NOVEMBER 11, 2019
8 min read
Data Science has turned into a hype technology in the modern world and has created much buzz in all industries. It combines statistics, mathematics, data analysis, machine learning, and visualization to extract insights from all the big data that a company obtains. The results of the research are used for the product and process improvements of the business. Data Science becomes accessible for fintech products because digital services provide rich possibilities for data mining.
The history of Data Science officially started with John Tukey’s book “The Future of Data Analysis,” published in 1962, after scientists began to concentrate on Exploratory Data Analysis and knowledge discovery in the Database. As computer science and technologies were developing, and the possibilities of the data extracting were growing, the process of data mining became more actual. Over time Data Science started to bring new specialties to the market, such as Data Scientist, Data Engineer, Data Architect, Data Administrator, Data Analyst, Data Manager, and Business Intelligence Manager.
Dashdevs helps fintech clients to develop digital ecosystems, web and mobile applications, and improve business processes. In addition to the basic development services, we also provide Data Science consulting help. As we often get questions from our clients and partners regarding the use of Data Science for business, we’ve decided to write an introductory article to describe the approach, share our best practices and tools, and give common data science use cases for the fintechs.
Data Science project architecture
The process of Data Science integration with fintech products starts with the analysis of goals and data sources. The former must describe the measurable targets that the business wants to achieve. For a typical fintech product, there is a number of data sources:
- User verification services. A fintech account can be created only for a verified user who has passed personal information to the KYC, FinCrime, AT, and AML services. These providers require a photo of the real documents, proof of address, and sometimes video or selfie.
- Card management services. A card can be issued, activated, blocked, closed, or re-issued so that we can get information about all these statuses. Card processing providers can expose information about the places where the card was used.
- Payment services. All fintechs products are about the movement of money and balances, so these service providers give data scientists rich information about financial behavior, which becomes the core of the modeling procedure.
- Mobile and web app analytics. The application analytics tools can give information about the usage of in-app features. Mobile and web applications have integrated analytics SDKs (software development kits) that are sending structured metadata per the triggers.
- Customer support tools. All modern customer support tools provide a wide variety of data concerning user requests, timelines, and resolution of customer problems.
- Open data sources. Sometimes we need additional information from government or official statistics that allows us to correlate our results with social and political changes.
All the data is gathered into the data warehouses. We want to put additional attention to three different terms - data lake, data warehouse, and data swamp. The data lakes have unstructured raw data from different data sources. Data warehouses have processed data that is structured and ready for the Business Intelligence (BI) processes. However, if data lake is overloaded with a massive number of unsorted data, it might become unusable. Such a messy data lake is called a data swamp. That’s why the process of data governance is one of the most critical parts of the Data Science process.
After the data is structured, it is ready for the next processes, such as business intelligence, machine learning (ML) data processing, and modeling processes. Business intelligence helps to get insights from the data.
Modern data warehouse
The process and the tool that is used for data structuring is the most crucial decision that is made by Solution Architect and Data Architects. Dashdevs outsourcing company commonly uses Snowflake as a data warehouse for fintech digital ecosystems. Hence we can give you our requirements for the warehouse. These criteria can help you to make the correct decision if you are choosing among several tools.
- Cloud solution. We try not to use on-premise solutions due to scalability and maintenance issues. Snowflake solution uses Amazon Web Services (AWS) platform for data storage that is called Amazon Simple Storage Service (Amazon S3). This solution is totally secure and easily scalable. On the other side, Cloud Computing possibilities are much more affordable than hardware. Fortunately, today fintech products are usually cloud-based and this fact simplifies cloud integration.
- Extract data from different resources. Various tools and service providers send the data in diverse formats such as XML and JSON. The data warehouse can receive structured or semi-structured data from different providers and transform it into a usable state.
- Ingestion services. Data providers work in different procedures and schedules, so we need to have a tool that gives the ability to load data continuously. The Snowflake has a serverless computational model Snowpipe that serves these procedures.
- Scalability. Data processing is not linear, so we need a scalable solution that supports multi-clustering. Data Engineering process can have activity and downtime periods.
- Integration with Data Science tools. We work with different data science instruments such as Spark, Python, R, Anaconda for data analysis, and modeling.
- Easy DB management. The process of database governance can be complicated. Sometimes we need to clone data or restore it after false actions. All of these actions can be done by Snowflake smoothly.
- Manageable sharing. Different teams can do different processing operations with different levels of access to information. For example, the compliance officer needs to have full access to the payment information, including payee and payment details. However, the marketer needs to know only the time and the address of the point of sale.
- Cost control. We don’t want to pay for the service if we don’t use it.
The process of data storing and data governance is one of the most crucial tasks for the Data Science, so it must be ordered appropriately.
Business Intelligence tools
The process of finding insights can be done by advanced analytics specialists such as BA or data analysts. Fintech companies are providing similar services to the customers, so they need to find their unique positioning on the market. The data from the warehouse or the data lake can contain essential insights. Consequently, marketers, product owners, and project managers are usual users of BI tools too. When we select the BI tool for the team, we usually pay attention to the following criteria:
- Integration with a data warehouse that helps to retrieve data seamlessly.
- Clear user-friendly interface is required for non-tech users with an uncomplicated model and chart creation process.
- Easy data management needs access to select, filter, and sorting options, because Data scientists process the data from different sources and for different timeframes.
- Secure access to the tool and manageable access control are required.
- The high-performance speed of data processing is a must for any BI tool.
The most popular business Intelligence tools are:
- Microsoft Power BI is a powerful tool that allows creating data-driven culture for businesses. Like all Microsoft products, Power BI gives a strong possibility for visualization tightly integrated with Microsoft Dynamics products. It has intuitive user experience and extensive possibilities for reports creation.
- Looker helps companies drive better outcomes through smarter data-driven experiences. This solution can be easily integrated with different sources of information. It has high performance and an exciting solution for generating insights. It supports multi-cloud hosting and hybrid environments.
- **Tableau **is designed for corporate and personal usage. It has interactive tools for visual analysis that is powered by the patented VizQL technology. Tableau helps to perform content discovery for the data from different sources. The solution can be deployed on-premise, in the cloud, and hosted.
Data Science use cases in fintech
Data Science becomes a trend for different fintechs because it can help them solve various business problems in a fast way. Here are the most frequent use cases in the fintech industry:
- Fraud detection is the most crucial problem for any financial institution, so they’re constantly looking for anti-fraud tools and different ways of automation in risk management. Different types of frauds try to impersonate, steal, or perform money laundering schemas. Efficient anti-fraud tools must have prevention, protection, and notification systems. The data warehouse receives data on the fly from payment processing systems, passes it through the models, and generates real-time results. Also, Data Science can help to define patterns of the fraud collaboration, build schemas and interaction diagrams.
- Deep learning of the customer’s performance allows conducting user segmentation, customer behavior modeling, as well as real-time and predictive analytics. BI tools allow visualizing the financial activity of the user in the digital bank ecosystem. The user’s financial behavior insights assist with building product strategy for fintech organizations. An additional parameter that can be provided to fintechs from Data Scientists is a customer lifetime value (CLV), which is a prediction of all the benefits that a business can get from the relationship with a customer.
- The risk modeling system helps to define if the user is reliable and can be granted with access to the additional services, higher money credits, and lower rates. Data Scientists can build models based on the product usage and open-source information from different sources.
- Product improvement strategy can be based on product usage analysis and market information. Data Scientists can build models and predictions of the feature changes in customer behavior and possible reaction to the fintech product changes.
- Process improvement can be based on the usage of the Digital twins’ approach, which is a trend of product development for the last few years. The financial organization or digital bank can track offline operation and customer support processes metrics, analyze them, and simulate the changes to evaluate future effects.
- Personalized marketing is one of the most powerful tools for fintech products promotion. Data Science gives a possibility to analyze the behavioral patterns of the user and suggests them the relevant financial products and services.
Fintech, as a young and fast-developing industry, is absorbing all knowledge and approaches that give an additional boost to their products and digital ecosystems. Unlike high street banks, the architecture of digital banks is more flexible and allows them to integrate with modern services and apply the latest data-mining techniques. Startups and mature businesses require Data Science consulting services that can empower them to organize processes and improve the products, so don’t hesitate to jump into the Data Science stream now. Feel free to contact us if you have any questions about data science in fintech.