Exploring Ally Financial’s Innovative Data Strategy

What is your day-to-day like and is there anything exciting that you’re working on? Could you give us some background on what you do?

Before we get into the day-to-day, I’d like to give some background about myself and the organization that I work for. Ally Financial is the leading digital financial services company and doing it right for our customers is the primary focus for Ally. We offer a lot of exciting products to our customers especially as we are in the Covid pandemic.

We wanted to improve the financial well-being of our customers, so we’re offering a lot of exciting products that could provide short-term relief to our customers. That’s one of the primary things that the organization is looking at, for example, if we have a customer who had a good relationship with us, we might offer a loan with different options or different payment options and we could also proactively work with customers and come up with a payment plan if there is any delinquency risk.

Those are some of the work-related things that we are currently undertaking. Specifically, my role at Ally, I’m the head of data architecture and my role involves working with our business partners, understanding their requirements, and coming up with architecture recommendations roadmaps, and priorities. I have a team of architects that work with me and currently, we are also working on a major cloud transformational initiative primarily focused on customers.

We are always looking at the customer experience when they access Ally services, how is the customer experience, what are some of the things we could do to improve our customers' experience, how can we improve customer loyalty. Those are some of our business focus, and as part of that, there’s a data strategy that we rolled out and we are looking at executing that data strategy. A major component of it is cloud transformation, moving all to the cloud and we are looking at making sure that as part of the data organization, we are looking at centralizing data and democratizing analytics.

Why don’t you tell me a little bit more about how you got interested in data and the tech industry in general?

All throughout my career, I have been focused on data and I started as a report developer and then I got into management consulting. I did management consulting for five to six years in the early stages of my career. Everyone was trying to solve problems and without data, I don’t think anyone could solve the problems as data is at the center of everything.

What are some of the innovations and what are some of the trends that you’re currently noticing in your industry?

Predominantly many organizations are looking at cloud transformations. We could see that as one of the big things happening in the data space and apart from that organization, looking at how do they make use of the data and how can they monetize the data?

In most cases, we all collect data, but that data is sitting somewhere else and many organizations recently have taken a different look, saying, “Hey, I have this data. how could I make use of this data and how can I improve my business or how can I increase sales or increase my revenues?”

So that’s one of the lenses people are taking a look at and also we would say there’s a lot of innovations in the space, especially when it comes to AI and machine learning and also with cloud computing. AI and machine learning have become very easy and flexible.

Specifically, in the financial services industry, we see that data is being used to do a lot of analysis to help understand the risk for any business. Like for example, we get the customer data, we talk to the customer over the phone, we get the interactions about the customers, but then based on their emotions we would be able to come up with the tailored or personalized product for the customer.

So that’s all happening in real-time within a fraction of a second or a millisecond. The technological innovation is especially tremendous in the space of AI and machine learning where the customer voice could be recognized and you could meet and greet the customer depending on the voice and also you could take the data about the customer and do segment analysis and be able to offer personalized products or recommendations for the customer.

So, this is all possible with AI and machine learning and we think many financial services companies wanted the customers to do self-service by themselves and we think that is feasible only through AI and machine learning.

IoT is another area where manufacturing organizations or maybe in any organization, IoT is being used tremendously, for example in the insurance space. There are various devices that transmit messages constantly and that IoT data is being collected to understand risk. An insurance company might have an underwriting policy where they could possibly come up with better premium rates.

What have been some of the biggest challenges around successfully executing data transformation initiatives, how did you go about solving them and where is the room to improve?

When we talk about data transformation initiatives, what are the business objectives, that is what I think is a key driving success factor for any of the programs. First, having a clear vision of what you want to accomplish, what is your timeline, is this a reasonable time to accomplish everything, and do you have enough resources available? Do you have the technical resources or the business resources who could provide support so that you could accomplish your goals. Those are the things to be looked at but beyond that, anyone that wants to improve their sales or support their customers to the best possible understanding, what the customer needs and when they need it is very critical for any business to be successful and providing personalized recommendations.

In the financial services industry, many are looking for personalized products and short-term relief and if we offer a product that might not suit their needs at this time then it’s possible that we might not be providing the right offers to the customer and that’s where an organization needs to understand what is the need and how can we satisfy our customer and that should be incorporated as part of the business strategy.

Once you have the business strategy then you think about how to make use of data to derive insights, what data do we have, what data do we need to collect, what are the KPIs or the metrics that we need, and how do we make use of data to do that. What sort of technology is going to help me achieve that? It’s like a triangle, you have data on one edge, and you have a customer on the other edge and then technology on the other edge.

What would you say takes up the most time when you’re working with data? Is it data quality or data preparation? What’s your biggest headache when starting a new project or doing data science?

Having a clear objective and vision is very important to understand. What are the business priorities? And then you understand what data you have available and what you need to collect and then also understand how you’re going to make use of technology to accomplish your goals.

Then execution or the implementation comes into the picture. So, you put everything clearly on paper. When you have a design established, and then you think about quality governance, metadata management lineage, etc. This is a common foundational thing in any data initiative where you would have to have good quality data.

If the quality of the data is bad then you’re not going to get the results no matter what sort of algorithms you run on the data, it’s going to be misleading and also you spend a lot of time in cleansing, making sure the quality is good. To me data quality should be good at the system of origin.

Before the data is transmitted from one system to another or one platform to another you put quality rules in place, so the quality issues do not propagate to multiple systems so you can address where it started and it doesn’t propagate so that would be my approach rather than taking a bad quality data and cleansing it in multiple places. For example, you might be gathering customer data or you might be applying for a particular product and there are some issues where quality is poorer. So the system that collects that data, that’s where the quality issues should be addressed rather than taking this data into a platform.

For class running models, there’s kind of an iterative process, you are looking for some level of accuracy. You wouldn’t get that accuracy immediately when you start testing your model so you have to have a mechanism to handle those in a way that you have the time available because most of the time there is a very short time frame to deliver and we’re all looking at agile type methodology when it comes to delivery.

When you’re dealing with machine learning-type initiatives, and if you’re still working in an agile fashion, make sure that you have enough sprints incorporated in the plan, so the testing could be done in multiple iterations and then you have the time accounted as part of the schedule. We also get a lot of unstructured data and semic structured data, with many systems that will spit out these sort of semi-structured data that needs to be passed, so by having to these data sets and having a way to incorporate auditing and data validation and reconciliation type rules so that you can ensure what you receive from your social system, you’re able to validate and match the data set those are some of the basic things.

Identity resolution when it comes to customer data, there could be different systems that might be collecting customer data, but identity resolution specifically is a foundational capability of customer 360. It’s not necessarily that identity resolution is something that everyone does on all sorts of data, it’s specifically for customers and as part of a foundational capability. Any organization would invest in having a customer 360 if they have disparate data sets about their customers, so as part of that a subset of the Customer 360 is where I did retail resolution.

What are your views on the current enterprise data ecosystem, do you think there’s room for improvement?

I think we have great technology choices available today and with cloud computing, everything is quite easy. Let’s say if I have to do a machine learning algorithm or apply a machine learning algorithm on a data set and make this happen constantly, it’s going to take intense compute capabilities, and also it’s going to spit out a huge output in terabytes.

We can do that very easily with the cloud capabilities available. We don’t need to wait for the infrastructure to be available. We can spin up even a GPU type instance very easily within seconds in the cloud and I think that’s one of the benefits of cloud computing, no infrastructure needed upfront and you can pay as you use, that’s another flexibility that we have got specifically as we measure in cloud computing.

There is a variety of services available to achieve what you want. Whatever tasks that you were doing previously that would take months and months are happening in seconds these days. So we think we are on the right path with cloud computing and the benefits that one can reap

especially in the data space with cloud computing is tremendous and this a lot of choices available readily available to use but there’s also a notion about protecting data and data security.

That’s if you have the right controls in place to address security then cloud computing has given us everything that we need to operate in an efficient way, but there’s always a notion of will my data be secured in the cloud. That’s the common thing that anybody asks when they haven’t taken a cloud transformation journey, it’s new for them, they’re thinking if this is going to be safe or secure, but I think there are techniques to make sure that you can keep your data safe and secure but data ecosystems have got a lot of great capabilities.

Today, there’s a lot of great AI-based analytics technologies available that could help you search and get the data that you need and you don’t need your business user to be working with IT constantly. Instead, they could go and sell the service themselves. That’s another area, the BI and Analytics, and the data visualization spaces, are constantly maturing and there’s a lot of innovations happening there as well.