Methodology
Every assumption, stated plainly
A tool that hides its assumptions is asking to be trusted rather than understood. These are all editable inside the tools themselves — if you disagree with a number here, change it and see what happens.
Default capital market assumptions
Starting points, not forecasts.
Expected equity return
7.0%
Nominal, before fees. Deliberately below the realised long-run US average — that average was flattered by valuation expansion which cannot repeat indefinitely, and every major forward-looking estimate is lower than the backward-looking one.
Equity volatility
17.0%
Annual standard deviation. Understates tail risk, because real returns are fatter-tailed than the lognormal family used in the simulation.
Expected bond return
4.0%
Investment-grade, nominal. Bond returns are far more predictable from starting yields than equity returns are from anything, so this assumption ages faster than the others.
Stock-bond correlation
0.10
Slightly positive. The negative correlation many investors assume held for a specific inflation regime, and 2022 showed what happens when that regime ends.
Inflation
2.5%
Used to convert nominal figures into today's money. Every long-horizon output is reported in real terms by default.
Historical data
1928–2025, annual.
Equities: S&P 500 annual total return, dividends reinvested.
Bonds: 10-year US Treasury constant-maturity annual total return.
Inflation: US CPI-U, December to December.
Where the numbers come from
Inflation is measured December to December
US history is not neutral history
Modelling choices
Real terms by default
Long-horizon outputs are inflation-adjusted. A projection showing a seven-figure nominal pot is not informative when the question is what it buys.
Fees as a continuous drag
Fees are modelled against assets rather than as a one-off deduction, because that is how expense ratios work — they scale with the balance and compound against you for as long as returns compound for you.
Exact Fisher relation
Real returns use (1+r)/(1+i)−1 rather than the r−i approximation, which drifts by close to a full percentage point at the rates people plug into retirement projections.
Seeded simulations
Monte Carlo runs are deterministic given a seed, so a result can be shared and revisited. A simulator that gives a different answer on refresh teaches people to trust the last number they saw rather than the distribution.
What we do not model
Taxes beyond a single flat rate you supply, property, insurance, a partner's finances, business assets, currency exposure, or jurisdiction-specific account rules. Each of these can reasonably reorder your plan, and a general tool that pretended otherwise would be guessing.
How the plan is ordered
The seven steps are sequenced by the certainty-adjusted return of the money that flows into each one, not by convention:
- An employer match is an immediate return on contribution — nothing competes.
- A one-month starter buffer stops the next surprise recreating the debt you clear.
- High-interest debt returns its own rate, guaranteed and tax-free.
- A full cash buffer has a low nominal return but high option value: it is what stops an income shock forcing an asset sale at the worst price.
- Tax-advantaged investing beats taxable investing for the same asset, at no extra risk.
- Moderate-rate debt and investing then compete on expected value, genuinely.
- Everything after that funds the long-run target.
Each step in your plan carries this reasoning inline, because a plan you understand is one you will still follow when it becomes inconvenient.
What we collect
The figures you enter stay on your device unless you create an account, in which case they are stored in your own document and are readable only by you.
Analytics records which tools are opened and completed, so we can tell what is useful. It never records the values you type. That is enough to steer what gets built next without turning a metrics pipeline into a place financial data can leak from.