Home > What’s UP > blog > How to unlock the power of ML in digital banking sales


08. 02. 2018

How to unlock the power of ML in digital banking sales

Banks need to make better use of machine learning in sales to be ready for the fierce competition from digital disruptors. We’ve put together a few pointers on how to get it right.  

Financial institutions are sitting on top of a huge mound of customer data such as info on demographic and social background, purchase history (including products purchased and amounts paid) as well as household income. At last, this data can be turned into revenue, thanks to the paradigm shift in sales from being reactive to proactive, and from instinct-driven to insight- and data-driven, SAP explained.

ML, a subset of artificial intelligence, helps analyse all available data and improves sales effectiveness and efficiency. Algorithms detect patterns and learn how to make predictions and recommendations by processing data and experiences. And as McKinsey pointed out, these algorithms can evolve in response to new data and experiences, and become better over time.

How algorithms do the trick

Generally speaking, an algorithm uses training data and feedback from humans to learn the relationship of given inputs to a given output (for example, how the inputs ‘time of year’ and ‘interest rates’ predict housing prices). There are different algorithms for different business use cases. Algorithms are used in various types of analytics that can be grouped based on their complexity or areas of application.

The potential payoffs are plenty. Banks can better predict possible customer churn and improve the effectiveness of cross-selling activities, Crowe said recently. Yet, the sales industry has only just begun playing with predictive analytics, and prescriptive analytics is hardly ever seen in the banking sector.

Use cases to boost revenue

Let’s start with an example. A bank has commissioned an analytics solutions company to end the migration of high-value mortgage customers to rivals. The company compared the attributes of loyal customers with those that had churned, and identified over 100 factors related to customer, product and transactional data.

The company worked with statistical techniques to hone in on these key factors and discovered, for example, that city-based middle-aged professionals with gold credit cards tended to churn more than others. The data was fed into a predictive modelling tool that uses neural networking techniques to predict churn behaviour.  

Then the number of factors was reduced to around ten and the model was applied to all mortgage customers, ranking them in order of their likelihood to leave. Based on this ranking, the bank launched a targeted marketing campaign, which cut the churn percentage by nearly half. For an outlay in the low six figures, the bank retained “millions of dollars of mortgage business”.

In another example, a US bank used ML to study the discounts its private bankers were offering to customers. Bankers claimed that they only offered discounts to top clients, and more than made up for them with other, high-margin deals. But the analytics showed something different: patterns of unnecessary discounts that could easily be corrected. After the unit adopted the changes, revenues rose by 8% within a few months.

Besides these use cases, ML tools can also forecast the close dates of sales deals based on previous wins and performance patterns. Using this and other key insights, they can calculate the probability of winning deals to predict expected revenue and create more accurate sales forecasts.

Another application of ML-enhanced analytics is lead scoring. By isolating the key qualities of the best prospects and customers, as well as identifying common traits shared by high-value businesses, this solution assigns a numerical value to each of these signals – industry, title, number of employees etc. The higher the score, the more likely the lead is to convert, according to Base.

Get the right data for ML

This sounds all very well but how to integrate ML into your sales strategy? First of all, collecting better data for ML analysis is key. Simply using existing data from a customer relationship management (CRM) system does not always bring the desired results as this data might be incomplete or flawed. Getting internal data cleansed before moving into AI is simply a must. But there is help. More and more vendors take public sources of data, organise it into data lakes and prepare it for AI to use, which might be yet another reason to opt for third-party providers.

ML tools work as correlation engines and can pinpoint which data combinations are likely to bring the desirable outcome. They can suggest the right channel, frequency and time to contact the customer as well as what tone should be used during the interaction. All these help banks create tailored offers and a cushy digital customer experience. ML can also evaluate the performance of marketing interactions, measure effectiveness and analyse customer reaction to marketing messages, as well as prioritise them based on performance.

ML helps sales teams too

Analytics can be used not only for data-driven and automated sales, but also to guide sales personnel by providing them with ML analysis. A chemicals firm, for instance, used analytical tools to give the field sales force insight into the overall business, which then enabled them to create their own strategies, implementation plans and own projects on the platform. These, of course, were monitored and tracked by managers. Within just a few weeks of implementation, churn was down and pricing was up, and within a year, it brought in an additional $50 million in EBITDA, McKinsey revealed.

Digital analytics can dramatically improve the chances of revamping sales, but hasty investments can be counterproductive and expensive. Not to mention that sometimes these investments do not seem to pay off at all. What is missing? Potential culprits include the front line who does not trust the data, overly complex insights or sales teams who simply feel that their own experience and expertise are being ignored.

For more on how banks can benefit from ML tools and what they can learn from other industries and competitors about implementing AI-driven technologies, download our white paper on AI in banking sales.

AI and ML in digital banking sales