How the projection works, and what it leaves out
Any app that draws a line into the future is making assumptions for you. Most do not say which. This is the list, both halves of it.
Any app that draws a line into the future is making a pile of assumptions on your behalf. Most of them do not tell you which ones. This page is the list, both what Valnivo models and what it knowingly refuses to model.
The part that is not random
The projection steps forward one month at a time. Income and spending grow at their own annual rates. Standing contributions to your investments come out of the settlement account, and whatever is left accumulates as cash. If that account cannot cover the plan in some month, the shortfall is sold out of your holdings, and the app reports the first month that happens.
Because only investment returns are random, the cash ladder and the contribution schedule are identical on every simulated path. That is deliberate: it is what makes the resulting figures explainable rather than merely produced.
The part that is random
Each holding follows a standard lognormal random walk in its own currency, using the return and volatility you set. Every holding shares one market draw each month, scaled by its asset class's sensitivity, and keeps its own independent component on top. A currency's central path follows the inflation gap between its country and yours, with volatility spreading the runs either side of it.
Note one deliberate asymmetry. Holdings carry the usual correction that keeps a volatile asset's median below its average. Currency legs do not, because parity is a statement about where a rate sits in the middle, not about an asset earning a risk premium.
Why the runs are drawn in pairs
Paths are simulated in antithetic pairs: the second of each pair uses the exact negation of the first's random draws. This matters far more than it sounds. A holding at 55% annual volatility over thirty years has an enormous spread, and independent sampling puts the median tens of percent off its true value. In testing, a bitcoin median read €4.07M against a theoretical €2.76M before pairing, and landed on €2.76M after. Same model, same number of runs, one fixed and one badly wrong.
What comes out
- The 10th, 50th and 90th percentile of net worth for every month to the horizon.
- A breakdown by country, read off the single simulated path whose total is closest to the median, so the parts always add up to the whole instead of being separate percentiles that do not reconcile.
- A money-weighted return, solved over your actual cash flows rather than assumed.
What this model does not do
- No tax, anywhere. No capital gains, no dividend withholding, no wrappers, no allowances.
- No mean reversion, no fat tails, no volatility clustering. Thirty years of constant lognormal volatility is a convenient fiction, and it is least defensible exactly where it matters most, which is the tail on a very volatile asset.
- Correlation is one market factor with a sensitivity per class, not a matrix.
- Currency shocks are independent of market shocks. In a real risk-off month they are not.
- The band is model uncertainty only. The 10th–90th percentile band is a percentile range of the model's own simulated outcomes. It is not an 80% confidence interval for your actual future financial position, and it says nothing about the chance that your assumptions are wrong, which is by far the larger risk. Likewise the middle line is the 50th-percentile simulated outcome, not a forecast of what will happen.
There is one presentation compromise worth naming. The chart's vertical axis follows the median lines and gives the uncertainty band a fixed amount of headroom. Past that the band is clipped, and the true range is quoted in words beside the chart instead. That keeps the median readable on a thirty-year view, at the cost of not drawing the extreme tail to scale.
Why say all this
A projection is a way of asking what follows from your assumptions, not a forecast of your life. Knowing which assumptions are baked in is the difference between using the number and believing it. Change one input at a time on the Future screen and watch what moves. That is the honest use of a model like this one.
Run it on your own figures
Valnivo is a free financial planning tool: it turns what you earn and spend into a picture of your finances 10, 20 and 30 years out, and shows what changes if you save more, stop earlier, or prices rise faster. It holds a rate of inflation for every country you use and a rate for every currency, so the arithmetic on this page runs on your numbers instead of an example. Nothing leaves your device unless you choose to sign in.