The Bank of England's July Money and Credit release landed this week, and the mortgage trade press has read it the way it reads most months: approvals down, affordability squeezed, buyers hesitant, somebody calling for stimulus. Mortgage Soup's write-up gathered the usual voices. OSB Group noted that house purchase lending was broadly flat on the year, Phoebus Software said demand "remains sensitive to affordability and the direction of mortgage rates", and MT Finance said the market "urgently needs" stimulus.
I read the same release and came away more optimistic than any of them. Not because the numbers are good, but because of what they say the problem actually is.
My read on the July numbers
Start with what fell. Approvals for house purchase came in at 56,100 in July against a six-month average of 60,800, roughly 8% below trend. Net mortgage borrowing dropped to £4.3bn from £7.7bn in June. The effective rate on new mortgages ticked up to 4.45%.
The article's one consolation was that remortgaging is "resilient" because approvals edged up from 34,100 to 34,500. That's a month-on-month comparison. Against the six-month average of 41,400, remortgage approvals are 17% below trend, a bigger shortfall than house purchase. Remortgaging isn't holding up. It's the weakest part of the market.
Now look at what didn't fall. Total lending to individuals is still growing at 4.3% a year, and has sat between 4.1% and 4.3% every month since April. Consumer credit is growing at 9.2% and accelerating, with credit card balances up 12.5% on the year. Households put another £3.8bn into their bank accounts in July, £2.2bn of it into ISAs. Mortgage repayments, at £21.3bn, are running above their six-month average while gross lending, at £25.9bn, is running below it.
| July 2026, seasonally adjusted | July | June | Previous 6-month average |
|---|---|---|---|
| House purchase approvals | 56,100 | 58,200 | 60,800 |
| Remortgage approvals (to a different lender) | 34,500 | 34,100 | 41,400 |
| Net mortgage borrowing | £4.3bn | £7.7bn | £5.3bn |
| Gross mortgage lending | £25.9bn | £26.9bn | £26.4bn |
| Mortgage repayments | £21.3bn | £21.2bn | £20.8bn |
| Net consumer credit borrowing | £2.0bn | £1.9bn | £1.9bn |
| Household deposits, net flow | £3.8bn | £6.2bn | — |
| Effective rate on new mortgages | 4.45% | 4.35% | — |
Source: Bank of England, Money and Credit, July 2026.
That is not the balance sheet of a population that can't afford to borrow. It's the balance sheet of a population that is borrowing unsecured, saving, and paying down mortgages while it waits.
It isn't a shortage of homes either. Rightmove's August index put the number of homes for sale at a twelve-year high, with London stock at its highest in sixteen years, while buyer demand sat around 10% below last year. Zoopla's August index had 5% more homes on the market than a year ago, buyer searches up 7% on the year, and sales agreed 6% down. More stock, more people looking, fewer deals getting done. Zoopla's own estimate is that the rise in mortgage rates since January has taken about 9% off a typical buyer's purchasing power. The demand hasn't gone anywhere. It's parked at the point where a lender has to say yes.
Where the remortgagers went
The remortgage shortfall deserves its own explanation, because the obvious ones don't hold. There's no evidence in any of this data of homeowners selling up to rent; the stock numbers above are about sellers who are still trying to sell.
What the data does show is that the Bank's remortgage figure only counts borrowers who move to a different lender. Product transfers, where a borrower rolls onto a new deal with their existing lender, aren't in it. UK Finance's first-quarter figures had 84% of all refinancing done by product transfer. A like-for-like product transfer with no additional borrowing doesn't require a fresh affordability assessment; a remortgage to a new lender does, at today's stress rate, against a proxy expenditure figure the new lender has never tested on this borrower.
So the 6,900 remortgages a month that have gone missing against trend are, on the most plausible reading, borrowers staying put because moving means being reassessed. That is the market telling lenders, in the plainest possible terms, that the cost of uncertainty is now higher than the cost of a worse rate. It's also every lender's lost opportunity to win another lender's customer.
"Affordability" is the wrong word for this
Read the lender commentary again with that in mind. Every one of them is describing uncertainty, about rates, about funding costs, about what a borrower can really carry, and calling it affordability.
The distinction matters because affordability is something that happens to a lender and uncertainty is something a lender creates. The standard mortgage affordability assessment takes an applicant's declared income, subtracts an expenditure figure derived from CATO and ONS population data or a bureau product such as TransUnion's affordability tools, and stress-tests the surplus at a rate above the pay rate. The expenditure figure is a proxy. It describes what a household of that shape, in that postcode, with that many dependants, tends to spend. It does not describe what this applicant spends. In our analysis, those proxies are frequently not representative of what the applicant's own bank data shows.
A proxy with no error bar leaves the underwriter with one rational move at a 4.45% stress rate: add buffer, on everyone. The buffer is not there because the applicant is unaffordable. It's there because nobody knows how wrong the proxy is for this particular person. That's what the missing 4,800 house-purchase approvals a month are: not borrowers who failed, but borrowers who couldn't be trusted enough to pass.
The market is rationing by uncertainty, not by affordability. And unlike rates and stimulus, uncertainty is something you can shrink now.
Why open banking hasn't already fixed this
The obvious answer is that open banking gives you the applicant's actual spending, so use that instead of the proxy. Most mortgage lenders have tried it and most don't trust it, and in our experience they're right not to trust the raw feed. There are three reasons.
First, transfers. A borrower moving £600 to a savings account and £400 to a partner's joint account looks, in a raw transaction feed, like £1,000 of outgoings. A borrower receiving a transfer back from that savings account looks like £600 of income. Read naively, transfers inflate or deflate the surplus depending on the direction of travel that month, and an underwriter who has been burned by that once will discount the whole data source.
Second, the single-account assumption. Open banking affordability tools are usually built as if the applicant runs one current account. Most people run several, and the one they connect is not always the one the money lives in. A surplus calculated on one account out of three is not a surplus, it's a guess with a decimal point. It's the same flaw that makes standalone open banking credit scores so unreliable: one account, read in isolation, treated as the whole picture.
Third, isolation. Open banking data read on its own is a digital bank statement, and a bank statement tells you about cash flow, not about commitments. It's most powerful when it's reconciled against credit file data: the loan that appears on the bureau but whose repayment doesn't appear in the transactions, the BNPL exposure that isn't on the credit file, the returned direct debit that a bureau search won't show you.
None of these are reasons to reject open banking. They're reasons to reject an uninformed implementation of it.
Measuring trust instead of guessing at it
This is the thinking behind our triple lock affordability model, and the point of it is often misunderstood. It is not a better single number. It is three independent views of the same applicant, built specifically so that the variance between them tells you how far to trust any of them.
The first lock is end-of-day balances: what is actually in the applicant's accounts on the day before payday, across multiple pay cycles, and how much of the overdraft they're living in. Transfers between accounts cancel out at that point. It doesn't matter how money moved during the month; what matters is what's left when the month is done.
The second lock is the aggregate surplus: full income against full expenditure, reconciled across every account the applicant holds, with duplicate transactions removed, over three, six and twelve month windows.
The third lock is population benchmarking against ONS estimates by postcode, marital status and dependants. Yes, the same ONS data that lenders currently rely on. The difference is what it's for. ONS is a fine third opinion and a dangerous only opinion. Used as a sanity check on a real number it tells you when an applicant's spending is out of line with their peers and worth a second look. Used as a substitute for a real number it tells you nothing about this applicant at all.
When all three agree, the buffer can shrink, because the uncertainty that justified it has gone. When they diverge, the divergence itself is the finding: an underwriter knows exactly where to look rather than declining on instinct. That is how you widen the pool of lending without widening the pool of risk. You're not lowering the bar; you're measuring where the bar actually is for each applicant.
The journey problem
There's a second reason open banking hasn't delivered for mortgage lenders, and it has nothing to do with data quality. It's sequencing.
In most mortgage journeys we see, open banking is a second line of defence. The applicant goes through a bureau-based affordability check first, an underwriter reviews the case, and only then, often days later and through a separate link, is the applicant asked to connect their bank. By that point they've had an initial decision, they've had a delay, and, if the case looked marginal, they've already started talking to another lender or broker.
The conversion hit isn't a risk-appetite problem. It's a plumbing problem. The most useful data in the whole application is being requested at the point in the journey where the customer is least likely to give it. Put it at the front, make it the first affordability view rather than the last, and the underwriter review becomes a confirmation rather than a second interrogation.
The next twelve months, for the lender who moves first
Here is why I think mortgage lenders should feel buoyant about the year ahead, and it's not a rate call.
The borrowers exist. Household lending is growing at a steady 4.3%. Households are comfortable taking on unsecured credit at 9.2% growth. They're saving. There are more homes on the market than at any point in twelve years and more people searching for them than a year ago. And a chunk of those would-be borrowers, roughly 4,800 house purchases and 6,900 remortgages a month against trend, aren't being approved for the one product where the assessment relies most heavily on a proxy nobody can check.
Every lender in the market is waiting for the same two things, a lower rate and a stimulus package, and when either arrives it will arrive for everyone at once. Uncertainty is different. It's shrinkable today, per applicant, by the lender who chooses to measure it rather than buffer it. That lender approves the marginal cases its competitors are declining, declines the ones its competitors are approving on a flattering proxy, wins the remortgagers who are currently too nervous to move, and does all of it before the applicant has had a reason to shop around.
The question for a credit committee this autumn isn't "can we afford to lend more at 4.45%?". It's "how much of our decline pile is uncertainty we could have resolved?". July's data suggests the answer is: more than you think.
Sources: Bank of England, Money and Credit, July 2026; Rightmove House Price Index, August 2026; Zoopla House Price Index, August 2026; UK Finance Household Finance Review, Q1 2026 (as reported by Mortgage Solutions). For how the triple lock model works in practice, see our affordability analysis use case.
← Back to News & Opinion