How FinTech companies can use a modern data stack to gain a competitive advantage

Big data is a huge advantage for FinTech companies and financial institutions – even more so when non-technical business users can access data and use it in their day-to-day work. According to a report by IBM, 71% of banking and financial markets firms say that the use of information and analytics gives them a competitive advantage, an increase of 97% compared to 2010. And Morgan Stanley has named the 2020s as the decade of computing, where the development of computing will be driven by leveraging internal data.

The potential use cases of data for finance

Data is useful both for understanding “what happened” and “what is coming.” The potential use cases are many, but fall into three broad categories:

• Reporting: Reporting answers questions about what happened? This includes performance reports for external target groups or compliance requirements, and internal reporting on customer behaviour.

• Analytics: Analytics drives into higher order questions about how or why something happened. For example, suppose you release a new product or feature. Has it changed customer behavior?

• Machine learning: This recent case is predictive. Since it addresses unknown unknowns, machine learning is open-ended and requires significant research. For example: how can you leverage data to make smarter decisions or improve the customer experience?

Forward-thinking financial firms have made each of these use cases a reality. USAA partnered with Finicity to facilitate reporting in Intuit products without having to share logins.

And JPMorgan Chase uses machine learning models to deliver personalized research and recommend services to clients without human intervention.

The “holy grail” for FinTech and financial firms is to enable non-technical users to explore data without much training or onboarding. This multiplies the number of data inhabitants in your organization. It also frees data engineers to focus on developing machine learning use cases, which can help uncover trends or patterns in datasets too large to analyze using traditional systems.

Achieving this requires careful thought about how your data and analytics stack is organized. Let’s explore the roadblocks facing FinTech companies, then consider how to structure your data stack to empower business users to make real-time data-driven decisions.

The roadblocks to fully functioning computer programs

There are three common roadblocks for FinTechs and financial institutions:

The self-service roadblock: Every time a business user wants to dig into a specific question, they have to ask someone on the data team to build a model for that specific question.

The roadblock to data freshness: In traditional data systems, data is extracted and aggregated before the query. Without these precomputations, visualizations would be too slow to be useful, but these precomputations introduce latency. Data freshness suffers. Business users cannot access same-day data and use it to inform decisions. This can be a big problem in FinTech. Imagine trying to analyze fraud or anomalous behavior detection data—latency issues prevent you from monitoring or taking action on these issues in a timely manner.

The Consistency Roadblock: Computer programs have a variety of end users, including customers, investors, and marketing teams. It is important to report consistent numbers across press releases, investor reports and websites. But the combined power of The Self-Service Roadblock and The Data Freshness Roadblock means you’re working with slightly different data each time you pull a report, so the final numbers are inconsistent.

Let’s consider how we can meet these challenges using existing technologies.

The architecture of a modern data and analytics stack

To overcome these shortcomings, there are three main areas in a data and analytics stack to address:

The data storage layer: Raw financial data typically resides in internal databases or external sources such as third-party data providers. The first step is to feed this raw data into a data warehouse where it can be processed into something usable. This storage layer is typically stored in the cloud due to the efficiency, scalability, and pay-as-you-go characteristics of cloud data warehouses.

The modeling layer: Data that sits in a warehouse is complicated. A single table can have 40-50 calculations into it. Exposing this to the end user will require them to have an expert level understanding of how the data fits together. Instead, we store that knowledge about the table operations in an intermediate layer. Data engineers build models of the data, and these models are exposed to end users in the business language they already use.

Consumption layer: Users’ interaction with data comes through this consumption layer. This is where end users answer questions and build content to present their decisions.

There are many solutions for each team, but the approach that emerges is to choose the best tools for each team. Ultimately, this enables many more people to create and share their own dashboards, reports or visualizations.

An example from the real world: Trumid’s data and analysis stack

Like many FinTechs, Trumid is ambitious. It seeks to digitize the corporate bond market, where most trades still take place via text messages and phone calls. The company had a data stack in place, but was stifled by the three challenges mentioned above. To enable self-service analysis, while keeping data consistent and up-to-date, it selected best-in-class tools for each component:

The data storage layer: There are many cloud data warehouse options. Trumid chose Google BigQuery because of its infinite scalability, along with its ecosystems and managed integrations with Google Cloud Platform and GitHub.

The Modeling Layer: Trumid chose AtScale. We built AtScale to allow data engineers to expose data using the language of business, while hiding complex table operations from end users.

Consumption layer: Options here include Power BI, Excel and Looker. Trumid chose Looker because of its ease of use, its ability to create quick visualizations, and the way its permission model allows you to expose data while dynamically enforcing rules about who can see that data.

Trumid’s stack is simple yet elegant, and allows the company to use data as a competitive differentiator.

One last thought

The FinTech market is huge, with hundreds of billions of dollars at stake. Innovation is the key to differentiating yourself as a FinTech solution and improving the way you run your business.

With a modern data and analysis stack in place, everyone wins. Non-technical users can ask questions, explore data to find answers, and then create content to present those decisions—without the help of data analysts or engineers. And data scientists are freed from the day-to-day tasks of being on the lookout for ad hoc questions. Instead, they can focus their time and energy on innovation to drive the future of the business.

To read more about a real-world implementation of the modern data and analytics stack discussed in this article, check out our webinar, Power Better Decisions with a Modern Data & Analytics Stack.

About AtScale

AtScale is a data analytics and business intelligence platform built to accelerate the flow of data-driven insights. Headquartered in Boston, the company has secured a total of $120 million in funding from investors including Morgan Stanley, Wells Fargo, Storm Ventures and Atlantic Bridge.

The platform was developed by Dave Mariani, a data pioneer at Klout and Yahoo! Mariani saw how BI tools like Excel and Power BI are unable to connect directly to big data storage platforms like Hadoop – imagine trying to drink the ocean through a straw. AtScale addresses this with a middleware solution that exposes the full picture of data to end users using the language of business.