Key takeaways
- An extra 0.50 percentage points of annual cost takes 13.96% of the final pot over 30 years, and that share holds whatever the growth rate turns out to be.
- Hargreaves Lansdown charges 0.35% on shares and ETFs, capped at £12.50 a month. Above £100,000 that beats Vanguard's 0.15%, which climbs to a £375 ceiling.
- Basic-rate relief at source turns £80 of net pay into £100 inside a pension, a 25% uplift. A 0.35% charge would take roughly 64 years to erode it.
- The FCA's 2017 market study found no clear relationship between charges and the gross performance of UK retail active funds, and some evidence dearer funds did worse after fees.
- Morningstar's cheapest quintile of US funds returned 8.7% a year over the decade to December 2024. The priciest quintile returned 5.6%.
Paying higher fees is defensible when the extra buys something priced in advance
Cost is the one part of investing you set rather than forecast, so the default answer is that cheaper wins. Then you find a platform charging more than a rival and apparently costing you less, or a fund whose ongoing charges figure is triple the cheapest option on the same index. When is the dearer bill the right one?
On published schedules and regulatory evidence, there's a narrow set of cases, and they share one feature. Paying higher fees survives scrutiny when the extra buys something you can price before you hand it over: a charge that stops climbing, a tax uplift written into the rules, a second charge that falls. The cases that fail share the opposite feature. There, the extra buys a forecast, and the forecast is almost always performance.
That's the test. The rest of this is arithmetic, starting with the hurdle, because the hurdle is steadier than most people expect.
What an extra layer of cost has to earn back
A percentage charge on assets doesn't subtract from your return. It multiplies it down. Each year you keep what's left after the fee, so after a run of years you hold that fraction compounded.
The useful consequence is that the share of the pot you give up doesn't move with market returns at all on this arithmetic. It depends on the size of the gap and the number of years. Here's the cost drag over time, expressed as the percentage of the final pot handed over rather than kept.
| Years held | Extra 0.25pp a year | Extra 0.50pp a year | Extra 1.00pp a year |
|---|---|---|---|
| 10 | 2.47% | 4.89% | 9.56% |
| 20 | 4.88% | 9.54% | 18.21% |
| 30 | 7.23% | 13.96% | 26.03% |
| 40 | 9.53% | 18.17% | 33.10% |
The chart above plots the middle column across the whole span. Put money on it: £100,000 left alone for 30 years at 5% a year, the highest intermediate projection rate the FCA lets a firm use for a personal pension, grows to £432,194 before platform and fund costs. Add 0.50 percentage points of annual charge and £371,853 is left, a shortfall of £60,341. For products other than pensions the FCA's ceiling is lower, at 4½%, which changes the pot but not the fraction.
So an extra half a point over three decades has to buy you something worth about a seventh of everything you end up with. That's the number any defence of a higher charge has to clear. Very few things clear it. Some do.
Test one: the charge stops. A capped 0.35% costs less than a 0.15% that keeps climbing
Headline percentages are the worst possible way to rank platforms, because several of them stop. Hargreaves Lansdown's stocks and shares ISA charges 0.35% on funds up to £250,000, and on shares, ETFs and investment trusts it charges "0.35% (capped at £12.50 per month)". That cap is £150 a year, and the firm states it plainly: from 1 March 2026, "you will never pay more than £150 per year in account charges for holding shares in each of our SIPP, Stocks and Shares ISA or Fund and Share Account".
Vanguard's UK self-managed accounts charge "£4 a month (£48 a year)" under £32,000 and "0.15% a year (max £375 a year)" at £32,000 and above. On the label, Vanguard is less than half the price.
Run the two schedules against an ETF portfolio and the ranking inverts. The Hargreaves cap starts binding at £42,857, because 0.35% of that is £150. Vanguard's 0.15% reaches £150 at exactly £100,000. Above that point, the platform advertising 0.35% charges less than the platform advertising 0.15%. At £300,000 of ETFs the bills are £150 and £375. Here, paying higher fees on paper is the cheaper outcome, and you can verify it before you move a penny.
AJ Bell's ISA schedule, effective from 10 July 2026, does the same thing at a different level: "0.25% (maximum £3.50 per month)" on shares, which is £42 a year, binding above £16,800. The identical 0.25% on funds has no such cap below £250,000, so a £250,000 fund holding costs £625 a year and the same money in shares costs £42. One rate, two prices, and the gap is close to fifteen times.
None of this makes a cap free. Dealing charges sit on top of the account charge on both platforms, so a heavy trader unwinds part of the saving. The point is narrower: a percentage tells you the price of the first pound, not the price of the last one, and the crossover is published. Our piece on platform fees walks the flat-versus-percentage version of the same crossover.
Test two: the wrapper hands back more than the platform takes
The second defensible case is a charge attached to a tax rule. GOV.UK describes relief at source in one line: "your pension provider claims tax relief from the government at the basic 20% rate and adds it to your pension pot". Put in £80 of taxed income and £100 lands, an immediate 25% uplift, with no market view attached to it.
Set that against a charge. A 0.35% annual fee erodes a pot by 0.35% of whatever it has each year, so it needs about 64 years to eat a 25% head start. Nobody's horizon is 64 years, which is why a pension that charges and a general account that doesn't aren't really competing on cost. They're competing on access, and one of them starts a quarter ahead.
The same shape appears wherever a charge sits next to a statutory benefit rather than a hoped-for return. What matters is that the benefit is defined in advance and doesn't depend on anyone's skill. Where it isn't defined in advance, the test fails, which brings us to the case people actually mean when they ask this question.
Test three: the extra buys performance, and this is the one the evidence refuses
The Financial Conduct Authority put the question directly in its Asset Management Market Study final report, published in June 2017. It asked whether investors "may choose to invest in funds with higher charges in the expectation of achieving higher future returns", then tested it on UK retail active funds.
The finding was blunt. "Our additional analysis suggests that there is no clear relationship between charges and the gross performance of retail active funds in the UK. There is some evidence of a negative relationship between net returns and charges. This suggests that when choosing between active funds investors paying higher prices for funds, on average, achieve worse performance." Across the three equity categories it analysed, where there was price variation, "the majority of results showed no statistically significant correlation between price and performance".
Morningstar's Mind the Gap 2025 study, published in August 2025, sorted US funds into fee quintiles for the decade ended 31 December 2024. The cheapest quintile earned 8.7% a year in aggregate total return. The priciest quintile earned 5.6%. On a dollar-weighted basis, which follows the money investors actually had in at the time, the same quintiles earned 7.6% and 3.8%.
Those are two different bodies, two continents and two methods, and neither found a price you could pay to buy return. The SPIVA scorecard covers the survivorship-corrected version of the same question in more depth. For this piece the conclusion is the narrow one: a fee justified by expected outperformance is a fee justified by a forecast, and the forecast has no support in the published data.
The strongest case for paying more, and where it thins out
The serious counter-argument isn't about fund selection. It's that an adviser's fee buys behaviour, not alpha. Vanguard has made that case for 25 years. Its 2025 retrospective states that its 2014 Quantifying Advisor's Alpha research "found that advisors following wealth management best practices can add up to, or even exceed, 3% in net returns for their clients".
If that figure is real, it dwarfs the hurdle. An extra 1.00 percentage point of cost gives up 26.03% of a 30-year pot, and 3% a year would more than cover it.
Two things about the number. It comes from the firm selling the funds those advisers use, and it's a model of best practice rather than a measurement of outcomes. Vanguard frames it as what advisers "can add", not what advised clients received. Second, Morningstar found the strongest version of the behavioural argument and then refused to over-read it. Cheaper funds did show smaller investor return gaps, but Morningstar's own takeaway warns that it "would be overly reductive to conclude that investors in inexpensive funds will capture more of their funds' total returns than investors in costlier funds", and notes that passive fund investors "were almost as susceptible, and in some cases more prone, to mistiming their transactions than investors in costlier active funds".
The honest reading is that a behavioural fee can clear the hurdle for some people and can't be verified in advance for any particular one. That's a different kind of claim from a published cap. Our piece on the behavior gap works through what mistimed transactions have actually cost.
Where a plan upgrade stops being arithmetic and starts being hope
Tiered platform plans are where the test gets its cleanest failure. Interactive investor's published schedule runs three service plans: Core at "£5.99 per month", Plus at "£14.99 per month" and Premium at "£39.99 per month". Premium includes free fund trades where Plus charges £1.49.
The step from Plus to Premium costs £300 a year. Recovering that in dealing charges alone needs 201 fund trades a year, which is roughly four a week. Premium also cuts currency conversion to 0.25% against 0.75% on the first £50,000 for Plus, so a portfolio converting a lot of sterling has a real case. A portfolio that doesn't convert anything is buying £300 of nothing a year, and over 30 years the drag from that runs alongside every other layer in the total cost of owning a fund.
Currency charges themselves tier the same way. AJ Bell's schedule charges 0.75% on the first £10,000 converted, 0.50% on the next £10,000 and 0.25% above £20,000. The small conversion pays three times the rate of the large one for an identical service, which is another reminder that a single advertised percentage rarely describes what anyone actually pays.
Fund charges spread nearly as widely inside a single low-cost range. Vanguard's own UK ISA lists ongoing charges figures from 0.06% to 0.79%. That 0.73 point spread costs 19.73% of a 30-year pot, so choosing badly inside one provider's shelf can matter more than choosing between providers.
What this evidence can't tell you
The fee schedules are dated snapshots and nothing more. Hargreaves cut its account charge to 0.35% from 1 March 2026, AJ Bell's terms took effect on 10 July 2026, and any of them could move again before you read this. Every crossover above is arithmetic on those specific published numbers, not a ranking of firms.
The performance evidence has real limits too. The FCA's study examined UK retail active equity funds in a defined period ending before its June 2017 report, and the regulator was careful to say the finding was not an argument that "passive funds were preferable to active funds". Morningstar's fee quintiles are US open-end funds and ETFs over one decade, and its own reading is that cheap funds tend to be used in contexts where investors go astray less often, which is not the same as price causing return.
The compounding table assumes a constant charge on a lump sum with no contributions, no withdrawals and no tax events. Real portfolios have all three, and a charge levied on a growing balance behaves differently from one levied on a static one. The direction is robust; the decimal places aren't.
These figures cannot tell you what any single fund or platform does next. They're dated schedules and a sample of past returns, not a forecast.
Nothing here touches whether a given fund tracks what it says it tracks, which is a separate cost that doesn't appear on any schedule.
What would change the conclusion
If a cap were removed. Every case in test one rests on a published ceiling. Take away the £12.50 monthly cap and Hargreaves' 0.35% on a £300,000 ETF holding becomes £1,050 a year rather than £150, and the ranking flips back to the headline order.
If the pension uplift changed. The 64-year erosion figure is a function of 25% and 0.35%. Basic-rate relief at source is set by rule, not by market, and the whole of test two moves with it.
If someone found a priced, contractual link between charge and return. That's what a performance fee claims to be, and it's the version of test three worth watching, because it converts a forecast into a term of business. The FCA's final report gives objectives, benchmarks and performance a chapter of its own in the remedies package.
The practical thing to watch isn't the advertised percentage. It's the point where each charge stops, and where your balance sits relative to it. That's a two-column sum on the back of an envelope, and it beats any ranking table, because the answer depends on a number only you have.