How safe is my retirement plan? Two ways to stress-test it

A plan that works on average is not the same as a plan that works. The average hides the order returns arrive in, and two retirees with identical pots, spending and average returns can end up £60,000 apart purely because one hit a bad run first. The honest answer is a probability: the chance your money lasts to the age you choose. Aim for 85% or more, and know which lever moves it most.

Figures for the 2026/27 tax year · Published 4 October 2026

Every projection you have ever seen for your pension assumed the market returns the same percentage every year. It never does. This article explains what that assumption hides, the two tests that expose it, and how to read the answer.

Why a single projection cannot tell you

A projection takes your pot, adds your contributions, grows everything at a steady rate, subtracts your spending and reports the age the money runs out. It is the right first step, and it is how Isaac’s dashboard works. But a steady 5% a year is a fiction. Real markets deliver +20% one year and −15% the next, and the average of those two is not the point. The order is.

Two people retire at 60 with £300,000, drawing £15,000 a year. Over the next ten years both earn the same returns, averaging 3.3% a year. One gets the bad years first: −15%, −10%, then a recovery. The other gets the same returns in reverse order.

• Bad years first: £187,000 left after ten years.
• Good years first: £251,000 left after ten years.

Same pot, same spending, same average. A £64,000 gap, because withdrawals taken in a falling market sell more units at the bottom, and those units are gone when the recovery comes. This is sequence-of-returns risk, and it is the single biggest thing a steady-rate projection cannot show you. It matters most in the five years either side of retirement, when the pot is at its largest and the withdrawals have started.

Test one: 1,000 futures

A Monte Carlo simulation runs your plan not once but a thousand times. Each run uses a different sequence of returns and inflation, drawn around the growth rates in your own plan, so a plan that assumes cautious growth is tested around cautious growth and an optimistic one around optimistic. Each run ends with a simple fact: at the age you chose, was there money left?

Count the runs where there was, and you have the answer: the chance your money lasts to 90, or 95, or whatever age you set. That one number is the thing to look at. Underneath it, the spread of outcomes year by year shows how wide the range is, and a worst-case line (the outcome one future in ten does worse than) tells you what a poor run looks like in pounds.

One subtlety: a plan does not have to last to 100. Isaac also shows the chance you are still alive at each age, using standard UK life expectancy. A 70% chance of money at 100 reads very differently once you see that the chance of being there to spend it is far lower. Plan to the age you can reasonably expect, with a margin, not to the oldest age imaginable.

Test two: every retirement year on record

Simulated futures are built from assumptions. History is not. The second test replays your plan as if you had retired in each year of the past 150, using the real sequence of market returns and inflation that followed, so your plan meets the 1970s inflation, the 1930s, 2000 and 2008 exactly as they happened. Each year’s returns are centred on your own plan’s growth rates so the comparison is fair, and everything is shown in today’s money.

The output is a story, hardest year first: “Retire in 1969 and your money lasts to 84. Retire in 1982 and it lasts past 100.” Then a count: in how many of those years was the money still there at your chosen age? If the answer is “every one”, your plan has survived everything the last century and a half threw at it. If it is “31 of 36”, you know which kind of decade breaks it, and you can test what fixes it.

The four levers, and which one usually wins

A probability is only useful if you know what moves it. Four things do, and the test should run each one on the same 1,000 futures so the comparison is like for like.

• Spend less. The strongest lever by a distance. Cutting £200 a month from a £2,500 budget typically adds more to the chance of success than any other single change, because it compounds across every year of retirement.
• Retire later. Two more years of work means two more years of contributions, two fewer years of withdrawals and two more years of growth. It is powerful, but it is also the lever people are least willing to pull, so test it before you rely on it.
• Save more now. An extra £200 a month helps most when retirement is a decade or more away. Close to retirement it is a small lever.
• Invest differently. Holding more in growth assets raises the typical outcome and widens the spread. A path that steps down as you age (more growth now, more caution later) cuts the worst outcomes with less cost to the typical one. Test it rather than assume it.

Isaac’s stress test offers the levers that match your verdict: if the plan is secure it shows how much room there is to spend more or retire earlier; if not, it shows what would make it more secure. Each result shows the chance before and after, and what happens to the worst one-in-ten outcome.

What a stress test is not

It is not a forecast. No one knows next year’s return, and a 92% chance of success can still produce a bad decade. What the test gives you is the shape of the risk and the size of the margin, which is what you need to decide how much to spend and when to stop. Re-run it once a year, after any big change, and after a bad year in the markets. If the chance has dropped, the levers tell you what to do about it long before it becomes a crisis.

Five questions to ask of your own plan

• What is the chance my money lasts to 90? To 95?
• In the worst one-in-ten futures, how much is left at 85?
• Which retirement years in history would have broken the plan, and why?
• Which single change adds the most to the chance of success, and can I live with it?
• How does the chance change if I de-risk after retirement rather than before?

If you can answer all five, you have a plan rather than a projection. If you cannot, the 4% rule is a reasonable place to start, and a stress test is where to finish.