Fractal Analytics’ Sankar Narayanan: ‘AI industry is a collective against inefficient decision making’

Banks have a lot on their plates right now. There are use cases for analytics and artificial intelligence today but they need to look externally for help. I'll admit that when I read that Apax was investing $200M in a company that delivers analytics and AI solutions to the financial industry, I had never heard of Fractal Analytics. My guest today, chief practice officer at Fractal Sankar Narayanan told me that a Forrester analyst called Fractal the best kept secret in customer analytics. Part hard technology and part professional services, Fractal works with many of the largest financial institutions around the world to improve their onboarding, marketing, and customer service, resulting in changes of hundreds of millions of dollars. Sankar and I discuss AI and analytics and why it's so hard for financial services firms to really embrace this set of technologies. He gives us concrete examples of the type of programs FIs can implement to improve company-wide decision making. Sankar also explains why AI isn't a catch-all for financial companies and how data engineering and behavioral science — together with AI — can scale decision making in financial services to the enterprise level. Subscribe: iTunes I SoundCloud I Spotify The following excerpts were edited for clarity. How Fractal is different than the competition I personally believe that the analytics and AI industry is a collective against inefficient decision making. We've had a lot of opportunity to learn from our peers and competitors. Every client we work with works with multiple suppliers, data providers, and tech firms. There's an opportunity to learn from each other. AI is gaining a lot of visibility across industry. I think AI is arguably insufficient if we're looking at it to solve all problems typical banks have. That's where our investment comes into play. There are two other core components that play together with AI for decision making to be very successfully executed at the enterprise level: data engineering and behavioral science. Behavioral science as holy grail How do banks understand and become relevant to every human they can service? Behavioral science is an emerging field we call 'design'. Design has two sub-components: a simplified user experience and building everything to understand and influence human behavior. These three components together will essentially help organizations scale decisions at an enterprise level. We've been in the analytics field since 2000. In 2009-2010, we decided that our business should have its strategy developed on the back of what our clients want us to do. So, we started measuring Net Promoter Scores to understand if a client likes us, will work with us again, and recommend us to his peers. Once we started measuring NPS, we focused on the right things to build. But it also helped us start measuring people's performance as a function of outcomes and not inputs. We stopped measuring things like Time Spent. It's an input. An outcome achieved would look at what type of output we delivered to our clients. In a sense, a behavioral science approach allowed us to see what were the big difficulties a client would face engaging with its clients and providers and how we could reduce that friction to zero. Tracking satisfaction along the whole customer journey Taking this outward, we started sharing this philosophy with our clients. Banks like to measure customer satisfaction at various touch points of customer interactions. You have call center customer satisfaction scores and branch customer satisfaction scores. I may have had a great experience talking with a branch executive but these scores don't tell anything about my journey. Where is that being measured? It isn't being measured systematically and consistently at most banks. We rarely see customer satisfaction tracked across the entire customer lifecycle. That's what we believe is the next wave of customer experience -- where banks become relevant not just at a customer's financial need, but at his or her life need. Bank impediments to analytics and AI A lot of times we see the learnings from analytics is relegated to the team doing the work. If those learnings are brought together across the bank, the speed to action is much higher. That's what we've observed. For example, risk and fraud are top of mind when it comes to algorithmic sophistication. But translating these learnings to say, marketing, isn't high because these two functions don't generally work well together. Marketing tends to be focused on trigger-actions as opposed to a fundamentally customer-first approach. Bringing this together at the organizational level is the biggest challenge. Working with banks of all sizes, we've observed their creative use of data is still quite low. We've had a number of conversations with banks that really showed a risk-averse approach to algorithmic data. But really, there's really little identifiable information that's needed for insights. Most AI initiatives are experimental and academic. There is a center of excellence team set up and it's funded from the top of the organization for this initiative. There is more benefit for AI when it's percolated in smaller experiments in different divisions in the bank. That way, the likelihood of it being implemented is a lot higher.