Making Financial Data Meaningful
An interview with David Firth by Sean Murphy
Around four in five of the calls landing in one high street bank's contact centre are not complaints, fraud or lost cards. They are people trying to make sense of a transaction they can’t recognise. The payment is almost always one of their own but a week later they have already forgotten and the transaction description is often a nonsensical collection of letters and numbers with no location and no memory to hang it on. Multiply that confusion across millions of customers and it becomes one of the more expensive customer service problems in retail banking, and it is the problem David Firth has spent a good chunk of his career dealing with.
Firth is Product Director at Moneyhub, the applied AI and data intelligence platform whose client list includes Nationwide, Lloyds, Standard Life and Scottish Widows. He joined in 2022 by way of Mastercard and TransUnion, having watched the company from a distance for years, struck by how a small business was scaling its technology across clients of that size. Banks, he points out, have sat on vast quantities of customer data for years and struggled to turn any of it into insights they can act on. Understanding this vast amount of data is not as easy as people think.
Moneyhub are experts in transaction categorisation and enrichment, and Firth is blunt about how hard it is to get right. Ten years ago, he notes, plenty of firms in open banking were happily claiming accuracy of ninety eight or ninety nine per cent, figures that fell apart the moment upon closer inspection. The trick was to park everything difficult in an uncategorised bucket and count what remained. Real financial lives are messier than the tidy models assume: many people have not just one current account but several, a Monzo balance here, a savings pot there, money shuffled between them in ways that defy easy reading.
What has changed is who measures the accuracy. Moneyhub ran a direct to consumer app for years, now operated by one of its clients, and used it as a live correction engine, customers tagging their own transactions and feeding more than a decade of judgements back into the model. That long ‘apprenticeship’ is now being plugged into core banking systems. The company categorises somewhere in the region of forty three million accounts for Lloyds and Nationwide, and when Firth says the accuracy is reaching ninety eight per cent he is quick to add that the number is not his claim. Banks run the tests and report the results back. In an industry used to vendors marking their own homework, having the customer do the marking instead is rare enough to be worth talking about.
Accuracy matters because more ambitious applications are starting to be built on top. Firth describes Paragon Bank's “Spring”, an instant access savings app that pairs account information with payment initiation. Link a current account and you see your savings in the context of what you actually have, and suddenly the friction of moving money your current account and savings collapses to a couple of taps. The behavioural effect is real: more frequent payments, more engagement, and pay by bank becoming the most used payment method on offer. The surprise, Firth admits, was the age profile. The heavy users are not the millennials everyone predicts but savers in their fifties, sixties and seventies.
From there, it is a short step to offering more anticipatory services to customers. If the intelligence layer can see a thousand pounds sitting idle and predict, from the rhythm of someone's payments, that it will still be there next week, why not offer to sweep it into a higher interest account and move it back before it is needed. This is where commercial variable recurring payments enter the story, the new framework that went live under the UK Payments Initiative and to which Moneyhub was among the first to commit. Firth is careful not to oversell the plumbing. It runs on Faster Payments, the same rails that have existed since 2008, so it is less a new road than permission to use the existing one for far more journeys. Commercial VRP opens up the possibilities to pay utility bills, fund pension contributions and investments.
For all the payments talk, it is pensions that seem to move him most, and specifically the Pension Dashboard, the long delayed government initiative that will let people see every pot they have paid into, private and state, gathered in one place for the first time. Britain is late to this. Several European countries have run dashboards for close to twenty years, and Firth points to Norway as the instructive case: a nation of barely six million where, he says, the service is used some thirty million times a year, with the overwhelming majority of that traffic flowing not through the state portal but through people's own pension providers, where they already feel some relationship. That distinction matters commercially.
Moneyhub has signed three sizeable pension clients to build exactly those private sector dashboards even though the public MoneyHelper service that anchors the system is not expected before the 2027/28 financial year, with the widely quoted autumn 2026 date marking the deadline for schemes to connect rather than the day anyone can log in.
When the dashboard launches, millions of people will see, for the first time, every pension they have ever paid into gathered on one screen, alongside a single figure the system calls an estimated retirement income: their state pension and every private pot, combined and translated into what it will actually pay them each year. Most have never seen that number and for many of them it will be a shock.
Which returns the story to where it started, with a person and a number they cannot at first make sense of. The whole point of enriching data was always to close that gap between what the figures say and what they mean. The aim is that people who can see their money clearly make better decisions with it, whether that is cancelling a forgotten subscription or paying more into a pension while there is still time for it to matter.
