The projection runs your plan a thousand times over different market sequences. Here is what each number it reports back is actually claiming.
Success %
The share of runs that funded every obligation all the way to your end age: spending, one-time expenses, income tax, and insurance premiums alike. A single unfunded dollar in a single year fails that whole run.
So it is a pass/fail tally, not a margin of safety. It says nothing about how the failures failed. That is why the next number exists.
Failure depth
When runs do fail, the app also reports the typical depletion age and the total shortfall, measured across the failing runs only. Two plans can both succeed 85% of the time and be nothing alike: one runs a little short at 93, the other is empty at 78. Read the pair together, never the success rate alone.
The median and the bands
The median is the middle outcome. Half the runs did better, half worse. It is not a forecast, and no single year of it is a promise.
Turn on the probability fan and you see the spread around it:
- The 25–75 band is the everyday range: half of all runs landed inside it.
- The 10–90 band is the stretch: falling outside it is about a 1-in-10 result, in either direction.
The 90% downside readout states that lower edge as one number: the 10th-percentile ending value, or, once more than 10% of runs fail, the age by which that run is out of money.
Confidence by age
The success rate is one number for the whole plan. Under the fan, the app also reports how that confidence should change as you get through your early retirement years: for each age, the share of outcomes that reach it on track and go on to fund everything. See confidence by age.
Per-age numbers are medians too
Every column of the year-by-year table is the median for that age, taken across all runs. Neighboring rows can come from different runs, so they do not add up the way one household's actual years would.
To follow a single coherent path instead, zero the volatility inputs. The Trace: zero volatility button does it in one click, and the medians collapse onto that one path exactly. See why re-running gives the same answer.
The mortgage group
A financed home adds two more readouts, beyond the ordinary numbers above.
The Housing panel shows two halves. Planned is a deterministic amortization run once at your assumed inflation, treating every entered payment as fully funded: when the loan ends, total contract interest, and interest saved by any extra principal or an early payoff. Simulation is what the thousand runs actually did with that schedule: the age the loan is repaid (median and the 10th to 90th range), the share of paths that repay it at all, and each payment's own eligible and fully-funded share. See extra principal payments for the full methodology.
Detail's mortgage columns split the same note into stocks and flows. Opening and closing balance are stocks: what the note owes at a point in time, each read at its own year's deflator, since closing balance is the next row's own opening. Interest, scheduled principal, extra principal (funded), payoff principal, payoff interest and unfunded planned principal are flows: what moved during that year.
The ordinary installment is treated as a contract: it is deemed paid ahead of other recurring spending on every simulated outcome, whatever else that year's budget cannot cover, so the note keeps amortizing on schedule. That simplification is made visible, not hidden, through two diagnostics beside the mortgage group: funded non-mortgage (how much of what a path actually funded went to spending other than the installment) and installment gap (how much of the installment itself that funded spending did not cover). A shortfall here shows up as a cut to your OTHER spending, never as a missed mortgage payment: there is no arrears state, so nothing carries forward with penalty interest.
When the three models disagree, the disagreement is the finding: it tells you the answer depends on which market history you believe, which is worth more than any single success rate.