To achieve the personalisation and the agility outlined in my last blog, and demanded by today’s customers, banks must look to automation. The only way to replace one-to-one branch relationships, and retain high levels of customer satisfaction, is to automate intelligent conversations with every customer. Banks have been moving in this direction for several years, and most use predictive analytics to automate at least some processes. But whilst they might have a few tens of analytical models running in they are quickly going to need thousands of models that support the automation of vast swaths of day-to-day banking interactions with customers.
Creating the automation needed to service these customers requires a focus on processes, platforms and people. And the time to act is now. The COVID crisis has dramatically accelerated both the desire and the comfort levels of customers dealing with banks through digital interactions. Recent research reveals that Internet banking is rated as a very important channel for banking relationships by 84% of customers. Mobile is not far behind on 77%. The same report found that an ‘omnichannel experience’ was the top expectation of 76% of customers. Automating personalisation to meet these new demands should be high on banks’ agendas.
The essential pre-requisite to delivering intelligent conversations, is adapting the processes within the bank. Too much data analytics is still undertaken at the department level. Models are created to automate mortgage solutions, for example, creating data assets which could be of use elsewhere (for example in risk and fraud teams) but are locked up in departmental silos. Conversely, data used to create those models is integrated and engineered for that specific model, most likely repeating work already done for very similar models. By creating Enterprise Feature Stores that can catalogue, manage and shared reusable data features, banks can dramatically increase efficiency in building models.
There is also mindset shift needed in creating agile, automated decisions. Banks need to embrace analytics model-based decision making as the norm rather than the exception. Real-time scoring of customer data, from varied channels is both possible and essential for the conversational analytics that will drive growth. It should be deployed as widely as possible, not restricted to a few siloed processes. As much ‘business-as-usual’ banking should be automated, with well designed, well governed, efficient and flexible models making the majority of decisions without human intervention.
Personalised, intelligent, automated conversations with millions of customers require data integrated across multiple channels at scale and at speed. For this automation to be effective, data must be collected, integrated, prepared and fed to analytic models and decisioning systems in milliseconds. These in turn must trigger interactions so that new offers, processes and decisions are presented to customers whilst they are still in the app or on the webpage.
The data platforms needed to support these ‘contextual analytics’ must be both scalable and flexible. Moves to cloud-based architectures may provide the ideal opportunity for banks to upgrade the provisioning of hardware and software services to support these always-on, near-real-time interactions. Cloud architectures are one route to flexible provisioning that allows analytics capacity to grow and shrink to support those conversations when it suits the customer.
But alone, basic cloud architectures are not sufficient. Banks need to ensure that the architectures they build are not only scalable but support fast and efficient scoring of millions of models in real time. Without the massively parallel processing capabilities of Teradata many cloud-based solutions cannot meet the demands of contextual analytics driving the billions of predictions necessary to deliver seamless customer experience.
New roles for people
People are the final part of the jigsaw. Certainly, there will still be a requirement for human interaction to manage more complex financial products and situations, and to handle the exceptions and edge-cases that it would be too expensive to model for automation. But removing manual, paper-based routines and tasks will help drive out significant costs for banks. The role of data scientists and the wider data management functions within banks will also change. Integrated data platforms need integrated data science teams. Already the model is shifting to agile development where small teams of experts in data modelling work directly with business functions to quickly create data products that meet immediate business requirements. Working with Teradata many banking clients are now embracing this ‘DataOps’ approach, along with the enterprise feature store concept, to rapidly transform their operations and deliver automated intelligence conversations.
Leaders embracing automation
For example, partnering with Teradata, one major European bank has created a true 360-degree view of the customer by consolidating data from across the bank into a single integrated customer data platform. Data from all channels (web, mobile app, call centre, branch, etc.) plus customer satisfaction evaluations feeds an intelligence hub which uses analytics and machine learning to deliver frictionless customer experience. This in turn delivers customised offers, next steps and services, through all channels in real time. All of this is automated at scale and speed. As a result, the bank can capture every data event and customer interaction in real time – over 20GB of rich data every day – and use it to automatically personalise nearly half-a-billion customer sessions. Creating these intelligent conversations, based on contextual analytics, delivered 30 per cent more conversions than in a control group. More than €5 million of additional new business was secured from the 50% of customers who were engages in these intelligent conversations than from those who received ‘traditional’ marketing communications.
Customers are not only moving to digital channels, but they increasingly prefer them. As argued in previous blogs, banks must move fast
to deliver through these channels whilst maintaining, or ideally improving, customer experience. Automation is no longer just a solution for a limited set of highly transactional tasks but must become the standard approach for most interactions. Only then will banks be able to square the circle of improving customer experience and defending against new competition whilst simultaneously reducing costs. Improving cost income ratios will be the focus of my next blog.