The unlucky year
Two retirees, the same pot, the same average return over thirty years, and completely different endings: the one who met a crash in year one may run out, while the one who met it in year twenty-five sails through.
Withdrawals are the reason. In a falling market you are selling units at depressed prices to fund the same lifestyle, so more units leave the portfolio for each pound you spend, and those units are permanently gone; they are not there to recover when the market does. A drop early in retirement therefore does lasting structural damage, while an identical drop later hits a portfolio that has already compounded and has fewer years of withdrawals ahead. This is sequence-of-returns risk. It is invisible in any calculator that reports an average, and it is the single largest reason otherwise sensible plans fail.
A crash with the bounce removed
The test applies the worst realistic case at the worst possible moment: a 35% fall (roughly 2008, or 2000–02) landing in your first year of retirement, with the portfolio afterwards growing only at your normal assumed return rather than enjoying the sharp rebound that historically followed. The figure approximates the 2007–09 global fall and the dot-com decline; deeper crashes have happened, but 35% is severe enough to be a real test without being an outlier nobody plans around.
Removing the bounce is deliberately harsh. Real crashes were usually followed by strong recoveries, and stripping the recovery out is what turns this from a simulation into a stress test: passing means the plan has genuine structural margin, not merely good luck. By the same token, a plan that fails here is not doomed, only fragile in one specific worst case. Reality has other shapes to offer too, a poor decade rather than a single fall among them, and Monte Carlo together with the historical backtest covers those.
The result is binary and blunt: either the money still lasts past 95, an age the ONS life expectancy figures suggest is worth planning for, or you get the age at which it runs out.
Yusuf, two years short
Yusuf plans to stop at 58 with £910,000 and spend £36,000 a year, just under 4%. His deterministic projection has him comfortable to 95 with money left over.
Then the crash lands. Year one takes the portfolio down to £591,500. He still withdraws his £36,000, which is now 6.1% of the reduced balance, and growth resumes at 5% real with no recovery bounce. The money runs out at age 94. Two years short, from a plan that had looked untouchable, and both of those missing years land at the age when he can least do anything about them.
What rescues the plan is more interesting than what breaks it. Cutting spending 15% for the first three years only, £30,600 instead of £36,000, carries survival past 95: three years of restraint, not thirty. Working one more year before retiring does roughly the same. So does holding two years of spending in cash to draw on instead of selling shares at the bottom, though it is worth knowing the calculation itself does not model such a buffer, even though a buffer is the single most effective defence against the exact scenario being tested. The practical lesson survives that gap: sequence risk is defeated by flexibility in the first few years, not by having a much bigger pot.
Which is why failing the test does not automatically mean more years of work. Check the cheaper fixes first, a cash reserve covering the first two years or a plan to trim spending temporarily if markets fall early; both usually beat several more years of full-time work. Nor is the answer simply more bonds. What you need is something other than equities to spend from during a crash, and historically, very bond-heavy portfolios failed long retirements more often, so a buffer for the early years beats a permanent shift.
What the test quietly assumes
Spending stays rigid all the way through the model, and almost nobody actually behaves that way in a crash; flexible approaches survive far better, and withdrawal strategies compares them. Guaranteed income is ignored too, unless you have added it as an event, and a state pension arriving mid-retirement changes the outcome substantially.
One boundary is worth drawing at the end. The sensitivity analysis ranks which inputs move your date; this page asks a different question entirely, whether the plan you already have survives one specific piece of bad luck.