Precision-targeted marketing through big data enabled machine learning
In today’s era of information-based competition, banking executives are increasingly looking towards big data to maintain their competitive edge.
In the retail banking area, there has been weak recovery following the 2008 financial crisis. General market growth has been modest, while the low interest-rate environment and new online market entrants have further compressed margins. It was always important to retain and expand existing customer relationships. But now it is critical.
Just when getting to know your customers better has become an imperative, face-to-face customer interaction via traditional branch networks is going down, as customers migrate to online banking.
Hyper Personalised Experience
Big Data analytics can help by enabling “Hyper Personalised Experience” at an unprecedented scale. Building on the investment in CRM over the last decade, newer technologies are available today to capture, analyse, and scale customer information, and predict cross-selling opportunities for banking products.
For this, the customer information footprint has to expanded from across internal transaction systems (i.e. “360 degree” view) to include data from the web, mobile devices, chat and messaging services, and social media to provide a “720 degree” view of the customer. Though the full “720 degree” view may still be a while away, incremental benefits can be obtained from increasing the depth of information available about the customer.
The goal is to determine the best possible product/pricing combinations to offer to customers, through the right channels, at the right time, so as to optimise both the value provided to the customer and the value of the relationship to the bank.
Analysing ever increasing data volumes
New York headquartered Novantas is a leading provider of analytic advisory services and technology solutions for financial institutions, including banks, brokerages and Fintech companies. It provides information, analyses, and automated solutions that improve revenue generation — across pricing, product development, treasury and risk management, distribution, marketing, and sales management.
In the deposits space alone, it helps optimise returns for over US$3.2 trillion of deposits on its client’s balance sheets. Novantas focuses on “customer science” and helps clients develop a deeper understanding of their customer needs and behaviours. It helps translate those insights into rapid and sustainable improvement in revenue, growth, credit quality, and profitability with proprietary tools, analytics, technology solutions, and scoring models.
Hank Israel, director of Marketing, Propositions, and Products at Novantas, said, “We help our clients solve practical, pragmatic business problems, such as identifying pricing and prospecting opportunities that can improve customer acquisition. As the market evolved, we had to expand the customer attributes we use and leverage AI [artificial intelligence] to find ways to bring raw data together and derive new insights.”
However, to take its analytics to the next level, Novantas had to modernise its data platform. Traditional systems could no longer be used with the increased volumes of data coming in. They also needed to process and analyse a greater variety of data, such as audio from call center recordings and unstructured text in payments transactions data. For example, by using natural language processing (NLP) to analyse call center recordings, Novantas could gain insights into customer sentiment on products and promotions.
There were requirements for real-time or near real-time analyses and also, simultaneous evaluation of many periods.
Building a self-service customer journey analytics solution
To address these issues, Novantas built a self-service customer journey analytics solution, called MetricScape, on a Cloudera-based Hadoop and Spark cluster.
The platform integrates customer accounts and transactions data from more than 30 institutions with third party data, and applies machine learning models to operationalise customer scores, such as customer potential value (CPV), for a variety of use cases, including offer optimisation, customer retention targeting, and cross-sell and upsell activities.
More than 1,000 business metrics per customer can be analysed with sub-second response time. The system can look at five years of data for six million customers and obtain insights within minutes.
Novantas deployed Cloudera on both Amazon Web Services and on-premise, to cost-effectively manage variable workloads as required. The organisation was able to monetise its new solution in only six months. This was 18 months earlier than originally projected.
MetricScape acts as a librarian for customer and financial data metrics, providing a business metadata governance layer that tracks things such as data lineage, definitions and dependencies. Its analytical workflow helped a team of data scientists pull together relevant data sets for analysis. It allows banks, in general, to look at customer account histories, the results of previous marketing campaigns and other data to segment millions of bank customers based on their likely responsiveness to planned promotional offers.
While all machine learning models and metrics are ultimately stored in Novantas’ MetricScape solution, IT staff wanted to give their data scientists the flexibility to choose the tools they used to create the models.
So data scientists at Novantas use Cloudera Data Science Workbench as their core development environment. This enables them to use whatever language they want to develop metrics and models and plug in their own libraries into this secure environment. Once the models are production-ready, they can then easily move them into MetricScape for use by clients.
More precise deployment of initiatives
The MetricScape solution provides a new level of precision-targeted pricing, to reach different groups of customers, such as those for whom price will drive incremental volume, and also those for whom the persistence of the deposit and/or utilisation of the credit will yield positive economics over the estimated lifetime of the balance (including the incremental cost of the incentive).
The solution has provided new opportunities for customer acquisition, through digitally refined targeting to find prospects with similar profiles. This has reduced the reliance on broad campaigns that erode value by paying incentives where they are either not required or where the customer’s behaviour will likely not deliver the intended economics.
Having identified target populations based on the appropriate offer (price or otherwise) for balance consolidation, the system’s machine learning capabilities equip the bank to rapidly test multiple dimensions of offers (price level, channel, message, conditions) against target populations, not only to optimise who should get an offer, but also to pinpoint which type of offer is the most economic over time.
At one major US bank with 10 million+ customers where this solution has been implemented, marketing execution costs were reduced by 50% by focusing on high potential customers. Focus on incremental accounts went down, lowering promo expense by as much as 60% and 12-month deposit growth by less than 25%. Reduced requirement for retention offers reduced promotion expense by 10% and balance retention by only 3% with nominal change in customer retention.
The Novantas solution has thus enabled banks to make their deposit business more profitable by leveraging big data and analytics technologies and techniques. Their models combined with the Cloudera Hadoop and Spark’s ability to store and process massive amounts of information at a low cost, have enabled banks to granularly target customers based on their responsiveness to price and the likely retained value from that offer.
These capabilities open up new opportunities for banks and might prove critical in an increasingly competitive environment.