What's Inside a Scenarios Simulation

A Black Box Is Not a Plan

Every retirement tool gives you a number. Almost none of them tell you how it got there.

You enter your details, the screen does something, and a confident projection appears. Maybe a green chart. Maybe a percentage. You're meant to trust it because it looks polished — not because anyone has shown you the maths.

That's the opposite of how a planning tool should work. If a number is going to influence whether you retire at 60 or 65, you deserve to know exactly how it was produced. So this post does what most fintech blogs won't: it walks through what actually happens inside a Scenarios simulation, from the moment you press run to the moment a chart appears.

If you'd prefer the reference version, the methodology page has every assumption laid out in tables. This is the narrative.

One Plan, One Thousand Futures

When you build a plan, Scenarios doesn't run it once. It runs it 1,000 times.

Each of those runs is a complete, self-contained simulation of your financial life from today until your chosen end age. Each one rolls a different sequence of investment returns and a different inflation path. Some are kind to you. Some are punishing. Most are somewhere in between.

The output isn't a single forecast — it's a distribution. The fan chart you see in the dashboard is the spread of those 1,000 futures, ranked into percentile bands. The 10th percentile shows what happens in tougher markets. The 90th shows what happens in friendlier ones. The median is the middle of the pack.

That shift — from one number to a range — is the whole point. A retirement plan that only works in one specific future isn't really a plan. It's a hope.

What Happens in a Single Simulated Year

Inside each of those 1,000 simulations, the engine moves forward one year at a time. Every year follows the same sequence:

1. Generate this year's market return, broken down by asset class.
2. Generate this year's inflation rate.
3. Apply your contributions while you're working, or your withdrawals once you're drawing down.
4. Apply the full UK tax engine — Income Tax, CGT, dividend tax, savings allowances.
5. Roll account balances forward, including dividends, interest, and the State Pension if applicable.
6. Move to the next year.

Repeat for 30, 40, 50 years. Then do it 999 more times with different random seeds.

Each step inside that loop is doing real work. Here's what's underneath.

Returns Aren't Drawn from Thin Air

Each asset class — global equities, UK equities, bonds, property, gold, cash, and "other" — has a calibrated long-run expected return and volatility. Global equities run at a 7.0% expected return with 18% volatility. Gilts and aggregate bonds run at 3.0% with 6% volatility. Cash sits at 2.0% with almost no volatility. These figures come from widely used capital market assumptions — long-run global equity studies, historical Gilt yields, and the kind of inputs institutional planners feed into their own models.

But asset classes don't move independently. When global equities fall, UK equities almost always fall with them. When equities crash, bonds sometimes hold or rise. The model captures these relationships through a 7×7 correlation matrix calibrated to post-2000 market regimes.

The maths to make this work is something called Cholesky decomposition. It's not as scary as it sounds. The engine generates seven raw random numbers each year, then multiplies them by the Cholesky matrix to produce seven correlated random numbers. Those become this year's asset returns. The end result: when global equities are simulated to fall 20%, UK equities tend to fall hard too, but bonds may hold steady. That's how realistic diversification gets baked in — not by assumption, but by mathematics.

This is also why a 60/40 portfolio in the simulation has lower volatility than you'd naively expect. The diversification benefit emerges from the correlation matrix, not from a fudge factor.

Inflation Is a Variable, Not a Constant

Most retirement calculators use a single inflation number — usually 2.5% — and apply it for every year of your retirement. That's a fantasy. UK inflation has been below 1% and above 11% in the last decade alone.

Inside Scenarios, each year's inflation is drawn from a normal distribution centred on a mean (default 2.5%) with its own volatility (default 2%). One simulation might run mostly tame. Another might give you a 6% inflation spike in year three of retirement, exactly when you don't want one.

This matters because real spending power isn't constant. The "Real" view in the dashboard deflates every figure by each simulation's actual inflation path, so you're seeing pounds in today's terms. The "Nominal" view shows the actual currency amounts. Both come from the same simulations.

For couples, both partners share the same economy in each simulation — same inflation, same returns. They're not living on different planets.

A Real UK Tax Engine, Not an Average

This is where most tools fall apart. Tax in the UK is genuinely complicated, and pretending it's a flat percentage produces dangerous results.

Inside the engine, every taxable pound is run through HMRC's actual rules for 2026/27:

• Income Tax — full bands, including the 60% effective marginal rate between £100,000 and £125,140 caused by personal allowance taper. Scottish bands also supported.
• Capital Gains Tax — applied only to GIA disposals, with cost basis tracked per account. The £3,000 annual exemption is used first. Unrealised gains compound tax-free.
• Dividend tax — using the actual dividend yield of your portfolio's equity allocation (1.8% for global, 3.5% for UK), with the £500 dividend allowance applied.
• Savings income — interest from bonds and cash in GIAs uses the Personal Savings Allowance (£1,000 / £500 / £0 by tax band) and the £5,000 starting rate for savings.
• Pension rules — £60,000 Annual Allowance, £10,000 Money Purchase Annual Allowance once flexible access is triggered, £268,275 Lump Sum Allowance for tax-free cash, Normal Minimum Pension Age 57 from April 2028.
• State Pension triple lock — modelled as CPI + 0.5%, applied on top of each simulation's stochastic inflation path.

You can also choose whether tax bands inflate with your inflation rate or stay frozen — the difference between modelling fiscal drag and modelling current government policy. Over 30 years, that single choice can shift the result by tens of thousands of pounds.

A Withdrawal Strategy That Actually Optimises

When you reach retirement inside the simulation, the engine doesn't just drain accounts in some arbitrary order. It applies a tax-optimised waterfall every year, filling cheaper tax bands before more expensive ones:

1. Guaranteed income comes in first — State Pension, defined benefit pensions, annuities, net rental income. Taxed as income.
2. Pension tax-free cash is taken from your remaining Lump Sum Allowance. Completely tax-free.
3. Pension income to fill the Personal Allowance — enough to use the remaining £12,570 at 0% tax.
4. ISA — tax-free withdrawals to bridge the gap without adding to taxable income.
5. Remaining pension income at your marginal rate (20%, 40%, or 45%).
6. GIA — using the annual CGT exemption first, then taxed at 18% or 24%.
7. Cash — the final buffer.

For couples, both partners draw independently — so both Personal Allowances and both ISA wrappers are used every year. The Withdrawals tab also lets you switch to PCLS or UFPLS, enter any LSA already used elsewhere, or override everything with a fully custom drawdown order.

This isn't a cosmetic feature. The order in which you take money out is one of the largest levers in UK retirement — bigger, in many cases, than your investment choice.

Spending Doesn't Have To Be Static

Most retirement models assume you take the same inflation-adjusted amount every year regardless of what markets do. Real retirees don't behave that way — they tighten up after a bad year and ease off after a good one. Scenarios lets you model that directly.

Instead of fixed real spending, you can switch the simulation to one of two dynamic strategies:

• Guyton-Klinger guardrails. The engine compares your current withdrawal rate to the rate you started retirement with. If markets fall and your rate climbs more than 20% above that starting point, spending is cut by 10%. If markets are kind and your rate drops more than 20% below it, spending is boosted by 10%. The defaults are the academic ones, but the thresholds, the boost and cut sizes, and the active window are all editable.
• Floor and ceiling. A simpler rule that caps how much your real spending can change in any given year — by default no more than +10% up or −15% down — letting you smooth out the ride without the rate-based mechanics.

There's also an optional cash buffer that holds two or three years of spending outside the invested portfolio, so withdrawals come from cash after a bad year instead of forcing sales of equities at a low. It can be set to replenish from the portfolio in good years.

These features won't dramatically change a healthy plan, but they meaningfully change the tail. A plan that fails 5% of the time on fixed spending often fails much less with guardrails, because the strategy mechanically responds to the exact scenarios that cause failure — bad sequences early in retirement.

What "Plan Confidence" Actually Means

After all 1,000 simulations finish, the engine counts how many of them successfully reached your end age without running out of money. That percentage is your plan confidence.

A plan that survives in 950 of 1,000 simulations gets a confidence score of 95%. That doesn't mean your retirement is "safe" — it means 50 plausible futures, drawn from the model's assumptions, still ran out. The fan chart shows the distribution: where the median lands, where the 10th-percentile pessimist ends up, where the 90th-percentile optimist sits.

A wide fan means the outcome depends heavily on market conditions. A narrow fan — usually from larger bond allocations or shorter horizons — means more predictability but probably lower upside.

90% and above is generally considered robust. 70–89% is moderate risk. Below 70% means a meaningful chance of falling short. None of these are rules — they're a vocabulary for talking about uncertainty honestly.

The Same Plan Always Gives the Same Answer

One detail worth knowing: the random number generator is deterministic. The seed is derived from a hash of your plan inputs. Change anything — your contribution, your end age, your asset mix — and the simulation generates a genuinely different set of 1,000 futures. Don't change anything, and you get the same answer every time.

This is deliberate. Some tools re-roll the dice every time you press run, which encourages people to keep clicking until they see a result they like. That's not analysis, that's slot machine behaviour. In Scenarios, your plan deserves a stable answer. If you want a different answer, change the plan.

What We Don't Pretend To Model

Every model simplifies the world. Honest tools say what they leave out:

• Returns are normally distributed. Real markets have fat tails — extreme events happen more often than a bell curve predicts. The model may understate the very worst outcomes.
• Correlations are fixed. In real crises, correlations spike — everything falls together. Long-run averages may understate worst-case diversification failure.
• No mean reversion. Each year is independent. Markets may mean-revert over long periods, which would tighten the long-run spread.
• Tax rules are frozen at 2026/27, optionally inflation-adjusted. Future governments will change policy. Sometimes dramatically.
• No longevity model. The simulation runs to a fixed end age. If you plan to 95 and live to 100, you have a five-year gap.
• No annuity purchase mid-retirement and a simplified GIA model that doesn't include bed-and-ISA, capital loss harvesting, or section 104 pooling.

Despite the limitations, Monte Carlo simulation remains the most widely accepted approach for retirement planning under uncertainty. It's what institutional pension consultants, IFAs, and tools like Timeline, Voyant, and Guiide all use. The honest answer is that it's the best framework available — not a crystal ball.

Why Any of This Matters

The reason for showing this much detail isn't to impress anyone. It's that financial decisions made on the basis of opaque tools are decisions made on faith. And faith is a poor substitute for transparency when 30 years of your life are at stake.

If a tool can't tell you what's inside it, it probably shouldn't be used to plan your retirement.

Scenarios is built so you can interrogate the answer. Every assumption is documented. Every rule is editable where it should be. Every result is reproducible from your inputs. The full reference is on the methodology page.

If you want to see your own numbers run through this engine, you can build a plan for free and watch the same process unfold against your real position.