Finally Moving Out Of Silos
Prahalad Thota, Head of Enterprise Analytics & Data Science, Wells Fargo, talks about how the Banking, financial services and insurance (BFSI) industry is adjusting to the latest changes in the data and cloud world
Financial services dipping a toe in the data lake
In the financial services— data, analytics and insights are very critical to succeed. If you’re a broad based financial services company, meaning you have all of the different financial products and deep relationships with your customers, you have a lot of data. Globally a lot of this used to be siloed. Every single product group or department used to collect their own data, store it in their own environments. They all made these decisions locally.
Over the last three to five years people have realised what a tremendous asset data is to be able to deliver products and services in a customised way, to manage risk. Financial services have come to realise that it’s really something that can differentiate how you operate. They are trying to organise data in a very systematic manner, understand what data is, define it and also have established governance on it. So not only Wells Fargo, but all financial services companies have a higher bar on how you manage data.
Secondly there has been somewhat of a limited move to the cloud. Other industries have gone to the cloud much more rapidly than I would say financial services. From a data security data encryption perspective, we want to make sure that our data is protected wherever it is. Financial services companies are starting to toe- dip into the cloud space. The cost of data storage is coming down with like some of the big data platforms like Hadoop. We can use some of this data and build very advanced sophisticated models and we can deploy them real time to make decisions that are good for the customer and good for the bank. We’re building the right models. We understand exactly how these models work. We can’t have black box models. We can’t say, we’ll put a model and it said to do XYZ with the customer. We have to see exactly what input are we using and have to be able to explain that.
Thirdly, more and more advanced analytical techniques are being used. Buzzwords Artificial Intelligence and Machine Learning are starting to become a reality. Companies are investing and trying to figure out how they can leverage ML algorithms to look at problems that in the past were solved with traditionally with things like statistical methods. I think in Silicon Valley alone there are thousands of startup companies, which are taking a small part of the problem in the entire data journey. It could be something around how do you combine data from different sources to all the way to applying ML algorithms, making it really automated and more efficient. We are trying to figure out what’s the right set of tools and technologies that we need to have to be able to do our work. We want to make sure that whenever we use data and create any analytics
We’re building models to predict the probability that a customer has a need for a particular product, feature or conversation. There are some simple algorithms, but in some cases we have to build very sophisticated models
or insights we have the right oversight from legal compliance and risk.
Key ingredients of the technology stack
When you look at any technology, there is the internally developed stuff, some things which are very specific to us. Then we buy enterprise solutions. Finally there’s open source. We integrate all of that into our technology stack that all our data scientists have access to.
I think there is actually a really strong set of talent in India when you look at data science. Apart from lateral hiring, we’re also targeting universities to bring in talent that has a quant background. A lot of programmes now offer data science almost like a minor in the engineering curriculum. Sometimes they miss some of the trends in the financial services context, which is okay as long as the talent has the quantitative abilities.The business context can be taught. In the US we look at data science, computer science, math, economics, even psychology. We hire the best talent and there’s a lot of training. The space is evolving so fast and so rapidly, that we find that we have to constantly look at what’s happening and bring additional training.
How do you organise your data? What is the strategy for customer and for businesses? You have to manage the data in the environments and making sure that you have all the processes. Analytics and insights looks at everything from advanced models to business intelligence. We’ve decided to expand our data science teams significantly and a big part of that will happen out of India.
Making customer experience better
In the bank right now we are aggressively pushing what we call the personalization initiative. We want to personalise our conversations, products and offers to the customers based on who they are, what are their true needs and how they like to interact with the bank. We are gathering a lot of the data about the customer—who the customer is, how long they have been at the bank, general attributes and their interactions over multiple platforms. We already had this information in different places, but we’re bringing all of this data together.
We are building models to predict the probability that a customer has a need for a particular product, feature or conversation. There are some simple algorithms to do that, but in some cases we have to build very sophisticated models. There are all these probabilities and preferences that we model and then feed into a decision engine which every day spits out what is the right conversation to have with a customer. Finally we have to actually check whether the customers are happier, more engaged and satisfied? Are we meeting their needs more effectively than in the past? We will roll out version one in June 2020 and that will keep evolving. We are applying
AI to look into things like the voice of customer, the way they talk to us in multiple ways, emails, chat messages etc. Our objective is also to build models that will predict fraud as it is happening real time.
We take permission from customers on where we can use what data. They can always opt out.
We’ve organised our data into 19 domains. For each of these, we have a business data owner and data management lead. Anytime anybody needs to use this data, they have to go to a fairly rigorous approval process. We also have our legal risk and compliance teams asking: Are you following all the laws of the land? Any analyst in the first week of joining the bank gets so much training on it. On top of all of this, there’s GDPR (General Data Protection Regulation).