The wrapper decision that costs more than the platform decision
Britain holds around GBP 872 billion in ISAs, and almost all the argument is about which firm to open one with. That is the smallest of the sums involved.
What is the average ISA actually holding?
HMRC counted 22.6 million adult ISAs in April 2026, against an ISA market worth around GBP 872 billion. Both are April 2026 HMRC figures. A tax authority’s count is not the sort of number a research site can go out and produce, and The Investors Centre is one of those: it opens and funds live accounts with its own money to test UK trading platforms rather than compiling rankings from providers’ published fee schedules. Divide one by the other and you get just under GBP 38,600 an account.
Treat that as an order of magnitude and no more. Balances of this kind are heavily skewed, so the mean is dragged up by a small number of very large accounts and sits well above the middle one. The typical ISA holds considerably less, and an average is the one statistic that cannot say how much less.
Three boxes, and GBP 20,000 to put in one of them
Wherever it sits, that balance is in one of three places: a cash ISA, a stocks and shares ISA, or an ordinary account with no wrapper round it. What follows takes GBP 20,000 and asks what each decision is worth over twenty years.
The cash ISA and the stocks ISA differ by more than their tax
Both are free of tax on the income and gains held inside them, so the tax difference between them is close to nothing, which makes the choice everybody calls a wrapper decision not really a tax decision. It is an asset decision wearing a tax label. One box holds a savings rate and the other whatever a portfolio does, and the treatment they are named after cannot tell them apart.
Worth pinning down first, because it moves the question. Asking which ISA is better is asking what you want the money to be doing, which is about your timetable and not about HMRC. The genuine tax decision is the third option, wrapped against unwrapped, and it earns less than its share of the talk.
Two sums, and one of them is ten times the other
Say two providers differ by a fifth of a percentage point a year in total charges, an illustrative gap and not a measured one. Put GBP 20,000 in each at an assumed 5% a year. Over twenty years 1.05 to the power of twenty is 2.6533, giving GBP 53,066. At 4.8% the multiplier is 2.5540, giving GBP 51,081: a gap of GBP 1,985, under 4% of the balance.
Now hold the provider still and change what is inside. Cash at 2.5% has a multiplier of 1.6386, so GBP 20,000 becomes GBP 32,772 against GBP 53,066 at 5%. The difference is GBP 20,294, more than ten times the provider gap, and it comes from a decision many people take in a minute by picking whichever option sounds safest.
Run it to thirty years and the gap is about GBP 44,500. Five per cent is an assumption, not a forecast.
What does it cost to leave an unwrapped account?
Here is the sum that genuinely is about tax. Same GBP 20,000, same 5%, no wrapper. The balance still reaches GBP 53,066, so the gain is GBP 33,066. The 2025/26 exemption from capital gains tax is GBP 3,000, with 18% and 24% by band above it. Sell the lot in one tax year and GBP 30,066 is taxable: about GBP 5,410 or GBP 7,220.
Sell across several tax years and the exemption applies afresh each time, so a patient exit takes much of that gain out for nothing. The unwrapped penalty runs from close to zero to about GBP 7,220 on identical holdings, decided by the order you sell in. It also omits the annual part: dividends and interest are taxable outside a wrapper in the year they arrive.
| The decision | The two versions compared | Difference after 20 years | When you find out |
| How much you add | GBP 20,000 once, or plus GBP 200 a month | About GBP 82,000 | Monthly, in your bank |
| What it is invested in | 2.5% against 5% a year | About GBP 20,300 | In the 2040s |
| When you start | Now, or in five years | About GBP 11,500 | Never, in a form you can act on |
| Wrapped or unwrapped | Identical growth, CGT on exit | Nothing to about GBP 7,220 | In the tax year you sell |
| Which provider holds it | 0.20 points a year of charge | Just under GBP 2,000 | Yearly, on the statement |
Five decisions, two of which nobody thinks of as decisions at all. Every row assumes the same rate for twenty years. Widen the charge gap in column two far enough and the bottom row climbs past the others.
Where cash earns its place
None of that makes cash a mistake. The arithmetic ran twenty years and most of the work cash does runs two. A deposit due in 2029, a bill in January, the spending held back so a bad year cannot force a sale: for those the only property that matters is that the number does not move, and only cash offers it. Cash held for a two-year purpose is doing its job precisely. Cash held for twenty because it was the default on the page is doing something else, and no published figure separates the two, so nobody writing about that GBP 872 billion can say what share of it sits in the wrong box.
Returns arrive in an order, and the order is not in the sum
Every multiplication above drew a straight line through something jagged. Returns arrive in a sequence rather than at five per cent a year, and the sequence is invisible to the compound interest formula, which assumes nothing enters or leaves. For somebody who adds and takes nothing for twenty years, that does not matter.
For everybody else it matters enormously. Two people with the same average return finish level if neither touches the account, and visibly apart if one was drawing an income or forced to sell in a bad year. Which is why access belongs before the arithmetic: anyone who might need the money on somebody else’s date is not who these figures describe.
One of these decisions can be settled by Friday
The provider decision is measurable now. Open an account, pay money in, deal, take money out, read the statement, and the cost is a fact within the week. The other two can only be modelled, and the model is settled in the 2040s by a market that has not done anything yet. Attention follows what can be settled: a fact about measurement rather than motives.
Where the research runs deep and where it runs thin
That shapes research, including the research quoted here. The work the team at The Investors Centre publishes runs deepest at the trading end, where it opens and funds live accounts with its own money to test UK platforms rather than compiling rankings from published fee schedules, and thinnest at the wrapper end, where there is nothing to open and nothing to fund. The ISA figures in this piece are HMRC’s, gathered rather than measured, and that is the honest label for them.
Which sum to do first
In descending order of effect: how much goes in, what it is invested in, when it starts, how the exit is taxed, and last, who holds the account. Close to the reverse of the order they get asked in, and the reversal is not irrational: the last has an answer tonight and the first four ask you to commit. None of it needs a spreadsheet. Two multiplications and a subtraction fit on the back of a bank statement. Do the provider sum second: it is the only one whose answer stays roughly right when every other decision has gone wrong.
…