In brief
The projection chart runs thousands of simulations with randomised annual returns and shows the resulting spread of outcomes as percentile bands — not a prediction of what will happen.
In plain English
Investment returns vary from year to year. A single projected line based on a fixed growth rate ignores this variability. The projection chart addresses this by running many simulations, each with a different randomly generated sequence of annual returns, and displaying the spread of outcomes.
This approach is called a Monte Carlo simulation. It shows how wide or narrow the range of possible outcomes is under stated assumptions, and where the projection is most sensitive to return variability.
How FutureClear models it
The simulation
Each simulation run generates a sequence of annual investment returns, drawn randomly from a log-normal distribution calibrated to the scenario's growth rate and volatility assumption. The simulation runs this process thousands of times. The result is a large set of possible paths — from runs with persistently poor returns to runs with persistently strong ones.
What the percentile bands mean
The coloured bands represent percentiles — ranked outcomes across all simulation runs at each point in time:
- p10: 90% of simulated paths produced a better outcome than this line. Represents a poor return environment.
- p25: 75% of paths were better. Below average, but less severe than p10.
- p50 (median): Half of paths were better, half were worse. The middle outcome, not a central prediction.
- p75: 25% of paths were better. An above-average return environment.
- p90: 10% of paths were better. A very favourable return environment.
The fan widens over time because small annual return differences compound into large differences over many years.
These bands describe the distribution of simulated outcomes. The p10 line does not mean "there is a 10% chance of this happening" — it means 10% of the randomly generated return paths produced a worse result than this.
Why log-normal returns
The simulation generates returns using a log-normal distribution. This is the standard model used by consumer retirement tools — including Vanguard, ProjectionLab, Boldin, and T. Rowe Price — because it:
- Prevents mathematically impossible negative portfolio values (a fund cannot lose more than 100%)
- Reflects the observed skew in equity returns, where large positive years are more common than symmetrically large negative years
- Calibrates well to historical return data for central estimates (the p25 to p75 range)
- Uses a transparent parameterisation: one mean return and one volatility figure
Assumptions and limitations
The log-normal distribution understates the probability of extreme outcomes at both ends — very large losses and very large gains. Real market returns have heavier tails than a log-normal distribution predicts (excess kurtosis).
In practice, the p10 and p90 bands may not fully capture how bad or how good the most extreme outcomes could be. The central estimates (p25 to p75) are well-calibrated to historical data. The limitation applies primarily to the outer edges of the fan.
Professional adviser tools (such as EValue and Moody's Analytics) use more complex stochastic models — regime-switching volatility, ARCH-type processes, or historical bootstrapping — that better capture tail behaviour. These models are designed for regulated financial advice and require assumptions that are difficult to validate or explain in a consumer self-service context.
For the purpose of understanding the range of possible outcomes and where a scenario is most sensitive to return variability, the log-normal model provides a sound and transparent foundation. The outer percentile bands indicate the direction and relative magnitude of extreme outcomes — they are not precise probability boundaries.
This property is documented by the Institute and Faculty of Actuaries (IFoA) and widely discussed in actuarial literature.