How much do I need to retire, and how wrong is that number?
Every retirement calculator hands you a single figure, and the confidence of that figure is the most misleading thing about it. The honest answer is not a number. It is a number with odds attached, and the odds are worse than most people expect.
The answer, as a table rather than a number
The popular rule is 25 times your annual spending, which is the 4% rule turned upside down. That multiple is real, but it is not a guarantee, it is a bet. Here is what different multiples actually bought on the plan modeled below.
| Save this multiple of your spending | Which is a withdrawal rate of | Money lasted to 95 in | Median pot left |
|---|---|---|---|
| 20 times | 5.0% | 56% of lives | $147k |
| 22 times | 4.5% | 65% of lives | $436k |
| 25 times | 4.0% | 74% of lives | $725k |
| 29 times | 3.5% | 84% of lives | $1.05M |
| 33 times | 3.0% | 91% of lives | $1.40M |
The famous 25 times leaves you failing about one time in four. Whether that is acceptable is a question about you, not about the arithmetic. Moving from 4% to 3.5% means accumulating four more years of spending before you stop, and it buys ten points of certainty. Moving the other way, from 4% to 4.5%, gives back nine.
Notice how steep that is. Half a percentage point of spending is worth roughly nine points of success. Retirement planning is far more sensitive to what you withdraw than to almost anything else you control.
Why your calculator gave you a bigger, calmer number
The plan being modeled: start at 30 with $40,000 saved, add $12,000 a year, retire at 65, plan to 95, and assume a 5% return after inflation. That is a reasonable, unremarkable plan.
Flat returns say $2.00M and 100%. Varying returns say $725k and 74%.
Set volatility to zero in the retirement simulator and every single one of the thousand simulated lives ends at exactly $2.00M, with a 100% success rate. That is the answer an ordinary retirement calculator gives you, and it is arithmetically correct.
Now allow returns to vary at 15% volatility while keeping the same 5% average. The success rate drops to 74% and the median life ends with $725k. Same average return, same contributions, same everything. The median outcome is a third of what the smooth projection promised.
That gap has a name: volatility drag. A sequence of ups and downs compounds to less than its own average, because a 20% loss needs a 25% gain to undo it. The smooth number is not optimistic. It is answering a question about a world where markets do not move.
The spread is the finding, not the median
Here is the same plan expressed as the range of lives rather than a single figure:
| Percentile | 10th (unlucky) | 25th | 50th (median) | 75th | 90th (lucky) |
|---|---|---|---|---|---|
| Ends with | broke | broke | $725k | $2.40M | $5.09M |
Identical inputs. Identical discipline. The unlucky quarter runs out of money before 95 and the lucky tenth dies with over five million. The difference between those lives is not behavior, it is the order the returns arrived in. That is sequence-of-returns risk, and it is the single most under-appreciated fact in retirement planning.
Every figure here is read from the retirement simulator by changing one input at a time and leaving the rest at their defaults. The zero-volatility case doubles as the model's own sanity check: with randomness switched off, all thousand lives must be identical and match an ordinary deterministic projection, and they do, at $2.00M. Method on how it's checked.
Which levers actually move the odds
Starting from that 74% baseline, here is what each change is worth. One of these results is not what most people expect.
| Change | Success rate | Median pot |
|---|---|---|
| Baseline: retire at 65, save $12,000, 5% return | 74% | $725k |
| Work two more years, retire at 67 | 78% | $905k |
| Save 20% more, $14,400 a year | 74% | $847k |
| Earn one point more, 6% instead of 5% | 84% | $1.67M |
Saving 20% more did not improve the odds at all
Look at the third row. Contributions up by a fifth, and the success rate stayed at 74%. That is not a glitch. The withdrawal here is a percentage of whatever you accumulate, so a bigger pot funds a proportionally bigger income, and the sustainability of the plan is unchanged.
This is worth being precise about, because it is easy to misread. Saving more still made you better off: the median pot rose from $725k to $847k, which is a richer retirement. What it did not do is make the plan more likely to survive. If you want better odds rather than a better lifestyle, the lever is a smaller withdrawal share, more years of compounding, or a higher return.
Two more years of work is the most reliable lever you control
Retiring at 67 rather than 65 took the odds from 74% to 78%, and it does three things at once: two more years of contributions, two more years of compounding, and two fewer years of drawdown. Unlike the return assumption, it is entirely within your control.
The 6% row is the largest improvement on the table, taking success to 84%, and it is the one you should trust least. Assuming an extra point of return is not a decision you make, it is a hope you hold. Working longer is a decision.
How much to trust any of this
These percentages carry real uncertainty
Each figure comes from a thousand simulated lives, which means it has sampling error. Running the identical plan with four different random seeds produced 74%, 78%, 77% and 77%. So the honest reading of "74%" is "roughly three in four, give or take a couple of points".
Take the shape of the tables seriously and the individual cells loosely. That withdrawal rate dominates, that volatility costs you a third of the smooth projection, and that saving more raises your standard of living without raising your odds, are all robust. The second decimal place is not.
What the model leaves out, all of it pointing the same way
Returns are drawn from a bell curve, which understates how often extreme years happen. Real markets have fatter tails than the normal distribution, so the true failure rates are probably a little worse than shown. There are no fees, and a 1% annual fee compounds against you for decades. There are no taxes.
Pushing the other way, there is no Social Security or pension here, which for most people is a meaningful floor under the worst outcomes. And the model assumes you spend the same amount every year regardless of what markets do, when in practice people cut back after a bad year, and that flexibility is worth more than almost any tweak to the portfolio.
Which is the real conclusion. The number you need is not fixed, because your spending is not fixed. A plan that adjusts survives conditions that break a plan that cannot. Use the multiples above to know roughly where you stand, then run your own numbers, and treat any single figure with the suspicion it deserves.
Common questions
Is 25 times my annual spending enough to retire on?
It is a starting point with roughly a one in four chance of failing. 25 times your spending is the 4% rule restated, and on the plan modeled here it survived to age 95 in 74% of a thousand simulated lives. If you want the odds nearer 9 in 10 you are looking at closer to 33 times, which is a 3% withdrawal rate. There is no multiple that makes the risk zero.
Why does my retirement calculator give a much bigger number than this?
Because most calculators apply the same return every year. Do that here and the plan ends at exactly $2.00M with a 100% success rate. Allow returns to vary while keeping the same average, and the median outcome falls to $725k and a quarter of the lives run out of money. The smooth number is not conservative or optimistic, it is answering a question about a world that does not exist.
Does saving more improve my odds?
It improves how much you get to spend, but not necessarily the odds. Raising contributions by 20% here lifted the median pot from $725k to $847k while the success rate stayed at 74%, because the withdrawal is a percentage of whatever you accumulate. A bigger pot funds bigger withdrawals. What moves the odds is working longer, spending a smaller share, or earning a higher return.
How precise are these success rates?
Not very, and it would be dishonest to pretend otherwise. Each figure comes from a thousand simulated lives, so it carries sampling error. Running the same plan with four different random seeds gave 74%, 78%, 77% and 77%. Treat any of these numbers as plus or minus a couple of points, and treat the shape of the table as the real finding rather than any single cell.