2016 03 14 news predicting crisis 883x432 1

Predicting crisis

14 April 2016

The article at a glance

‘Incident Dashboard’ helps Barclays predict whether incidents may snowball into big issues, as discussed at Cambridge Trust in Banking conference. Just about …

‘Incident Dashboard’ helps Barclays predict whether incidents may snowball into big issues, as discussed at Cambridge Trust in Banking conference.

2016-03-14_news_predicting crisis-883x432

Just about every adult, at some point, encounters a bank’s automatic teller machine (ATM) that doesn’t work. Some people might tweet about the inconvenience on their Twitter account, while others simply move along silently to another ATM that functions.

Yet it’s the potential for seemingly small incidents like an ATM not functioning to snowball into big crises that can harm a bank’s reputation and trust among customers. So Barclays, the transatlantic bank headquartered in London, has been implementing an “Incident Dashboard” that uses big data analytics and artificial intelligence to help predict through social media mentions when a problem may become a major issue.

“We always make a determination on the technical severity of an incident and its potential impact on the overall system, but there are also issues that while technically negligible can be a big issue for customers – so that’s what this incident system is designed to predict,” says Elisabetta Osta, Managing Director, Information, Insights and Innovation, at Barclays.

Mrs Osta discussed the Incident Dashboard system at the Cambridge Trust in Banking conference, held last autumn at the London Stock Exchange and organised by the Executive Education division of Cambridge Judge. She was recently named a Visiting Fellow in Marketing at the School.

The Incident Dashboard for Barclays was developed by predictive analytics company Market IQ, which participated in the bank’s Accelerator programme in London in 2014. The programme is aimed at supporting breakthrough financial technology (fintech) innovations.

The Incident Dashboard ingests data from the social universe, monitoring millions of data points in real-time data from sources like Twitter, Facebook and blogs, as these sources represent the leading edge of sentiment.

Using Market IQ’s Natural Language Processing algorithms and Artificial Intelligence framework, the Incident Dashboard categorises social datasets to focus only on relevant information: this weeds out very general tweets such as “I love my bank” or “I hate my bank” in order to focus most closely on specific topics that may develop into issues. The system also figures out what people are really saying on social media that is relevant to the bank and filters through the noise to help Barclays efficiently react to relevant and high-priority issues.

“People tend to talk in a very natural manner when speaking in the social media universe,” says Fahad Kamr, CEO of Market IQ. “If an ATM is down, they may not say that but may instead say: ‘Standing at branch, can’t pay my babysitter’ or ‘Can’t get my money out’ – so the system figures out what they’re really saying.”

The Incident Dashboard assigns a probability to the likelihood an incident will develop into something serious, and this figure is seen by key decision makers at Barclays in order to prioritise incidents and organise responses.

“This is another way for us to establish the severity of an incident,” says Elisabetta Osta. “We are looking as much at the customer impact as the technical impact. It’s a way for us to listen to customers in a very real-time way.”

On the issue of trust in banking, the topic of last autumn’s conference, she said there are “different levels” of customer trust that are vitally important for banks to maintain.

“Customers want to know that their money is safe with us, that they can access their money without a glitch and pay without a glitch, and this is the fundamental level of trust. Then secondly there’s value for money, and then reputational trust as reflected in the media is the third level.

“You need to do well in all three – but the Incident Dashboard and our work with Market IQ is particularly designed to help identify problems and respond more quickly in order to assure higher levels of customer service and trust.”