Key takeaways
- On a £10,000 UK profile priced at June 2026 Bank of England rates, highest-rate-first cost £4,094 in interest over 47 months; smallest-balance-first cost £4,781 over 50 months.
- That £687 gap is 6.87% of the sum borrowed, or £14.61 a month across the faster plan's 47-month term.
- Reshaping the same debts so the smallest one carries the highest rate makes both orders finish in 43 months for £2,722. The gap becomes £0.
- Across 1.4 million UK cardholders, 51.5% of surplus repayments went to the dearer card, where 97.1% would have minimised interest. Most people run neither strategy.
- Brown and Lahey measured a 13% performance gain from smallest-first ordering. This profile needs 10.8% more paid each month before the snowball breaks even.
Highest-rate-first saved £687 here, and nothing at all on two other profiles
Which debt do you clear first — the one charging the highest rate, or the one with the smallest balance?
The arithmetic isn't in dispute. Paying the dearest debt first, the avalanche, minimises interest by construction. The only live questions are how much the alternative costs, and whether anything buys that cost back.
Here is the size of it. On a £10,000 debt profile priced at the Bank of England's June 2026 effective rates, the avalanche cleared in 47 months and cost £4,094 in interest. The snowball — smallest balance first — took 50 months and cost £4,781. The gap is £687, and 3 months. That's 6.87% of the sum borrowed, or £14.61 a month across the faster plan's term. Both totals are ours, computed from published rates, and the assumptions are set out below.
£687 isn't nothing. It also isn't the difference between coping and drowning, and on two of the three profiles modelled here the gap was exactly £0.
The rates, the balances and the payment rule behind the £687
The Bank of England publishes what UK borrowers actually pay, monthly. In June 2026 the effective rate on interest-charging credit cards was 21.49%. On interest-charging overdrafts it was 21.17%. On new personal loans to individuals it was 9.67%. Those are market averages, not a quote for you.
The profile assumes three debts totalling £10,000:
- £5,000 on a credit card at 21.49%
- £3,000 of overdraft at 21.17%
- £2,000 of personal loan at 9.67%
It assumes a fixed £300 monthly budget covering all three, no new borrowing, and no interest-free balance transfer. Minimum payments follow the FCA's rule for credit and store cards, CONC 6.7.5R, which has applied to agreements made on or after 1 April 2011: at least the interest, fees and charges applied that month, plus 1% of the amount outstanding. Everything above the minimums goes to one target debt at a time. Interest accrues monthly at a twelfth of the annual rate.
Applying a card's minimum-payment rule to a personal loan is a simplification. A real loan carries a fixed instalment you can't shrink, which cuts the surplus both strategies have to work with. It cuts it equally, so it doesn't favour either order.
This profile is the worst case for the snowball by design. The smallest debt is also the cheapest, so the two orders disagree at every step. It's a common enough shape: a fixed-rate loan taken out years ago, sitting beside revolving debt that crept up afterwards.
Changing the shape of the debts takes the £687 to zero
Flipping the balances so the smallest debt is the dearest — £2,000 on the card at 21.49%, £3,000 of overdraft at 21.17%, £5,000 of personal loan at 9.67% — makes the two strategies one strategy. Both clear in 43 months for £2,722 of interest. The difference is £0.
Give someone three credit cards all priced at 21.49% and the same thing happens for a different reason: 52 months and £5,359 whichever end you start from. When rates are equal, order can't change the cost.
So the useful first question is about the debts, not about the borrower. The gap between the two strategies is a function of how far the rates diverge, and how that divergence lines up against the balances. Where the dearest debt is already the smallest, there's no trade-off to agonise over. The chart above shows the three orders run on the first profile, which is the only one of the three where the choice costs anything.
UK repayment data says most people run neither strategy
John Gathergood, Neale Mahoney, Neil Stewart and Joerg Weber studied monthly statements for 1.4 million UK cardholders from January 2013 to December 2014, linking multiple cards held by the same person. Among those holding exactly two cards, the average gap between the higher and lower rate was 6.3 percentage points — roughly a third of the 19.7% average purchase rate in the sample.
Minimising interest meant putting 97.1% of every payment above the minimums onto the dearer card. People put 51.5% there, which is close to indistinguishable from splitting it down the middle. 85% of the sample should have put everything on the high-rate card. 10% did.
What they did instead was match. The share of repayment going to each card tracked the share of balance sitting on it. Run that rule on the first profile and it takes 48 months and £4,340 — £246 worse than the avalanche, and £441 better than the snowball. The strategy most people actually follow has no name in the popular debate and sits between the two that do.
The money at stake in their data is smaller still. Annualised interest savings from a perfect reallocation averaged £65 for two-card holders and £248 for five-card holders. At the 90th percentile the figures ran from £167 to £927. For a typical two-card borrower, the entire prize for optimising is about £65 a year.
The field evidence for the snowball is one dataset, and it never sees an interest rate
David Gal and Blakeley McShane obtained records for a random sample of 5,943 clients of a US debt settlement firm, all enrolled before 1 January 2007. Their finding: closing accounts predicted getting out of debt, while the dollar balance of what was closed did not. Once you control for the fraction of accounts closed, the fraction of dollars paid adds nothing.
The number attached to it: a year after enrolment, a client who had consistently paid down the smallest balances was 14% more likely to complete the programme than one paying down balances at random.
That's real evidence and it deserves better than a shrug. It also carries three limits. It's correlational — nobody randomised the repayment order, so motivated clients and small-balance-first clients may be the same people. It comes from debt settlement, where accounts are negotiated down rather than repaid in full. And it never observes an interest rate, because the data doesn't contain them. Gal and McShane say as much: where rates and balances diverge far enough, the rational strategy wins, and their proposal is that people be told both things and left to choose. Behavioural findings in household finance often shrink when someone looks hard at them, which is the same pattern behind where the loss-aversion figure comes from.
Brown and Lahey measured the motivation at 13%. This profile needs 10.8%.
Alexander Brown and Joanna Lahey built the cleanest test of the mechanism. Subjects typed 150 cells of text arranged in five columns of unequal length, in ascending, descending or equal order, starting in December 2011. Ascending — smallest first — was fastest: 11.08 seconds a cell against 12.50 for descending, which they convert to a productivity gain of roughly 13%. 71% finished the whole task under the ascending order, against 48% under descending.
A second group of 70 subjects, in January 2013, was allowed to pick. 22% chose ascending, the least popular of the three orders. The arrangement that helped most was the one people wanted least — which is the standard argument for writing a rule down in advance, the same logic behind what the evidence shows about an investment policy statement.
Now price it. If the snowball really does buy a 13% lift in what gets paid each month, what's that worth against £687? On the first profile, a snowball payer needs £332.49 a month — 10.8% above £300 — to end up paying no more interest than an avalanche payer at £300. Matching the 47-month term takes only £309.66, a 3.2% lift.
10.8% is below 13%. On this profile, on those numbers, the snowball comes out ahead. That is the strongest counter-case against the arithmetic, and it's stronger than the avalanche camp usually admits.
Brown and Lahey ran the same exercise themselves and found the same shape. With two $10,000 loans, one fixed at 10% and $300 a month available, a 13% payment boost repaid both loans faster than rate-first ordering for every second-loan rate up to 16%, and lost from 17% upward. Measured on total money spent rather than time, the crossover arrived earlier, at 12%.
In UK practice the first sort is neither by rate nor by balance
Both strategies assume every debt is the same kind of promise. In the UK they aren't. Citizens Advice sorts debts into priority and non-priority, and priority debts are, in its words, "debts that can cause you particularly serious problems if you don't do anything about them" — rent, mortgage, council tax, energy, and some phone and internet bills. What follows non-payment is eviction, disconnection or a magistrates' court, not interest.
A council tax arrear charging no interest at all outranks a card at 21.49% on that logic, and no rate-based rule captures it.
Regulation adds a second sort. Since 2018 the FCA has defined persistent debt as a period of 18 months in which a customer pays more in interest, fees and charges than in principal. Firms have to prompt at 18 months, remind at 27 and intervene at 36. The 2016 market study behind those rules found 650,000 people had been in persistent debt for three years or more, with a further 750,000 making systematic minimum repayments over the same stretch. It also found 8.9% of cards active in January 2015 — 5.1 million accounts — were on course to take more than a decade to clear.
For that group the avalanche-versus-snowball question is the wrong one. The binding constraint is the size of the payment, not its destination.
What this arithmetic cannot tell you
Start with what doesn't exist. No randomised field trial has assigned real borrowers to avalanche or snowball and followed them to payoff. Everything above is a lab task, a correlational field sample, or arithmetic. Anyone who says the debate is settled is overreading all three.
Brown and Lahey's 13% came from typing cells in a room for under an hour, and their paper flags the figure as illustrative rather than transferable. Their choice study also found the subjects who gained most from ascending order scored highest on self-control — which, as they note, may mean the people most in need of a motivational crutch are the least likely to get one from it.
Gal and McShane's clients are American, enrolled before 2007, and in settlement rather than repayment.
The £687 is a model, not a measurement. It assumes the £300 never changes, no new spending lands on the card, no interest-free transfer is available, and the rates hold for four years. Change any of those and the number moves. Those rates are market averages published for June 2026; the rate that matters is the one on your own statement, which may sit well above or below them.
The comparison also assumes there's a surplus to direct. If £300 barely covers the minimums, both strategies collapse into the same thing, and the question stops being about order. The same tension runs through whether part of an emergency fund should be invested while expensive debt is outstanding.
Free debt advice is regulated, and 14% of the people who used it in 2024 paid for it
Debt advice in the UK is a regulated activity, and the free services aren't a thinner version of the paid ones.
GOV.UK directs people to MoneyHelper, run by the Money and Pensions Service, to find a free debt adviser. Citizens Advice publishes the priority-debt framework quoted above. StepChange describes its advice as "completely free and impartial". National Debtline, run by the Money Advice Trust, describes itself as "a registered charity providing free, impartial, expert debt advice to more than 100,000 people each year". England and Wales also have a statutory Breathing Space scheme, in force since 4 May 2021, giving up to 60 days of protection from interest, charges and enforcement, applied for through a debt adviser.
The FCA's Financial Lives survey found 1.7 million adults — 3.2% of the population — used a debt advice or debt management service in the 12 months to May 2024, rising to 22% among adults in financial difficulty. One in seven of those users, 14%, paid for it. In the same survey 33% of credit holders described debt advice as a last resort.
What would change the conclusion
A field experiment would end it. Randomise real borrowers with real multi-debt portfolios into the two orders, measure who finishes, and one paper settles what a decade of lab work hasn't. Until then the best evidence for the snowball comes from a typing task and the best evidence against it comes from a spreadsheet.
A wider rate spread changes the balance of it. The £687 came out of an 11.82-point gap between the dearest and cheapest debt in the profile. Widen that gap and the avalanche's edge grows faster than any plausible motivation effect. Narrow it and the edge disappears, as it did on two of the three profiles here.
An interest-free balance transfer changes the question entirely. It collapses the rate spread the avalanche exists to exploit, and replaces it with a different problem about the transfer fee and what happens when the promotional period ends.
The number worth watching isn't which debt you start with. It's the gap between your highest and lowest rate. A few points, and this argument costs more attention than it's worth. Above ten, the arithmetic starts to outweigh the psychology — and that's a fact about your statements rather than about your willpower.