The pandemic made it clear that traditional banking is a thing of the past. Online banking had already been on the rise, but a 200% jump in new mobile banking registrations in April 2020 established that customers are able and willing to change. As more Americans bank virtually, banks are fighting to meet customer demand?
And beyond the challenges set forth by the pandemic, “digital natives” like Rocket Mortgage, Venmo, Stripe, and Robinhood are all vying for business. These technology-forward organizations position themselves differently from traditional financial institutions and are attracting a younger user base for their services.
But a traditional bank has advantages over these challengers:
- Familiarity and history: Your personal relationships and history with customers mean that your bank is often more aware of their history. And customer questions can be answered in person instead of being routed through a call center.
- Deep and rich data: Historical data can prove invaluable for ML efforts. Customer deposit amounts, payments, and balance information can be used to predict future behavior.
- Preference for personal banking: Customers, especially those with a high net-worth, may have discomfort with digital channel dependency for wealth management. A new brand might be a risk, and they could feel uncomfortable not having a specific person to call if something goes awry.
As we get back to our “new normal,” traditional banks can use the rich data and relationships they have with customers to their advantage. Forward-thinking leaders are reimagining what it looks like to do business, and they’re using machine learning to elevate the customer experience.
Discover how you can use machine learning to create engaging and profitable relationships with your customers.
Every bank can find value from machine learning
Machine learning might sound like a type of data analysis useful to only the largest of organizations, but its concepts can scale to meet the needs of small and mid-sized banks too. When we use the word machine learning (ML), we are referring to machines and systems that can learn from “experience” supplied by data and algorithms. In banking specifically, ML algorithms can be used to identify patterns in data beyond what humans are capable of observing, and these learnings can be applied to new data sets.
It is now possible to improve the customer experience using ML. By parsing customer transaction data, ML can identify clues and patterns ahead of time, even before the customer considers taking action.
For example, the process of buying a home and obtaining a mortgage might begin with small savings accumulation or an increase in deposit amounts from wages. ML models can assess banking-specific data like credit patterns, risk tolerance, and price sensitivity, and can be coupled with demographic data like age, median income, and distance to branch.
The goal of using ML data in this use case is to target prospective customers with offers most relevant to their situation and stay ahead of customer demands.
Knowing where to begin — and where to focus efforts
Machine learning has such a wide variety of applications, it can be difficult to know where to start. Identifying a use case for customer-focused ML expenditures is a good first step. In general, we have found that you can benefit from starting with a use case with low or medium relative complexity. Examples focused on improving the customer experience include:
- Predicting service line interest (HELOC, mortgage refinance, etc.)
- Streamlining loan approval processes
- Increasing lines of credit
- Improving fraud alert notifications
With so many use cases to choose from, it can be easy to get lost in the planning for each example. Instead, try focusing on one area at a time.
Using your strengths, combined with ML concepts, you can deliver an optimal customer experience that digital challengers just can’t match.
Cross-selling across the relationship with machine learning
You can leverage machine learning to determine not only which customers would be a good fit for a mortgage loan, but also the other products that customers might need. Thinking about a mortgage loan, a Home Equity Line of Credit (HELOC) might be a good match for a new homeowner. In any case, the message and product can be tailored to meet the customer’s specific needs.
Another part of cross-selling is to personalize the offer based on the customer’s history and propensity to buy. Perhaps an interest rate would be meaningful for one type of customer, while a waived application processing fee would entice another. For individuals who identified as being interested in high revenue products, the marketing effort can be even more personalized, like a phone call or an in-person event invitation.
Applying machine learning in real life
The following illustration is an example of how an internal dashboard might appear to a banker or service representative. For any specific product, each person has a percentage likelihood that they will take action.
Individual model scores are shown, along with next steps, such as outreach about an investment account, or mailing a promotion about mortgage rate refinancing.
In this example, marketing inputs, like website data, are combined with transaction and deposit information. When a banker or service representative encounters a customer, either in person in on the phone, they can suggest specific next steps, or ask if the customer has questions. Having a dashboard with this information enables banking employees to be empowered to guide the conversation with data in real time.
Related Case Study: Machine learning predicts outcomes at Primary Financial
How does a financial services firm improve sales targeting to predict its clients’ desires to invest? Machine learning was the answer for PFC. Learn more.
FAQs about machine learning and banking
Does the machine learning process work fast enough to enable real-time benefits?
For all but the most complex scenarios, yes! Normally, ML is fast enough to be integrated into real-time transactions.
Does machine learning get in the way of compliance requirements?
In general, no. By using existing data that you obtained or using your data in coordination with third-party data, you are not running amiss of privacy and compliance concerns.
How do we ensure the use case we pick machine learning is right given that there are so many to choose from?
We recommend focusing initially on those low-cost, high-ROI use cases with a low-medium relative complexity. Given additional experience, context, pipelines, and an understanding of how advanced analytics programs operate, then more complex initiatives can be undertaken.
Data reliability can be a concern. Using low-quality data is not advised, but it is possible to start projects with small data sets. Engaging with a third party to evaluate your situation is advised in situations like this.
Reimagining customer insights & relationships
Banks that employ machine learning will have a portfolio of more customers than ever who are positioned for a variety of banking products delivered in a digital, personalized, and meaningful way. Now is the time to act and implement machine learning to meet customers where they are, using the contact methods that they desire and delivering the products and services to best meet their needs.
Need help building your use case and plan? Access our Machine Learning Use Case Guide for Banks.
Want to dig deeper? Check out our webinar on this topic.