How big your pension should get before the tax arbitrage thins out, why the year you contribute matters, when across your career to put the money in, and what the wrapper is worth against an ISA.
The pot that funds an income sitting exactly at the top of the 20% band every year from withdrawal age to planning age, uses the whole tax-free lump-sum allowance by the end, and depletes to zero at that point. Bigger, and the marginal pound comes out at 40%; smaller, and basic-rate room goes unused.
Each year the pension must supply headroom = band top − state pension − other income of taxable income. Under phased UFPLS a quarter of every withdrawal is tax-free, so the gross withdrawal is headroom ÷ 0.75 and the tax-free slice headroom ÷ 3. The target is the present value of that stream at the in-retirement growth rate.
Tax-free cash is simply a quarter of each withdrawal, running until the lump-sum allowance is used up — shown in the stats as the age the cap is exhausted. After that point withdrawals are fully taxable, so the pot drains more slowly and the chart shows a visible step down. Because the pot keeps growing while it is drawn, total withdrawals across retirement far exceed the pot itself, so phasing reaches the cap from a pot well below four times its size — and extracts more tax-free cash than taking the maximum lump sum upfront would. Where the withdrawals do not reach the cap, the remainder is genuinely out of reach: you can only ever take a quarter of what you crystallise, and there is no more pot to crystallise. The model does not inflate the target to chase it.
In short: Because the pot keeps growing while it is drawn, total gross withdrawals exceed the starting pot, so phasing normally exhausts the cap part-way through retirement — and extracts more tax-free cash than taking the maximum lump sum upfront would. Where the withdrawals do not reach the cap, the remainder is genuinely out of reach: you can only ever take a quarter of what you crystallise, and there is no more pot to crystallise. The model does not inflate the target to chase it.
Income tax and NIC thresholds are frozen in cash terms until April 2031 and indexed afterwards at your chosen rate.
Your salary is about to cross a tax threshold. You could save money by delaying this year's contribution to the higher-relief year. Same pension, same spending in each year — and extra money to spend or put in your ISA.
| Year | Salary | A pension | B pension | A relief | B relief | A take-home | B take-home | ISA flow | B spendable |
|---|
Route A contributes the chosen percentage of salary in both years. Route B contributes nothing in year 1 and enough in year 2 that the total landing in the pension — employer top-up included — is identical. Relief is applied band by band as the contribution lowers income, so a large year-2 sacrifice earns a blended rate.
In year 1 route B keeps more take-home than route A; that difference goes into an ISA. In year 2 route B keeps less, and draws the ISA down to cover the gap so spending matches route A exactly. What is left in the ISA afterwards is the gain — money you have for the same pension and the same two years of spending. No return is assumed on the ISA over the single year, so this understates it slightly.
Building on the idea in tab 02, this page models strategically moving pension contributions to your highest-relief years across your career, and compares that with a standard fixed percentage contribution. The difference is cash you can spend — or invest in an ISA as another retirement asset. Before all this, remember to capture your full employer match as far as you can.
Route A contributes your chosen percentage of salary every year to withdrawal age. Route B always captures the employer match, and places the rest in the years with the highest relief. Both are solved so the value of the pension at withdrawal age is identical, employer contributions, match and top-up included. The difference is take-home you keep. The headline treats it as invested in an ISA at the accumulation growth rate from the year it was saved, and reports its value at withdrawal age; the today's-money figure deflates that by inflation.
A contribution grows to withdrawal age at the same rate whether it sits in the pension or in the ISA you would otherwise have kept it in, so the growth factor drops out of the comparison and only the relief rates matter. That is a deliberate simplification. If you smooth your spending by drawing on the ISA during the years of heavy pension contribution — even though your income is higher then — you may reasonably hold that ISA more conservatively until it is needed. A lower return on it means the true gain is somewhat smaller than the model states.
The tab 01 target is the pot that keeps every withdrawal inside the 20% band. Contributions above it come out at 40%, with no tax-free element once the cap is used. You may still judge that worthwhile in a very high-relief year — 62% in, 40% out is a positive trade — and choose to keep contributing past the target. Tab 04's matrix prices exactly that case; this tab does not, so that the two questions stay separate.
Bisection on the every-year contribution rate until route A reaches the tab 01 target, given the employer's contributions. If a match is set, the rate needed to capture it is shown separately. If the employer's contributions alone already reach the target, the required rate is zero beyond the match and no contributions of your own are shown.
1 − (take-home given up) ÷ (your sacrifice + NIC top-up): the relief the tax system and the employer's NIC saving give you on your own money. The regular employer contribution and the match are excluded — free money, not relief. They still count when the two routes' pensions are equalised.
Three inputs build a real path — growing to the plateau age, flat afterwards — converted to nominal with inflation. Unpinned years in the edit table follow the three inputs; ticked years are pinned. Thresholds are frozen until April 2031 and indexed afterwards at the tab 01 rate.
You will sometimes see a year with no contribution sitting next to a very large one. That is usually the annual allowance at work rather than an arbitrary spike. Reaching the 60% band from a high salary can need more headroom than a single year's £60,000 allows, so the model contributes nothing one year, banks that allowance as carry forward, and uses the combined total the next year to make the descent in one go. Where no such descent is needed and the model is genuinely indifferent between years, it now spreads the money evenly across them rather than filling one arbitrarily.
Think of each year as offering pension at a price: the take-home you give up per pound that lands in the pot. The price is lowest in high-relief years. The complication is that within a single year the price changes as you go — from a salary in the 45% band, the first pounds are relieved at 47%, the stretch down through the £100,000–£125,140 taper band at 62%, and anything below that at 42%. So the model works out, for each year, the cheapest average price for every possible depth of contribution, treating a descent through the expensive 47% slice into the cheap 62% band as a single package priced at its average. It then buys pension in whichever year and package is cheapest, moves to the next cheapest, and stops when route B holds the same pension as route A. Where several years offer exactly the same price the model has no reason to prefer one over another, so it spreads the contribution evenly across all of them rather than filling one to capacity first — the same total cost, but an allocation that reflects the indifference rather than an accident of ordering. Where several years offer exactly the same price the choice between them makes no difference to the total, so rather than filling one and leaving its neighbours empty the model spreads the contribution evenly across all of them, passing a year's share on to the others once it runs out of room. Money can be moved into any later year, and — if allowed — into an earlier one.
Each year's total contribution must fit within that year's allowance plus unused allowance from the previous three years, oldest first. The allowance is £60,000, tapered where threshold income exceeds £200,000 and adjusted income £260,000, to a floor of £10,000 — reached only at £360,000 of adjusted income. Salary sacrifice is added back into threshold income. These figures are held frozen unless you switch on indexation.
Tab 03 assumes today's rules last your whole career. They might not. This page asks what a rule change would cost you — measured two ways. Against doing nothing clever: was the timing strategy still better than a flat contribution? And against perfect foresight: how much did you lose by not seeing it coming? Then it weights those losses by how likely the change is, year by year.
Every scenario is scored by running three contribution plans through the same post-change tax rules. The flat contribution puts the same percentage in every year and never reacts. The timed, responsive plan is exactly what tab 03 produces: it allocates in ignorance until the year the change lands, then carries on under the new rules. The foresight plan knows from day one and allocates optimally against the rules as they will actually be. The gap between the first two is what timing was still worth; the gap between the last two is the cost of not knowing.
The timing strategy front-loads contributions into high-relief years, and a rule change almost always removes relief rather than adding it. So a plan built in ignorance has usually already banked the benefit before the change arrives. Regret is largest when the change lands early, before much has been placed.
Rather than a single guess, the model uses an annual hazard rate — the chance the change happens in any given year — rising in a straight line from today's level to the level you expect by the end of your career. The chance of surviving to year k without the change is the product of one-minus-hazard over the preceding years, so the probability it lands exactly in year k is that survival probability multiplied by that year's hazard. This shape matters: a low hazard now with a higher one later means the change is far more likely to hit mid-career than next April, which is exactly when a long-horizon saver is most exposed.
The expected gain is then each outcome weighted by its probability, including the case where nothing changes at all. Expected regret is the probability-weighted shortfall against foresight — the price of the uncertainty itself, and the number worth comparing against the headline gain on tab 03.
Take £100 of gross pay and route it two ways. Into an ISA it arrives net of income tax and, under salary sacrifice, employee NIC. Into a pension it arrives whole and is taxed on the way out. The cell is (pension after tax out) ÷ (ISA after tax in) − 1. Even at the same rate in and out the pension wins, because a quarter of every withdrawal is tax-free.
A large single contribution is relieved band by band as it lowers income — from the 45% band, down through the 60% taper, and on into the 40% band — so it earns a blended rate below the headline. The blended row applies that rate to the same four exit scenarios, using the tab 01 tax year and the selected NIC treatment.
The top row is a contribution made while your income is inside the personal allowance: there is no relief going in, so an ISA keeps the whole £100 and the pension can at best draw level — it wins nothing and loses whatever is taxed on the way out. The first column is the mirror image: a withdrawal taken entirely inside the personal allowance, taxed at nothing. Real retirements sit between the two, because the state pension and other income fill the allowance first.