Financial institutions have always known the data they possess could yield tremendous insights. They just haven’t figured out how to tap this goldmine. The vast amounts of customer data — device details, login information, transaction histories, payment behaviors, etc. — give banks and credit unions a truly unique opportunity to learn more about- and deepen relationships with consumers.
Fortunately, new tools and technologies have ushered in a Golden Age of data analytics, with machine learning and statistical analysis helping financial institutions gather actionable insights.
Traditionally, matching accountholders with the right products or services hinged on human intelligence, intuition and inference — often poor substitutes for scientific conclusions derived from a robust pool of data.
But if you subscribe to a video streaming service like Netflix, you will see machine learning in action every time a movie or TV show is recommended for you. Using “machine intelligence,” algorithms sift through the dizzying array of available choices to curate a tailored set of offerings just for you. The recommendation models that streaming services and shopping apps utilize has been curiously absent in financial services… until recently.
noun. a set of data that describes and gives information about other data.
The first step to generate actionable insights is to gather a large set of individual data points — as much information about customers, accounts, periods of time, and business processes as you can. This “raw material” can then be mapped as metadata to reveal underlying patterns at a very high level. This process is similar to how a football quarterback’s performance can be objectively calculated over a season with a “quarterback rating” — completion percentage, touchdown-to-turnover ratio, yards and touchdowns per game and more.
Abstractions like these allow for rapid comparisons between participants in a complex activity (professional football, in this case, or financial services) where digging into each individual detailed record in a database would be unwieldy or impossible. When looking at metadata in relation to accountholder behavior, a financial institution can identify a customer’s specific behavioral traits and identify correlations and trends at a much higher level than if it focused on any one individual piece. This metadata can be visualized in an easy-to-use platform with a highly intuitive user interface in such a way that just about anyone can investigate and learn about customer behavior without needing extensive math or IT skills.
The next step is translate these insights into action by crafting personalized messaging and marketing campaigns that match specific solutions with customers’ needs. Financial marketers can create campaigns that target people based on which products they adopt, their transaction frequencies and/or their transaction amounts — and do it all in a relatively simple, straightforward way.
Harnessing the Power of Data-Driven Decisions
The quantitative observations financial institutions can collect about their customers extend well beyond just a single service or product silo. Data analytics can be used to identify the deposit accountholders who are most likely to also acquire lending products like auto loans or home mortgages, and suggest offerings before that customer makes a big purchase decision.
Take auto loans for example. A bank or credit union can anticipate who would be a candidate for an auto loan program by examining their similarities to other customers that were already approved for auto loans. Examining patterns credit reports and how checking accounts are used, you can draw comparisons with other customers just before they applied for a car loan. It’s all about uncovering antecedent behaviors that work at a correlative level — leading indicators of interest in a product.
Many data analytics platforms use a class of algorithmic techniques called “collaborative filtering,” which analyzes data based on principles of “similarity,” “nearness” and “neighborhoods.” Two notions of similarity are utilized:
- User-based similarity, which defines nearness among users
- Item-based similarity, which defines nearness among products
Individuals are considered “similar” if they purchase many of the same products, while two products are considered to be correlative if consumers tend to purchase them together.
An analytics tool using these concepts can derive recommendations for varying segments based on the behavioral similarity of a user and the user’s nearest “neighbors” — those who may not be identical, but have enough similarities that the system can extrapolate predictions with a reasonable degree of reliability.
Back to the auto loan example. The data analytics platform might yield several observations about auto loan purchasers, such as:
- 4.9% of those at the financial institution have multiple auto loans
- 10.8% of those with mortgages have multiple auto loans
- Those with mortgages are 2x likely to purchase multiple auto loans
The relatively high correlation between mortgages and multiple auto loans indicates a similarity between the two products, and the financial institution can identify individual users who would statistically be significantly more receptive to targeted marketing messages promoting car loans.
This forms the baseline for recommendations that are more personalized and relevant, but these same behavioral comparisons can be expanded across multiple axes to refine, enhance and broaden auto loan recommendations. Let’s shift our focus over data we can analyze from debit- and credit card transactions:
- Who spends the most time driving (derived as a function of amount spent on fuel)?
- Who spends the most on automobile repairs?
- Who is currently financing an auto loan directly with the manufacturer?
- Who travels within a typical driving distance most frequently?
- Who has recently paid off an automobile loan?
Using this type of machine learning and statistical analysis will present additional targeting opportunities. For instance, a financial institution can drive interchange fee income by crafting communications toward accountholders who infrequently make purchases via a checking account to promote the advantages of using a debit card with a rewards or cash-back program… and look for similarities between people who will be the most receptive to such offers.
A financial institution can also target users based on that customer’s most- and least-utilized channel. If a customer rarely logs in via a mobile device, the financial institution can create a campaign that touts the use and benefits of its mobile app. Couple this approach with A/B testing, and you’ll not only be generating more conversions, you’ll be well on your way to creating a sophisticated and automated data-driven digital marketing strategy.
Rich Data Analytics = Better Products and Better Relationships
Leveraging machine learning and behavioral analysis reduces your financial institution’s reliance on generic marketing messages that get broadcast to large, unsegmented audiences. It empowers banks and credit unions to increase the number of relevant and effective interactions they have with consumers. The result is a set of deeper relationships that translates to improved cross-selling success, increased revenue and a much-improved customer experience that will only get better as analytics platforms learn more about the data financial institutions collect.
Adam Anderson is EVP/Chief Technology Officer at Q2 Holdings, where he is responsible for technology strategies that support teams responsible for innovation and customer satisfaction.