ING Bank is planning to launch an automated digital loan service for small and medium-sized businesses in Spain. In the coming weeks Banco Santander will do the same in the UK market. But instead of developing the service themselves, both are planning to piggyback on the online platform of Kabbage, a U.S.-based fintech start-up that lends money to businesses and consumers using a wide set of online data and algorithms to measure credit-worthiness.
The launch of these new services underscores not only the disruptive power of fintech start-ups but also how Big Data and machine learning are changing the financial services industry.
If pundits are right, banks will not only leverage Big Data and machine learning to make better and quicker risk assessment decisions, they will also add natural language interfaces and open their APIs to allow for the building of new kinds of financial services. For example, a consumer might use Facebook Messenger on her mobile phone to contact her bank and find out in real time whether it is possible to get a loan for the TV she is considering at the moment they are standing in the store, says Yann Ranchere, a partner at Anthemis Group, a digital financial services investment and advisory firm. Such services may be available as soon as next year, he says.
How the financial services industry will become increasingly data driven and what types of new services will be on offer in the future will be discussed during a November 5 Web Summit session, on the money stage, moderated by Informilo’s Jennifer L. Schenker. Panelists include Kabbage Co-founder and CEO Rob Frohwein, Jane Zavalishina, CEO of Yandex Data Factory and tech industry veteran JP Rangaswami, Deutsche Bank’s first Chief Data Officer.
Fintech start-ups like Kabbage, MODE, Kreditech and Cignifi have been gaining traction by using Big Data and machine learning to automate loan decisions, replacing outdated, inefficient manual processing. Since launching in 2008 Kabbage, which is based in Atlanta and operates globally, has issued over $1 billion in loans. It now generates $100 million a year in revenues.
The collection of data points — including connections into small businesses accounting packages and social media accounts — combined with machine learning doesn’t just help determine risk, “it also allows you to create a hyper-personalized experience,” says Frohwein. “We believe we can get to know customers way, way better than traditional banks.”
The new partnerships with ING and Santander (which also include investment by the traditional banks into Kabbage) are an acknowledgement that banks can no longer do everything in house. The services will not be white-labeled.
Big Banks and Big Data
Why aren’t banks developing these services themselves? Banks have been pouring millions of dollars into Big Data for years, says David O’Connell, a senior analyst at Boston-based Aite Group, a research and consulting firm specialized in financial services. But “finding the data, organizing it, making it consumable and then making Big Data deployment behave in such a way as to adapt as business requirements change and the data and the data sources change” is easier said than done.
A Q1 2014 global survey of 141 senior IT executives conducted by the Aite Group found that 76% of North American bankers are dissatisfied with Big Data analytics, compared with 47% of EMEA bankers and 32% in the Asia Pacific region.
Zavalishina, the CEO of Yandex Data Factory, a business unit of Russian search engine Yandex that specializes in machine learning, says she believes she knows why there is a difference in dissatisfaction between American and European banks. “The American buyers started early and invested more and that is why they are more unhappy,” she says. “During the first wave of Big Data hype they collected a huge amount of data. They have spent three years focusing on how to collect, store and clean data but this was all about cost, not value. Big Data is meaningless without machine learning. The real way to extract value is machine learning.”
Machine learning is based on algorithms that can learn from data without relying on rules-based programming. The tidal wave of Big Data has increased the potential of machine learning and the need for it.
Few European Banks Have Embraced Machine Learning
“When we explain to customers about the different ways they can improve their businesses with machine learning then they are very interested,” says Zavalishina. “But one of the first things they say is, ‘we have data spread around and the data is not clean, it may contain errors, it is not structured. Once we get the data organized then we can start doing this.’ This is the wrong way to think about it. First you have to think about what you want to do with the data and experiment with some data in real use cases. To make this work and build a predictive model you actually need the ability to experiment. Technology can do what people can’t: analyze, test, redefine and reapply thousands of hypotheses to determine the best next course of action. But using these technologies to their full potential requires change in existing business processes — the introduction of the “culture of experiment.”
Most banks have yet to embrace a culture of experiment but that said, some European banks have started using machine learning and are already seeing benefits. Yandex is working with Russia’s biggest bank, Sberbank, with 45 million customers. In Europe more than a dozen banks have replaced older statistical model approaches with machine-learning technologies and, in some cases have experienced 10% increases in sales of new products, 20% increases in cash collection and 20% declines in churn, according to a McKinsey briefing paper. The banks have achieved these gains by devising new recommendations engines for clients in retailing and in small and medium-sized companies. They have also used machine learning to create micro-target models that more accurately forecast who will cancel service or default on their loans and how best to intervene, says the briefing paper.
However, the best talent in machine learning is not in the banks, says Anthemis Group’s Ranchere. It’s in search engines as well as Silicon Valley giants like Amazon, Facebook and Apple. “The cutting-edge stuff sits with those players. [However] tools that used to be very complex are being commoditized at an exciting pace,” he says. He points out that Amazon is now offering machine learning as a service and IBM’s Watson is being used to manage risk and provide personalized guidance and investment options at financial institutions.
Yandex: Search Is All About Machine Learning
Yandex Data Factory says it is well-positioned to provide customized machine learning services to clients in a variety of sectors, including financial services, pharmaceuticals, gaming and transportation, because the company has 15 years’ experience as a search engine. “Search is all about machine learning – it requires handling huge Big Data sets and it is all about predictions,” says Zavalishina.
Not surprisingly U.S. search engine Google also appears poised to help financial services companies with machine learning offerings. An October 18th article in the Financial Times outlines how BlackRock, a multinational investment management corporation based in New York City, is in discussions with Google’s UK operations to create a joint venture that will explore how machine learning might be used to improve investment decisions.
Earlier this year BlackRock hired Bill MacCartney, a former senior research scientist at Google, to lead research on applying natural language processing and machine learning to quantitative investing.
Analyzing massive amounts of data additionally requires interfaces that process information more like a human than a computer — i.e., the understanding of natural languages and the generation of hypotheses. A U.S. start-up called Kensho, which has raised funding from Goldman Sachs and top-tier VCs, including Google Ventures, gives a glimpse of what is to come. Its market data analytics system claims to be able to find answers to more than 65 million question combinations in an instant by scanning more than 90,000 actions such as drug approvals, monetary policy changes, and political events and determine their impact on nearly every financial asset on the planet. Questions can be posed by typing them into a simple text box.
The Rise Of New Ecosystems
The ability to collect and parse Big Data is already possible. Indeed, banks have more data about consumers than ever before. So how come consumers are not yet able to access more personalized services from the big banks?
What is missing is ecosystems, says Ranchere. Banks need to become platforms and open their application programming interfaces — i.e., expose their proprietary software to outside developers so they can develop their own applications.
It is starting. Silicon Valley Bank is working to build out its APIs so that its high-tech customers can customize how they interact with the bank. That effort is being accelerated with the addition of the team of Standard Treasury, a fintech start-up that focuses on helping banks open their applications to others.
Meanwhile, France’s Crédit Agricole, Citigroup, Bank of America and Silicon Valley Bank have started sharing APIs with nimbler, unregulated tech start-ups so they can innovate much faster on retail banking services.
“There is a sense that the mood is changing,” says Ranchere. “Change is coming.”