Leading players in the banking industry such as JPMorgan Chase and Bank of America are already making good use of big data. Frankly, by this point, if your business is failing to exploit the opportunities presented by big data then you are considerably behind the pack. We suggest looking at four practical applications with due attention to big data sources, specific algorithms, and models used to get valuable insights, as well as business outcomes.
1. Personalized customer experience
Big data technologies make it possible to analyze the data generated each time a customer gets in touch with a bank in order to identify behavior patterns and deliver personalized customer experience. The task is not an easy one, taking into account multiple available channels. A bank’s analytical system should ingest the data created when a customer visits a local branch, surfs the website, uses mobile banking, withdraws cash at ATM, contacts a call center or a chatbot, etc. Though this plethora of data sources is already impressive, banks seek for partnerships with telecoms and retailers to get even more information, such as spending habits and geolocation data.
Collecting data is only the tip of an iceberg – the most difficult part is to get valuable insights. To make this possible, banks use many advanced analysis techniques, such as multivariate models, cluster analysis, and decision trees. For example, the analytical system may find out how customer profile data correlate with the deposit interest rate and period.
Such precious insights enable banks to design effective and targeted promotions, recommend relevant products, charge individual fees, offer personalized credit limits and interest rates. For example, having analyzed customer’s location data, purchase history and response to previous promotions, a bank can incentivize the customer to purchase at retailer X and get 15% cash back. If 20% of customers frequently shop at retailer X, a bank may consider issuing a co-branding card with this retailer. This is what Citigroup and Costco did and their card produces great results: 2.4 million new cards issued and increased cardholders’ spend in 2017.
Capital One shows another example of personalization: they get valuable insights into every customer’s spending history and share them with the latter. The bank uses push notifications to inform a customer if they were charged twice for the same product or service or if a recurring charge increased.
2. Credit scoring
The idea behind credit scoring is to create a predictive model designed to assess the chances of a customer’s default on payment based on rating an applicant’s characteristics and their attributes. Undoubtedly, the more characteristics and attributes, the more informed a bank’s decision will be. That is why banks rejoiced with the advent of big data. Traditional scoring models mainly rely on socio-demographic, employment and education data, as well as on a potential borrower’s credit history (what credit types they have already applied for, whether they paid on time, etc.).
In contrast to traditional ones, big data-driven models can deal with many more data sources without compromising on the analytical system’s speed. For example, banks are able to analyze their applicants’ behavioral data, such as activities on social media and web search history, as well as external data, such as records from government databases, statistics bureaus, etc. to get a more comprehensive picture of an applicant as a consumer of banking services. While this quantity of data sources is already impressive, banks can also analyze an applicant’s behavior outside the banking sphere, for instance, by looking at their e-commerce and microgeographical data to understand the applicant’s spending patterns, favorite products to buy, and places to go out. To run this complex analysis, banks can make use of machine learning techniques.
Thanks to big data analytics, banks are able to take into account many more aspects to make well-grounded decisions much faster. For example, Bank of America was able to reduce its loan default calculation time for a mortgage book of more than 10 million loans from 96 hours to just four. And the time required to score a portfolio of 400,000 loans with multiple scenarios applied shortened from 3 hours to 10 minutes.
3. Risk management
It is extremely important for banks to do a great risk management job. Numerous risk groups such as market, investment, and liquidity ones, require the bank’s ability to crunch big data and build predictive models tuned to deliver accurate forecasts. For this, banks define thousands of risk factors and apply trend analysis, scenario planning, and assumption testing to identify possible outcomes.
For example, banks can use historical simulation to manage investment risks. It allows banks to trace back the influence of key factors such as interest and exchange rates changes. The model requires a bank to examine each day of the past year (or even several years) and analyze what movements happened compared to the previous day and what were the outcomes. If a bank chooses a 1-year period for analysis, this results in running 250 scenarios.
JPMorgan Chase, for example, analyzes data about thousands of companies, millions of customers in more than 60 countries in order to spot trends and dependencies. They apply the combination of big data and artificial intelligence to see, for instance, where rising prices are affecting consumers the most.
4. Fraud protection
Banks use big data analytics to identify suspicious activities and prevent fraud. An efficient fraud-fighting strategy includes a combination of measures. One of them is to recognize unusual activities in a customer’s behavior pattern, for example transferring almost negligible amounts. If this activity is fraudulent, the amount may be too small for an account owner to notice. Still, if the fraudulent scheme is successful with millions of accounts, the cumulative loss is painful.
To double-check suspicious behavior, a bank can analyze the content a customer posts on social media. For example, one tweet that announces an upcoming vacation can explain a sudden change in the customer’s spending pattern; or geolocation data with international calls service turned on may show that the customer is indeed in another country.
Big data can be helpful at the stage when a bank has already identified a fraudster. In this case, the bank can resort to connectivity analysis to find out other players who are likely to be involved in a criminal scheme.
The scoring model is also applicable for fraud detection. In this case, it is crucial to identify hundreds of attributes that are a subject of scoring. These attributes may include the period that a customer is with the bank, the value of their account, the number of banking products used, etc. The system analyzes every attribute in real time and returns some score. The key ingredient is to have this score recalculated with every new activity of the customer. Combining the scoring model with machine-learning based factor combinations helps to come up with an even more effective approach.
In a nutshell
The four examples described by no means make the list of use cases exhaustive – big data consulting practice knows much more. However, we’ve picked personalized customer experience, credit scoring, risk management and fraud protection as these issues are of high priority for banks and big data opens new opportunities for enhancement in those areas.