An average is a poor planning tool. A plan that works at a 7% average return may fail in a third of paths that average 7% but arrive in an unhelpful order. Simulation makes that spread visible.
Typical output: run 10,000 paths for a $1,000,000 portfolio withdrawing $40,000 a year for 30 years, and report that 86% of paths end with money remaining. The 14% failure rate is the number worth acting on, not the median ending balance.
The results are only as good as the assumptions. Drawing returns independently from a normal distribution understates crashes and ignores the way bad years cluster. Bootstrapping from actual historical blocks, and stress-testing with a deliberately poor first decade, gives a more honest picture. See sequence-of-returns-risk and risk-of-ruin.
Related: sequence-of-returns-risk, safe-withdrawal-rate, risk-of-ruin, backtesting, mean-variance-optimization