Saturday, November 18

Feature: machine learning unlocks big data in retail banking

Google+ Pinterest LinkedIn Tumblr +

A recent report about digital fails caught our attention lately. Written by strategy and marketing consultants Simon-Kucher & Partners, the Wharton School and University of Rochester, the Global Pricing & Sales Study 2017 delved into why only one in four companies achieve topline growth with digitalization.

Researchers found that the largest percentage of companies, at 48%, invested in increasing sales process efficiency but only 33% saw visible topline impact, described as “highest investment in low return initiative”.

Whereas investing in optimizing prices with big data and monetizing digital products only had some 11% and 8% of the companies’ investment dollars, but which resulted in 46% and 45% topline growth for those efforts, or, “low investment in high return initiatives”.

In other words, more companies are investing more money in digitalization strategies that aren’t working and avoiding investments in big data.

 

“That’s the disconnect for many people investing in this: it’s obvious to automate but you need to redo the content with it.”

 

Jan Haemer, one of the study’s authors, and a director at Simon-Kucher, said that there are a number of reasons why companies aren’t investing big data initiatives: they don’t know what data they have and how it can be used.

“People don’t look at all the options that they can get on it, and focus rather on those things that are evident,” Haemer said. “Everyone talks about sales force or another one of the new types of CRM (customer relationship management), as opposed to saying: what else can I do with digitalization?”

For financial services firms, the results could be sobering. Chatbots, for example, fit into the category of “highest investment in low return”.

The intelligence lies in how and what offers are being made to the customer, and how that’s
optimized, said Haemer. That’s not to say that chatbots aren’t a helpful idea, but automating an existing inefficient approach for a bad product won’t work.

“It might make it more efficient to serve the customers, but it is still not capturing the true benefit that you can get,” said Haemer. “That’s the disconnect for many people investing in this: it’s obvious to automate but you need to redo the content with it.”

 

“It’s a pure money play. This industry is so young, the question is who has the most money to throw money against the problem for the next five to ten years?”

 

Machine learning is the only way

The only way to really do useful things with big data is to apply machine learning to it, said Jens Rassloff, global head of alliances and Microsoft global alliance lead partner at KPMG.

Just recently, KPMG closed a deal with Microsoft and leading UK bank, and a deal with an Australian bank is in works. Both focus on a whole new way of thinking about analytics and machine learning on customer data.

In the case of the Australian bank, small and medium-sized enterprises are provided free tax preparation services, with the aim of enrolling hundreds of thousands of participants and working with the data to know everything about this group of customers. It was a two-year project in the making.

“I have seen many, many banks, many insurances, many industry enterprises, which were completely overwhelmed with the data they had in the past already,” said Rassloff. “Data is the new currency and we need to get our arms around the data of our company. I would say there is less than 30% of the clients today who really have a clue what to do with those data.”

The cloud market for storing all that data, he added, is dividing into horizontal and vertical strategies.

For example, IBM and Oracle want to apply machine learning to their existing products for services tightly aligned to business use cases, while Google, AWS and Microsoft want to artificial intelligence by making affordable, easily consumable and available cloud services.

“I think there is a need for both, and from our side, I would say that those who can continue to offer affordable, ready-made, secure cloud services over the long term will be the winners,” he said. “It’s a pure money play. This industry is so young, the question is who has the most money to throw money against the problem for the next five to ten years?”

 

“Their challenge is that it’s too many steps away from the data scientist group.”

 

Finance under pressure to adapt

Financial services are under immense pressure to adapt to a changing environment, whether that’s due to security and risk, as well as the slew of regulations coming in.

Sean Ma, product manager at data platform Trifacta, said that he’s seeing financial services having to adapt to a much wider range of requests and use cases than other sectors, with a noticeable trend towards making use of customer-generated data such as customer service chat logs or customer visits trapped in weblogs.

Trifacta recently collaborated with Google on Cloud Dataprep, an intelligent data preparation and cleansing service now in public beta. Trifacta has worked with banks such as Royal Bank of Scotland (RBS) while Cloud Dataprep worked with HSBC.

“Traditional sources of data, while very useful and form the bedrock of what they’re doing, today they need access to things like weblogs and customer service chats,” he explained.

The big challenge is how do you take this often unstructured or semi-structured data, and take advantage of the information so it can be fed into downstream machine learning algorithms for analysis?

HSBC, for example, is moving everything into the cloud and are one of Google’s biggest customers. In the case of RBS, the bank found that almost all of its data from customer service complaints was locked away in legacy systems, with only a handful of data scientists being able to access it.

 

“Overtime, the system gets more intelligent and derives better suggestions to new pieces of data for other users too.”

 

It’s a bit surprising, said Ma, how much data is actually locked away in large institutions, and that the majority of people who need it don’t get it.

“Their challenge is that it’s too many steps away from the data scientist group,” he noted. “People can’t access what they need, and that’s because they don’t have the tools, they don’t have the skillset, and they don’t have the rules in place to govern who has permissions to the data.”

“By actually analyzing those webchats and gaining access to 100% of that, this information is gleaned out and spread throughout the organization,” Ma said, adding that this results in better customer service and deeper customer insights.

And making information easier to work with should probably be more of an operational focus for large institutions with big data, so that it’s more accessible for people who aren’t data scientists. Ma pointed out that this is an area where machine learning is making a difference for Cloud Dataprep.

As users click, select, interact, drag and otherwise interact with their data, Trifacta’s background algorithms are at work learning from those choices in an anonymized secured data set.

“Overtime, the system gets more intelligent and derives better suggestions to new pieces of data for other users too,” he said.

 

Share.

About Author

Leave A Reply