Its average is the growth rate in your plan, and its spread is the volatility you set. The default is 10% a year for invested pots, our assumption for a balanced fund that holds shares, bonds and other assets. Cash defaults to 1%.
• Invested pots move together. Pensions, ISAs and taxable investments share one market shock each year, as they would if they held similar funds. Cash has a shock of its own. A pot you have not split is treated as one balanced fund at the default volatility.
• Split pots follow their own mix. If you split a pot into equities, bonds, commodities and cash, each part has its own expected return, volatility and correlation to the market. That lets bonds soften a bad year for shares rather than repeat it.
• Each year is drawn independently. We do not assume that markets bounce back after a fall, or that a fall continues.
• The runs are seeded. The same plan run with the same settings gives the same answer, so a change in the result is caused by a change in the plan, not by chance.

What stays fixed

Inflation, tax rules, the State Pension and your spending stay exactly as the plan sets them in every run. Only investment returns vary. Inflation and tax are covered by the assumptions above, which you can change and rerun. The one exception is a crash scenario that names its own inflation, below: its prices rise in every run alike, for the years it covers.

What counts as success

A run succeeds if every year of the plan is fully paid for, up to the age you set for the last person in the plan. A run fails if any single year falls short, even if money is left later. That is a stricter test than whether the pots stay above zero.

The test does not split spending into essential and discretionary. Every pound you planned to spend counts. You can test a leaner plan by copying it with lower spending and comparing the two.

Alongside the success rate, Landfall shows:

• How precise the rate is. It comes from a sample, so it is shown with the number of runs and a margin at 95% confidence. 93% of 250 runs is give or take 3 points, and more runs give a narrower margin.
• The spread of outcomes. A chart of your total assets each year shows the middle 50% and the middle 80% of runs, with the plan’s own line for comparison.
• When runs fall short. The share of runs that have run short by each year of the plan.
• What is left at the end. The range of what the plan leaves behind across the runs.

A success rate of 90% means that 90% of the simulated runs met the test above. It does not mean a 90% chance that your retirement will turn out as planned.

Sequence risk

When bad years come matters as much as how bad they are. A fall in the first years of retirement, while you are drawing on your pots, does far more damage than the same fall twenty years later. Because each run draws its years in a different order, the stress test captures this.

In our review of a sample plan, 7% of runs had a crash-sized four-year run of losses in the first five years of retirement, and 60% of those runs failed.

Adding a crash

You can also add a crash scenario: a fixed bad run of returns at a moment you choose, such as a sharp crash, a lost decade, a 2008-style fall, a stagflation or one you define. With a crash applied, Landfall reports two success rates side by side, one with the crash and one for the same runs without it.

The stagflation is the crash in which prices rise while markets fall, which is the combination that hurts a retirement plan most and the one a fixed inflation rate can never show. Its preset runs inflation at 9%, 7% and 3% over its years, and you can type a path of your own beside any crash. For those years spending, inflation-linked pensions and the State Pension follow that inflation rather than your assumption, care fees keep their margin above it, and bank savings earn their own rate. The plan page says the path beside the result.

Each holding in a split pot follows shares by its correlation, but never less than 0.3 by default, because in 2022 bonds fell alongside shares. That floor is a judgement, not a measurement. You can change it, down to below zero for a holding that gains while shares fall.

How far a holding falls is then scaled by its volatility against shares, so a steady bond fund often falls less than shares and a holding more volatile than shares falls further. A pot you have not split takes the crash in full, and cash is untouched.

The maths

Each year’s return R, for a holding with growth rate g and volatility σ, is drawn as:

Z is a standard normal draw, made from the year’s market draw and the holding’s own in proportion to its correlation. This makes g the arithmetic mean of the annual return.

The median compound rate is lower:

So 5.5% growth at 10% volatility compounds at about 5.0% in the middle run. σ is the spread of the log return, so the annual return itself spreads slightly wider, about 10.6% at those figures.

In a crash year, a holding in a split pot earns:

Here c is the crash’s return for that year, ρ the holding’s correlation to equities, f the floor, σ the holding’s volatility, and the σ beneath it the default volatility for equities. Cash is untouched, and a pot you have not split earns c.

What the stress test simplifies

Each simplification moves the result in a known direction. We would rather tell you which way than claim the model has none.

Care costs: the one place we use estimates

Care fees are the only cost in Landfall taken from industry figures rather than official ones. Growth and inflation are estimates too, but they are assumptions you set, not prices someone charges.

Government sets the means test, but it does not publish what care costs someone who pays for it themselves. Leaving care out would be less honest than estimating it carefully, because for many people it is the largest cost of later life. So care follows stricter rules than the rest of the plan:

• It is off unless you switch it on. A plan without care says so.
• The means-test rules are exact. They include the capital limits, tariff income and the personal expenses allowance, read from the statutory guidance exactly as tax rules are. Each nation’s are read from its own source, and where a figure is ever missing, Landfall says so beside the result rather than quietly borrowing another nation’s.
• A pension counts as the income it could buy. From State Pension age, the means test counts a pension as the annuity it could buy, as the regulations for England, Scotland and Wales set out. It is not counted as savings.
• The council pays its own rate. Once the council helps, the rest of a home’s fee is a top-up that someone else, such as family, pays, or the person moves to a home that costs less. Landfall shows the top-up beside the result and never takes it from your plan.
• The home pays through a deferred payment. Your home never counts while you or a partner live in it. Once someone moves into a care home for good with nobody left at home, it counts after a short grace period. Fees their savings and income cannot meet are then deferred against it and repaid from the estate, with interest where the nation charges it. Nothing is sold.
• Fees are always a range, never a single number. You choose low, typical or high for each type of care, along with how long care lasts and when it starts.
• Each range needs two independent sources. One report is a data point, not a range. Two reports that share the same underlying data count as one source.
• The method is fixed. “Typical” is the average of the reports’ UK averages. “Low” and “high” are the lowest and highest regional averages any report gives. Nobody adjusts a figure because it looks wrong.
• Ranges expire after twelve months. Every care figure is shown with the date it was read and the label “industry estimate, not official guidance”.
• We do not model reforms that are not law. That includes the £86,000 lifetime cap on care costs. It applied to England only, never came into force, and was cancelled by the government on 29 July 2024. Any changes that follow the Casey Commission will be handled like any tax announcement.

At the point of need, how to pay for care is a regulated decision. Landfall shows the shape of the problem, and points you to a specialist later-life adviser for the decision itself.

What we leave out, and how to tell us we are wrong

Gaps are recorded, not hidden. Where we have not read a source, the figure is either left out or marked as an assumption. The current list is the table under “What the model covers”.

Corrections are recorded. Every change to a figure is logged with its date, its source and the reason for it. When we find that we got something wrong, we fix it at the source.

Tell us. If you think a figure, a rule or a result is wrong, email support@landfall.money with the source you are relying on. We read every report against the law, and we will tell you what we found.

Changes already in law are included from the date they take effect. Measures that have only been announced are not. A Budget can change the figures, never the principles set out here.