Retail banks are in a unique market situation where they have expansive customer bases across numerous segments and possess vast data sets. Yet, only 10% of these banks are deploying artificial intelligence and machine learning (AI/ML) models at scale used for any competitive advantage.
This includes the ability to glean more insights on customers, harmonise traditional and digital channels, or personalise offers. This lack of utilisation stands in stark contrast to the retail industry, who faced similar challenges a few years ago.
Data success in retail
By applying machine learning and data science models, Amazon has been able to exceed a 35% online conversion rate by creating personalised content and recommendations around its products across its digital channels. In the same vein, Netflix reports that 80% of its users follow its recommendations rather than using search, bolstering customer retention.
In both cases, personalisation contributes significantly to their revenue streams. In food and beverage, the Starbucks mobile app sends over 400,000 variants of hyper-personalised content to their loyal customers.’ Such a vast repository of food, beverage, and seasonal offers will provide the allure of a 1:1 personalised experience.
By Q4 2022, the Starbucks mobile app had been adopted by 27% of its North American customers. And its mobile app (e.g., order, pay and delivery) accounted for 72% of total sales volume in Q4, marking a substantial year-on-year increase of 35%.
The financial performance, buoyed by mobile apps, data and personalisation, defied analyst expectations.
What can banks learn?
Banks can learn from adjacent sectors like retail and adopt best practices regarding personalisation. The first step towards this transformation involves creating more robust data management structures that aggregate data sources to provide a comprehensive customer 360.
Data can also be democratised and decentralised to improve timelines and usefulness for individuals, teams, or departments. The idea is for the lines of business to have access to the most recent file and for this data set to be closest to the source, even in original format.
The advantages are in having access to data in real-time, contextualising data for the most relevant or highest value use cases for multiple business units.
Real-time data can support any number of sales and marketing campaigns, and stream contextually relevant next-best actions to contact centre agents to improve customer service. Security teams have more tools in place to detect fraud, improve overall cyber-defences and multi-factor authentication.
Catering for the next generation
A starting point can include customer demographics, purchase history, location, time, search and browsing times across digital channels.
Since the pandemic, mobile banking apps have been the primary means for engaging customers, especially from Gen-Y and Gen-Z customers who try and use this channel almost exclusively.
This group is the banking sector’s fastest growing segment and have the loudest voice in the room when it comes to digitisation. They are also the most prone to shift allegiances if their digital demands are unmet.
Banks must therefore ensure their user interfaces and personalisation strategies meet or exceed standard set by the retail industry and other digital frontrunners.
Unlike Gen-X or Baby boomers, this latter segment consumes experiences, not material things.
Paytm, an Indian-based financial services increased the sales and click-through rates of its ecommerce service, Paytm Mall, by implementing personalized recommendations on the homepage of its site.
Since deploying Amazon Personalize, a managed AI/ML service to create a personalization model that generates recommendations for each customer, the company increased conversion rates to between 5.5 to 6% - which is nearly three-time the success legacy banners ads were able to achieve.
The good news is banks have the platform and the audience. Data and analytics are the binding force connecting what already exists.
Collaborative filtering, for example, is a common technique used to predict products based on collected data sets from other businesses with similar profiles or preferences (e.g., businesses who bought Y, also bought Z) to create the right product package.
Hyper-scalers, FinTech’s and digital-only neobanks that are new entrants in the banking market have a distinct advantage. What they lack in name recognition and brand, they make up for in arguably something more powerful in today’s evolving digital landscape: agility.
Built on cloud-native technologies, such as containers and micro-services, these modern entities are rolling out services much faster at lower unit costs, with lightning speed automation (e.g., opening a bank account in 90 seconds) and more nimble organisational structures that reflect a start-up over a bricks and mortar legacy brand.
Inadequate technological infrastructure for customer 360
Mainframes have underpinned core banking since the 1960s and are prevalent to this day.
While they have delivered robust and secure capabilities for maintaining a ‘system of records,’ they become the Achilles Heel when banks need to service customers through a ‘system of engagement,’ the prerequisite for digitalisation in modern banking.
Any personalisation that can be done tends to be limited to a single set of products, or an individual promotion. This approach yields significantly less data sets for a model to draw on, dramatically reducing the potential for ‘trigger’ or push recommendations, such as a next-best action, promotions, and custom content.
Without a comprehensive view of customer behaviour and preferences through an abundance of data, average campaigns will yield mediocre results at best, and the big picture metrics, such as customer lifetime value (CLV), will be overlooked.
Cost of compliance inhibiting new technological investments
With the rising cost of compliance, banks’ priorities are increasingly skewed towards managing existing products and mitigating risks, leaving less room for change. A recent study estimates that banks spent USD $274.1 billion in financial crime compliance alone, but this is only part of the picture.
A growing set of requirements have a profound impact on the types of technology incumbent investments banks can make, how and where they can be deployed, level of security protections and who can access what systems. While compliance is critical to the operation of a bank and protecting customers, compliance costs and new requirements continue to climb unabated.
As such higher operating costs and lower customer returns reduces the available budgets required to invest in digital technologies necessary to reverse the fortunes.
Fragmentation and skill gaps
From an industry perspective, there are other uphill battles. Banks and their peers have no real standard or protocols on how to build, deploy, deliver, and govern AI/ML models at scale. This market is as fragmented as the APIs ecosystem.
The absence of robust frameworks to govern the use of data and analytics further complicated the picture.
Even if these systems were in place, data quality is often a barrier and then there is the lack of internal skilled resources, such as data scientists and programmers, to integrate data to a compelling AI-driven campaigns. Many banks as well as multiple sectors also centralise and secure data management with a single department, such as IT.
As a result, data is not used on the frontline where it is needed the most - despite banks having the means to be as agile and data-driven as their modern counterparts.
Real-time data streaming can connect the dots
Taking into account the barriers, the future of banking is in the ability to interact with data in real-time and to expand usage across all channels. This includes traditional, such as in-branch, kiosk or contact centre agents and digital including mobile apps, online and social.
The starting point is having a corporate strategy around data, beginning with governance, data ownership, and access.
With governance and business processes to one side, the technical solution architecture is designed to retrieve data from multiple sources. Think of it as an additional layer that manages and coordinates data access across these sources.
By keeping the data in its original format and location, banks can access it faster and maintain its user-friendly nature. Unlike traditional methods that consolidate all data into a central 'data lake', this approach keeps data distributed across its original sources. This means we don't have to move everything to one central location before analysing or processing it.
Data should be seen and treated as a first-class citizen and used by the stakeholders who have a best understanding of the data asset and applicability to their role in the business or the customer opportunity.
83% of financial services firms believe 'accessing real-time data streams were either important or extremely important for building rich customer experience.
Other benefits found in the study include improved operational efficiency, higher customer engagement, faster reaction times and better overall insights.
Moving data from one location into a centralised structure, converting into different formats before streaming back to the front line will add considerable time (this is referred to as the data time gap).
Levelling the playing field with speed
Today, many organisations experience a data time gap of many days. The impact of a high data time gap is that events are less likely to be in the moment and contextually relevant. There are also likely to be added investment costs to this approach.
Distributed data management platforms – data streaming, data visualisation, data catalogue and active meta data platforms – as a starting point, is important for incumbent banks to level the playing field against their neobank challengers.
Changing the data paradigm from traditional request response architecture to include event-based architectures, is key.
Event-Based Architectures offer multiple use cases beyond ‘in the moment’ recommendations to supporting other areas of business banking where a business process can be improved with real time data and by enacting AI/ML models in real-time.
Emerging use cases include high-speed trading, payment processing, fraud alerts and customer notifications like order tracking. Some banks have started to use this in credit scoring, merchant services, claims processing, automating compliance reporting for, such as EU MiFID II in the case of one bank, and auditing.
Research has shown that personalisation through AI/ML-based models and algorithms, can lower the rate of customer churn and increase annual revenue uplifts from 10% upwards.
Another study on marketing campaigns that used this approach reported 5 to 15% higher revenues while launching them two to four times faster.
Data integration for customer engagement
Integrating disparate data points into business processes and across the value chain is the future for banks – and many other customer facing sectors - but requires a solid understanding of active meta data to realise the full potential.
The mission is to better capture, process and respond to events in session (to move from one channel to the next) and in context (having the trigged action match the customer need).
While the immediate requirements will support consumer and business banking in improving CX and harmonizing channels, new use cases are emerging in areas such as capital markets.
This evolution in strategy allows for the design and customisation of investment strategies, as well as wealth management through predictive actions, as institutions like Morgan Stanley have learned.
Focus on SMEs too
While consumers are an attractive starting point, over 95% of all businesses are generally the small and medium enterprises. They often show the highest propensity for cross-selling services and profitable.
This segment will tend to have more data sets than consumers, for example, that can be used for hyper-personalisation. Business checking, savings, credit cards, loans, merchant services, commercial real-estate, and interbank transfers show the strongest potential for elevating a bank’s ability to serve this segment - which is often underserved and overlooked.
Introducing a long tail recommendation engine from product research to point of sale based is a strong entry point for banks. Other sectors, such as telecommunications start here to increase revenues per client, overall stickiness and enable other strategies that lower churn rates.
Also, creating a long tail of recommendations, such as Amazon and other online retailers have learned, deliver an immediate ROI for investments by moving otherwise static products.
Play the long game
Banks should be careful not to devote their entire analytics and personalisation capacity to pushing short-term product sales. Delivering personalised insights that offer support and guidance with no strings attached builds trust and helps banks to become clients’ provider of choice – and then product conversions come naturally.
While AI/ML-based models for personalisation and customer insight is the future of digital banking, skills are often in short supply.
This is especially the case in creating a data governance model to support fast moving to real-time streaming, re-aligning business processes to support new agile operating models, and creating frameworks to de-centralised data ownership structures so that data can be unlocked and used for better outcomes.
- Jaffery, Bilal. Deloitte, Omnia AI Practice. 2020. Connecting with meaning.
- November 2022. Starbucks (SBUX) Q4 2022 Earnings Call Transcript.
- Structured, Unstructured or Semi-Structured.
- 2022. Amazon Case Study. Paytm Boosts Homepage Sales with Personalized Recommendations Using Amazon Personalize.
- 2022. LexisNexis, True Cost Financial Crime Compliance 2022 | LexisNexis Risk Solutions
- Like oil data brings enormous potential when refined, processed, and contextualised from its raw state
- 2022. Confluent. Business Impact of Data Streaming: State of Data in Motion Report 2022
- Brodski, Sonia (2019). BCG. What Does Personalization in Banking Really Mean
- McKinsey. 19 July 2022. Getting personal: How banks can win with consumers | McKinsey