In 2016, market research firm GfK’s¹ comprehensive research showed that customers expect their primary financial services providers to know about their lifestyles and financial goals. According to the research, customers are willing to share data with their providers to help them understand their preferences. At the same time, though a majority of companies in the Banking, Financial Services and Insurance (BFSI) industry want to offer personalized services to their customers, less than 20 percent consider themselves equipped to deliver on the personalization promise.
Personalization in banking comprises of a gamut of approaches that can significantly improve customer engagement, satisfaction and retention. From real-time alerts for one-time activities to analyzing spend patterns to offering financial guidance and providing timely reminders for late payments, several banks have taken steps in this direction. However, the one element that is bound to make the most impact on customer centricity is the Next Best Product (NBP) offering.
NBP enables banks to create more attractive product bundles for customers, provide enhanced levels of customer service and more effectively cross-sell and up-sell their products. By focusing on NBP, banks can develop best practices that can boost customer centricity. This can increase the customer satisfaction rate, and positively impact customer yield and product efficiency.
The NBP Offering
A perfect NBP offering incorporates response modeling, customer lifetime value and takes into account the minimum profitability per customer or customer segment. But the spindle on which such an offering operates constitutes the following three elements:
Single View of the Customer
The NBP Algorithm
Good quality data is the foundation of all successful marketing and outreach activities. This comprises of both first-party as well as third-party data. The spectrum of data² can include:
Current / previous products and services subscribed: For example, opening credit card accounts or making direct deposits
Balance and transactional history: Mortgage payments to another financial institution, recurring monthly payments
Demographic and self-reported data: Age, income, familial affiliations, social media presence to indicate behavioral and psychographic attributes
Channel preferences: Attributes such as preference for online banking, mobile apps and more favorable reaction to promotional content on a specific marketing channel
Clickstream data: Pages visited, length of stay on each page, products explored and depth of visit
From a simple rule-based system to a combination of predictive models, there are a range of analytical models for an NBP solution. Regression-based approaches, neural nets, discriminant analysis, decision trees and collaborative filtering are a few common examples. While most techniques work well, the two primary requirements in the BFSI industry are ease of use and high predictive accuracy.
There are two types of recommendation systems: memory-based and model-based. Memory-based techniques use the data about clicks / subscriptions / purchases to compute similarities between users or products. This data is then used to recommend a product to a user who has not been served with that content before.
Model-based techniques use several machine-learning algorithms to predict the probability of users subscribing or purchasing a product. Popular model-based techniques are Bayesian Networks, Singular Value Decomposition and Probabilistic Latent Semantic Analysis. However, the predictive power of these techniques is not as high as memory-based techniques.
Similarly, there are two types of collaborative filtering — User-based Collaborative Filtering (UBCF) and Product-based Collaborative Filtering (PBCF). UBCF assumes that people who agreed in the past, will agree again. Thus, to predict a customer’s reaction for a product, the opinions / actions of similar customers are considered. Though this is one of the most popular and easily implemented approaches, it has a major disadvantage. For new customers, sufficient information might not be available to decide who is similar or dissimilar to them.
PBCF operates on the concept that a customer is likely to have the same opinion for similar products. The similarity between products is decided by looking at whether other customers have purchased / subscribed / expressed interest in them. As the number of customers and products increase, the computation time of the algorithm doesn’t grow exponentially. Ultimately, it is capable of addressing the two most important requirements of the banking industry: quality of prediction and high performance.
It is critical to think through the operational, technical and organizational aspects of how to deliver targeted offers to each customer across several channels. Customers need a coherent experience across all channels.
The target groups for cross-sell and up-sell marketing campaigns are formed by combinations of product similarity scores and customer segments, and their current relationship with the bank in terms of the products they have purchased.
These groups can be strategically targeted with measurable personalized campaigns designed for cross-selling / up-selling of suggested products:
Strategic communication: Customers want to interact with banks in real time across communication channels of their choice. Hence, selecting the right communication channel for marketing is an integral part of any NBP solution. Understanding the customer journey and intervening at the right moment can add tremendous value for both customers and banks
Measure results: The response rates of personalized cross-channel marketing campaigns using NBP recommendation should be analyzed to explain the lift in response rate. Ultimately, the aim is to assess how the cross-sell / up-sell strategy impacts long-term customer loyalty. Since the machine-learning algorithm is used in the analytics engine, the process can self-learn and dynamically adjust to customers’ changing behavior over time. However, it is important for the cross-sell / up-sell strategy to evolve too. So, continuous measurement and refinement is critical.
Along with the above, hyper-personalization engines can help personalize outbound campaigns, essentially targeting a segment of one. Such engines can be implemented with great success in banking especially for High Networth Individuals (HNIs) and Key Opinion Leaders (KOLs).
Blueprint for Action
An effective NBP solution based on the tenets described above will enable banks to develop a deeper and more valuable relationships with their customers. Customers, meanwhile, can get the products and services they want.
To ensure the successful implementation and adoption of NBP and enable scaling up of the solution, banks can follow these key measures:
Identifying products, services and offers to be considered: Analyzing the ownership percentage at various levels of product hierarchy will help ascertain which product or service or offer should be considered in the NBP portfolio (such as credit card / savings account / platinum account / reward card)
Time period chosen for analysis: The time period chosen might be dictated by historical data availability. Also, it can be primarily based on the range of products for which the model is built
Model evaluation: Validation should be done on two data sets — one taken from modeling population as a holdout and the other taken from a time period outside of the modeling population time period
Piloting the models before implementation is crucial
Assessing impact of models: The effectiveness of NBP models should be tested via a control vs. test experiment
In conclusion, NBP holds the key to enhanced customer engagement and maximizing revenues for the banking industry in an intensely competitive and disruptive environment.
Join the conversation
25 January 2022
24 January 2022
19 January 2022