Retirement planning with simulation: the complete guide

How much do you need?

The standard first pass is 25 times the annual income you want, which corresponds to drawing 4% of the portfolio in year one and adjusting that amount for inflation afterwards. If you want $60,000 a year, that is $1.5 million.

Treat that as a sanity check rather than a target, because every input behind it moves. Lengthen the horizon and the sustainable rate falls. Add a 1% annual fee and it falls again, by roughly the size of the fee. Make spending flexible and it rises. The safe withdrawal rate calculator runs the arithmetic in both directions, and the definition covers what the rate does and does not promise.

A projection is not a plan

Compound a single assumed return over thirty years and you get one confident number. It will be wrong, and not in a way you can correct with more decimal places, because markets do not deliver the average — they deliver a sequence, and the sequence changes the outcome even when the average does not.

A Monte Carlo simulation replaces the single path with thousands. The output is a distribution: a median, a range, and the share of runs in which the money lasted. In one household we modelled, a conventional projection of $4.3 million turned out to sit at the 71st percentile of the simulated outcomes — the median was $2.9 million. Same inputs, very different plan.

The engine matters too. If each year is drawn independently from one distribution, the model will understate the clustered bad decades that actually break plans. Killion uses a regime-switching model with fixed seeds so runs reproduce.

The risk that actually ends retirements

Two people can earn identical average returns over thirty years and end up in completely different places, purely because of when the bad years arrived. Sequence of returns risk is near-harmless while you are saving — a crash is a discount when you are still buying — and it is the single largest threat once you are selling assets to fund spending.

We isolated it by scheduling the same 40% crash at different ages in an otherwise identical plan. Landing at retirement age 65 it cut the success rate from 86% to 71%. Landing at 48, the same crash cut it only to 81%. The event was identical; the timing did the damage.

The defences are all forms of flexibility: a cash buffer so you are not forced to sell into a fall, a spending rule that cuts after bad years, or working slightly longer. None of them require predicting anything.

How you spend it matters as much as how you invest it

Most planning attention goes to allocation. In practice the spending rule is at least as powerful, because it decides who absorbs market risk: your portfolio or your lifestyle.

We ran flexible and rigid withdrawal policies through 5,000 identical market lifetimes. The flexible 4% policy lasted in 100% of runs; the rigid 4% policy failed in 1.1% of them. The flexibility was not free — it bought that safety by spending less during bad decades — but it is a trade most people would take, and it costs nothing to decide in advance.

The catch is that a flexible rule only works if you can genuinely flex. If nearly all your spending is fixed costs, a rule that instructs a 15% cut cannot be followed. Work out what share of your spending is discretionary before choosing a policy. Killion models five named rules, including Guyton-Klinger guardrails — see withdrawal strategies.

Fees: the certain loss

Returns are uncertain. Fees are not. A 1% annual charge is deducted in good years and bad, on a balance that is growing, and it also removes the compound growth those deductions would have produced. That second effect is what makes the total so much larger than the headline percentage suggests.

We simulated 5,000 forty-year saving lifetimes at 0%, 0.5% and 1% fee drag. The dollar gap at the finish line is far bigger than the number sounds, and it is worst for exactly the people who saved hardest and started earliest. Run your own figures on the investment fee calculator.

Testing against real history

Simulation and backtesting fail in different ways, which is exactly why running both is worth the effort. Simulation explores futures that never happened but inherits its assumptions. Backtesting assumes nothing but has only one sample — the history that actually occurred.

A plan that survives both is robust for two independent reasons. A plan that looks fine in simulation and fails in the 1966 and 1929 starts deserves another look. Killion backtests to 1928, testing every start year with a complete horizon.

One caution when reading a survival rate: historical windows overlap heavily, so the effective sample is far smaller than the number of windows implies. Use it as a robustness check, not a probability.

Building a plan, step by step

1. Get the balance sheet complete. Every account, every liability. A plan missing a mortgage flatters itself.
2. Get spending honest. Split it into fixed and discretionary — the split determines which withdrawal rules are actually available to you.
3. Pick a horizon deliberately. Plan to an age you would be uncomfortable outliving, not to life expectancy, which half of people exceed by definition.
4. Run the simulation and read the bad tail first. The worst decile is the world you would have to live in. If that is survivable, the plan is robust.
5. Backtest the same plan. Check it against the starts that broke other plans.
6. Change one thing at a time. A paired comparison against identical markets tells you how often a change helped, which is far more useful than watching a median move.
7. Re-run when reality changes, not on a schedule. A plan is a model of your life, and it goes stale when your life does.

A full walkthrough of the mechanics is in the help centre, starting with how to read your projection.

The full research library

Every article below is original simulation work with its method stated. Browse by topic or read the full blog.

By topic

• Retirement withdrawals — Safe withdrawal rates, the 4% rule, and flexible vs rigid spending policies.
• Sequence of returns risk — Why the order of market returns can matter more than the average — especially near retirement.
• Monte Carlo & projections — Why single-number projections mislead, and how probability-based planning works.
• Investing costs — Fee drag, expense ratios, and the lifetime dollar cost of 1%.
• Investing behavior — Time in the market vs timing the market, and the cost of panic selling.
• Strategy comparison — Apples-to-apples benchmarks of allocation and spending rules on identical markets.