Lloyds Banking Group’s core business dates back to the 1700s. The group – which comprises 16 brands that collectively serve over 28 million customers – doesn’t have any legacy technology quite that old, but does have its share of dated IT systems and data processes, including some paper-based ones. 

LBG, however, is now working on a major transformation with the aim of becoming the UK’s biggest fintech, and has invested £4 billion in its ongoing pivot to digital banking. It has already transitioned 46% of its applications onto the cloud and decommissioned over 500 legacy applications. 

Enterprise data underpins every one of these programmes and decisions. 

The Stack sat down with LBG’s Jo Riseborough, group data management and culture director, to talk about how she sets the data management and culture strategy for 65,000 employees to create foundations for “well-understood and well-controlled” data as a consistent base layer of digital transformation.

No “garbage in”, please!

Riseborough joined Lloyds from insurer Aviva in 2022.  

She emphasised that well-managed data is central to Lloyds’s digital transformation project as a key unlock for data-intensive workloads like training machine learning models and building GenAI products. 

“Across the last three years, we've come such a long way in terms of understanding our data, having the lineage mapped, knowing that our key data is of the right quality,” she told The Stack, fresh off taking home the “Large company of the year” prize at the British Data Awards. 

She’s not resting on her laurels though. 

The bank’s data management transformation “won't finish at the end of this year or at the end of next year – that journey will continue,” she said. 

Better controls for better outcomes 

Riseborough sits in the Chief Data and AI Office, but her team operates a hub and spoke model; she runs a centre of excellence in the “hub” that is also responsible for writing frameworks and standards on data management. 

Each of LBG’s business areas, between 10 and 12 in total, have their own data teams which execute these standards according to the type of business unit and data they process. 

“We help with oversight and track the business areas progress,” Riseborough said. “When I joined, Lloyds was still very much establishing its foundations in terms of how it wanted to manage its data,” she added. But the bank is now on track to have data management controls over 90% of its most critical datasets by 2026. 

As one of the biggest banks in the UK, LBG has millions of data items. 

Creating a baseline of well-controlled, trusted data to work from for critical business use cases was a key step in Lloyd’s data journey. 

Things like ‘know your customer’ or financial crime processes, are “the data we can't live without”, she said. “We went through a process of identifying the most critical data to the bank, which is around about 1,000 data items. 

“And then we deployed controls over those.” 

This framework improves the control environment other teams can work from, “but ultimately, what that results in is a real kind of build in trust around our most important data,” she told The Stack. 

Having trusted and reusable data products sitting in the cloud means colleagues can apply to their own business use cases and the business can more easily identify use cases where AI can “shift the dial in terms of things like delivering operational benefits and delivering great outcomes.”

The data products themselves follow a three-stage build: a first stage that copies the source data, a second that cleanses it for consumption, and a third "consumption data product" tailored to a specific use case, such as customer service, or a financial or regulatory process. Access at that third stage is governed and updated regularly. 

Riseborough described consumption data products as "a really big unlock for the group" because they spare analytical and engineering teams from recreating data for every new purpose, streamlining the path from identifying a use case to having usable data.

Producer and consumer model 

On how the group applies consistent data quality controls at pace, Riseborough explained the team has established a “producer-consumer” model approach to distribute accountability for data management. 

Senior leaders are responsible for ensuring the data their teams produce is well-governed and fit for purpose. Consumers of that data, typically back office functions like finance, are responsible for maintaining that high data quality and flagging issues. 

This dual accountability means that the business can make data-driven decisions or deploy customer-facing products with confidence. 

To track the success of this model, her team monitors adoption of accountabilities across business units as well as controls over critical data and its quality. (She highlighted each business unit is responsible for flagging what data is considered “critical” to their respective teams.) 

“The most powerful thing that we have been able to leverage, particularly on the quality of data over the last few years, is having a visible metric at our GEC level or our board level that kind of tracks how we're measuring data quality and how we're remediating data quality issues when we find them. 

“We've set a fairly low appetite, I guess, in terms of once we find something, we want to make sure that we remediate that as quickly as possible, which sometimes involves a number of parts of the business collaborating together to make that happen.”

How to monitor data quality at scale 

Riseborough said working with vendors who are on the cutting edge when it comes to applying automation and AI to data management has helped the bank adopt newer technologies. 

She said the bank self-builds some tools but uses Collibra for data lineage and Ataccama for data quality monitoring. 

“Often the vendors move at a really fast pace and they're already thinking one step ahead about how we automate more, how we use AI and things like AI governance. Sometimes they are a couple of steps ahead of us.”

Ataccama is the group tooling capability for tracking data quality across business units.

“In terms of automation, we're not quite there yet in terms of automating the remediation of data quality issues, but the control is automated. The Atacama tooling is automated. So it will just run automatically across data sets all of the time.” 

She said this is essential for the bank to be able to monitor data quality in real-time and remediate issues as soon as possible “so we don't find out a number of months later.”

Over the past three years, the group’s use of the tooling has massively increased, with business units being set their own rules and tailor the tool to their unique data control needs – supporting the data accountability goals Riseborough’s team is working toward. 

What's next

The forward-looking work clusters around automation and AI. Riseborough said the team is exploring the use of agents to support the build of data products as a future state, expecting it to enable faster builds than is possible today, while still working through questions of efficiency, data quality, and the controls such builds require.

A specific proof of concept is running out of the bank's technology centre in India, where colleagues are looking at how to resolve data quality issues "without the need for delivery and engineering teams", the capability Riseborough flagged as not yet in place. She said she was looking forward to seeing how it develops before the end of the year. 

Asked how she keeps pace, Riseborough was disarming about the limits of any one role. "My job is only so big. I can't be an expert in everything, but I'm really fortunate that I work with a set of experts," she said, pointing to peers specializing in platforms, AI, automation, and tooling, alongside industry conferences, peer conversations, and roundtables. 

Her day-to-day challenge, she said, is keeping pace with everything Lloyds needs from data management without "forgetting about the fundamental kind of building blocks."

Delivered in partnership with Ataccama.