What these results are actually telling you

Lesson one

A success rate is not a probability about your life

You will see figures like "87% success" all over this site. Here is what that sentence actually contains: the calculator invented several thousand plausible runs of good and bad market years, played your plan through each one, and in 87% of them the money outlasted the period you specified.

That is a statement about a model, not a forecast about you. And the model contains one assumption that is definitely false: it assumes you carry on spending exactly as planned no matter what happens. In the 13% where the money ran out, the simulated version of you watched the balance fall for a decade and changed nothing at all.

Real people don't do that. Someone heading for trouble notices around year ten or twelve and trims something — a smaller car, a cheaper holiday, one fewer year of helping the kids. The adjustment is usually modest, and made early it's usually enough. This is why the withdrawal strategy calculator exists: it puts a number on what that willingness to flex is worth, and the number is generally larger than people expect.

And it says nothing about how badly the failures failed

This is the part that gets missed. A "failure" in these simulations means the money hit zero at some point before the end. Running out at 94, in a paid-off house, with a state pension still arriving, is a manageable situation. Running out at 78 is a crisis. Both count identically in the percentage.

So when you see a success rate, look for the accompanying detail — when the failures happened, and what was left in the ones that survived. Two plans with the same 80% can be nothing alike.

Lesson two

Real and nominal figures are not comparable, and mixing them wrecks the answer

This is the most common mistake anyone makes with any retirement calculator, and it is worth understanding once, properly.

A nominal figure is the number that would appear on the statement. A real figure has had inflation stripped out, so it's expressed in what money will buy. Same money, two ways of describing it — and over thirty years they diverge enormously.

The same plan, described both ways

$1,000,000 invested, growing 7% a year, with inflation at 3%. Spending starts at $50,000 a year and keeps pace with prices.

The statement says $7.6 million. The honest figure is $3.1 million — that's what it would buy at today's prices. Both numbers are correct. Only one of them is any use for deciding whether you can afford to stop working.

Most calculators on this site handle it for you by asking for inflation separately and growing your withdrawals with it. Where a chart is drawn in today's money it says so. If a figure isn't labelled and it matters to your decision, assume nominal and divide it down.

Lesson three

The fan chart, and why its width matters more than its middle

When a calculator runs thousands of simulated futures, it can't draw them all. Instead it draws the shape of the spread: a line through the middle outcome, and shaded bands showing where the luckier and unluckier ones landed. Something like this.

This is $1,000,000, drawing $45,000 a year in today's money, over thirty years. Four thousand simulated futures. The plan survived in 78% of them.

Now read the spread rather than the line. The typical future ended with about $760,000 still in the account. The luckiest tenth ended above $3.4 million. The unluckiest tenth ended with nothing — and among the futures that ran out, the earliest did so in year 12, while the typical failure got to year 24.

Which has a practical consequence: the near end of the chart is the part worth acting on. The first five to ten years are where the fan is tight enough to plan against, and — because of sequence risk — they're also the years that do most to determine which part of the fan you end up in. Year twenty-eight will look after itself, and you'll have re-run all of this fifteen times before you get there.

Lesson four

The answer changes when you change an input. That is the point, not a bug.

People often find this unsettling. You nudge the return from 6% to 7% and the final balance moves by hundreds of thousands, and the whole exercise starts to feel arbitrary. If the answer is that sensitive, what's it worth?

Quite a lot — but not the answer. The sensitivity is the finding.

The useful thing any calculator here can tell you is which levers actually move your outcome and which ones barely register. That's a question about your particular situation, and it has a definite answer. Try it: take one calculator and change one input at a time, by an amount that's realistic rather than dramatic.

• Spend 10% less → usually an enormous effect, and it's the lever most within your control.
• Work one more year → typically large, because it adds a year of saving and removes a year of drawing at the same time.
• Hold two more years of cash → small effect on the median, large effect on the disasters. Exactly what you want from insurance.
• Return assumption 1% higher → big effect on the number, and no effect at all on reality, since you don't control it.
• Pick different funds within the same allocation → almost nothing, which is the opposite of how much attention it usually gets.

Run that exercise once and you come away with something durable: a short list of the two or three things that genuinely matter for you, and permission to stop worrying about the rest. That's worth far more than any single projected balance, and unlike the balance it doesn't expire.

Lesson five

What these models leave out

Being specific about the gaps is more useful than a general disclaimer, so here they are.

• Your own behaviour. The largest variable and the one nothing here models. The simulated you never panics, never sells at the bottom, never quietly increases spending in a good year. The real you might do all three, and the cost of that usually exceeds every other factor on this list.
• Tax. There is no tax logic anywhere in these calculators — no brackets, no allowances, no means testing. Where a page asks for a tax rate, it's a number you type in and the calculator multiplies by it. The tax assumptions audit sets out exactly what is and isn't modelled, and which way each simplification pushes the answer.
• Fat tails. The simulations draw each year's return independently from a smooth distribution. Real markets produce crashes worse and more often than that implies, and real bad years cluster together rather than arriving politely spaced out. Read the unlucky end of the fan as optimistic.
• Sequences that repeat. Because each year is drawn independently, the model doesn't reproduce the long grinding stretches that history actually contains — the 1970s, or Japan after 1990.
• Fees, beyond what you enter. Platform charges, fund costs, adviser fees and trading spreads all compound against you, and they're the only certain number in the whole exercise.
• Life. Divorce, redundancy, a parent needing care, a business that does better than expected, an inheritance. None of it is in here, and some of it will happen.

None of that makes the exercise pointless. It means these are instruments for comparing options, and poor instruments for predicting the future — which is exactly how they should be used.

Lesson six

Why this site gives a different answer from your adviser's software

It will, and neither of you is wrong. The difference will almost always be one of these, in roughly this order of size:

• Different return assumptions. One percentage point over thirty years is a difference of about a third in the final balance.
• Real versus nominal. If one model quotes in today's money and the other doesn't, the two figures aren't measuring the same thing at all. Check this first — it explains most large discrepancies.
• A different planning age. Running to 90 rather than 100 changes everything downstream.
• Tax. Their software probably models it. This one deliberately doesn't.
• Historical versus simulated returns. Some tools replay actual market history; these generate plausible futures. The two approaches give systematically different answers, and both are defensible.

When two models disagree, the productive move is to find the assumption they disagree about rather than deciding which tool to trust. That conversation is usually more valuable than either answer — and it's a good question to put to an adviser, because the honest ones enjoy it.