The clause nobody reads out
Almost every retirement calculator plans to a fixed horizon (usually 90 or 95) and reports whether the money lasts that long. Buried in that is an assumption worth saying out loud: that you are alive the whole time.
For a 65-year-old, the chance of reaching 95 is less than one in five, as the ONS life expectancy data makes plain. So a plan that "fails at 92" is failing in a scenario most people in that position never experience, and optimising hard against it means trading years of your actual life for insurance against a future you are unlikely to see.
This view puts mortality back into the picture. Every future year is split three ways: rich (still alive, money intact), broke (still alive, money gone) and dead (no longer here). Only the middle one is the failure worth engineering against. Morbid? A little. But it is more honest than the alternative, which is silently assuming you live to 95 and working extra years to fund it; the chart simply makes an existing assumption visible.
The mortality curve deserves its caveat before anything else gets one. The Gompertz model here is calibrated to UK/US unisex life tables, and population tables are not you: your health, family history and circumstances can shift the curve substantially in either direction. It approximates period tables, too, which understate how long today's savers are likely to live, so the survival figures below are on the pessimistic side of the honest range.
Same risk, different question
Presented conventionally, a plan reports its chance of running out before 95. With mortality applied, the same risk typically shrinks to a fraction of that headline, because most of the failures land at ages most people never reach. The financial risk did not change; the framing did. And the second framing answers the question people actually have, which is not will my money last to 95 but will I ever be alive without money.
That is not a licence to under-plan. The alive-and-broke chance is still real, and it lands at the point in life when you are least able to do anything about it. Nor is the answer to plan to a shorter horizon, which hides the risk instead of pricing it; keep the long horizon and read the alive-and-broke figure, which already accounts for survival probability.
The overlay cuts the other way too, honestly. It tends to show that the years you should worry about are the early ones, when you are almost certainly alive and the portfolio is most vulnerable to a bad sequence. Late failures are partly protected by mortality. Early ones are not protected at all.
A second worked example puts numbers on the decision that usually follows. A 58-year-old whose plan fails 42% of the time has a 10.4% chance of ever being alive without money, and pricing his "one more year" against that figure is where the chart stops being a curiosity.
Ade's mid-eighties
Ade is 61, retiring now with £680,000 and spending £34,000 a year, a 5% withdrawal rate. Conventional Monte Carlo gives him 47% success: read that way, the plan is a coin flip, and the instinct is to keep working. With mortality applied the same plan shows a 13% chance of ever being alive and broke, because on these population tables the chance of reaching 95 at all is about 8%. His alive-and-broke risk peaks around 85: old enough that survival is uncertain, young enough that it is far from unlikely.
What Ade does with that is a personal call, not a mathematical one. Some see 13% and decide the plan is sound. Others see it and buy an annuity to cover essentials from 85, which converts the tail risk into a known cost. Guaranteed income later in life is, in fact, the main thing that pulls the alive-and-broke number down: a state pension (whose start date you can check against your birth year on GOV.UK), a defined-benefit pension, or an annuity bought at older ages, each turning an uncertain tail into a fixed cost. The number does not make the decision; it just stops the decision being made against a misleadingly scary 53%.
Two of the model's blind spots matter most in exactly Ade's decade. Spending is assumed flat, when in reality it usually falls with age, which the spending smile models separately. And care costs are not modelled at all, the very expense most associated with the late years this chart covers. The simulation itself carries the same caveats as Monte Carlo, since it is built on the same engine.
Couples, and everyone still working
For two people the model tracks either-partner-alive survival, which is why a couple's effective horizon is longer than either individual's: the money must last until the second death, not the first. That single fact usually matters more to a couple's plan than any adjustment to the withdrawal rate. See the couples calculator.
For the bands themselves, and what each one is telling you at a given age, how to read the chart goes through it region by region.
The chart earns its keep before retirement as well. Working "one more year" for safety has a cost the sensitivity analysis can quantify, and this view shows what that year buys in the years you are most likely to be alive for.