In the wake of the widespread shift to digital banking and commerce caused by national lockdowns due to the Covid-19 pandemic, fraud teams have seen a sharp increase in cybercrime.

This has been an upward trend for some time. Since 2011, financial crime has more than tripled, to $32 billion in 2020. By 2027, it is expected to exceed $40 billion by 2027. For these reasons alone, banks must do all they can to protect themselves against fraud.

Robust data analytics can be used as a means to not only detect and prevent fraud, but to significantly increase overall business performance. Indeed, by monitoring fraud data, forward-thinking financial institutions (FIs) can build an intimate understanding of their customer’s risk profiles and leverage this information to develop new and tailored products.  

While this is by no means a new concept, what may surprise many is the fact that the information collected by organisations on customers – in order to run their fraud algorithms – is often the most comprehensive available, and can include behavioural, social, geo-location, online and even mobile activity data. This information is a treasure trove of opportunity, and can be leveraged to improve a marketing department’s analytics, and help provide highly targeted campaigns and online recommendations.

When it comes to personalisation-as-a-service, however, the implementation of recommendation engine software is often necessary – which serves content or product offers to customers, based on estimates of what the individual wants or needs. Similarly, loyalty programmes can also be offered to drive customer retention. In the financial services arena, a loyalty programme leverages marketing and fraud data to reward repeat customers, with points, gifts, and other exclusive perks.

Yet, when implementing these digital initiatives, banks will need to overcome a number of challenges. For instance, recommendation engines work most effectively when supplied with in-depth datasets on each customer. As such, FIs will need to unify siloed data pools across their organisation, to create a 360-degree view of their customer base. Building an in-house “digital brain” may provide the answer.

Another, overarching, challenge banks will encounter during this digital transformation is ensuring an appropriate balance is struck between customer service and data security. The key is ensuring enough security measures are inserted into the process, without the experience becoming cumbersome for the customer.

If successful in these efforts, financial services firms can turn the pandemic-induced spike in financial fraud into an opportunity to significantly enhance customer experience and win customers’ loyalty.

Implementing recommendation engines

One way FIs can leverage fraud and risk data to enhance the customer experience – particularly regarding retail banking, credit card, and investment services – is by using it to power a cutting-edge recommendation engine. The larger and more in-depth a bank’s pool of fraud and marketing data, the more accurately recommendation engine software can predict customers’ behaviour, and deliver a tailored product offering.

However, not all recommendation engines are created equal. Installing a simple plugin will not be as successful as implementing sophisticated software to provide bespoke recommendations. Here are three steps financial services firms should take to ensure they deploy a best-in-class recommendation engine.

1. Collect detailed customer data:

The level of product personalisation rests upon how much data is collected from the fraud and marketing teams – including data transactions, social demographic, geolocation, behavioural trends, web traffic, and so on. Naturally, the further back in time the data extends, the better. For example, if data has been collected on a customer for 10 years, as opposed to one, the ability of the recommendation engine to predict the customer’s behaviour will be far more accurate.

There are a few key categories of data that can be used to create personalised recommendations:

• Aggregated data (such as search queries, or category and product views).
• User-specific data (like aggregated data, user data includes specific user interactions, such as which categories and products the user viewed or purchased)
• Static product data (this is supplied by the client in the product feed, which includes metrics like price, availability, and other product attributes).
• Third party data (including demographic, behavioural, social, geo-location, online and mobile activity data)

2. Use AI technology to optimise algorithmic recommendations:

Retail banks can also use machine learning-optimised algorithms to deliver highly relevant product recommendations. For instance, if a customer is completely new to a bank – and limited data on them is available – it may be appropriate to serve the customer a rundown of the institution’s best-selling retail products.

However, if it is a long-term customer, they will prefer to receive personalised recommendations, based on, for example, their previous engagement with the bank’s product suite. This may include products similar to their recently purchased items; products related to their recently viewed products; or even top sellers from their recently viewed categories.

With an innovative recommendation engine banks can not only develop personalised services and integrated offers that address customers’ unique needs – they can be delivered at the right time. This may mean serving competitive mortgage offers when a customer is looking to buy a new house; or recommending high quality growth funds when a customer is saving for their child’s further education, for example.

3. Implement overriding capabilities into machine learning processes:

It is also important that the software enables its users to define recommendation rules for behavioural or demographic segments of choice. For investment services firms, this may involve restricting recommendations to only show technology IPOs, and attractive stock buys, to investors interested in the technology sector.

In an interview with Finextra, Stuart Tarmy, global director, financial services industry solutions, Aerospike, said: “Implementing ‘best practice’ customer-360 analytics and recommendation engines separate the most successful companies from the rest. Indeed, the effectiveness of any recommendation engine depends on three key areas. First, companies need access to a large and robust source of customer data; second, they require finely tuned and performant software algorithms; and third, they need to implement a scalable technology platform that can quickly process the data and provide recommendations in real-time. Often, recommendations need to be executed in under 20 milliseconds in order to meet customers’ expectations.”

“In the most successful cases”, added Tarmy, “we have seen real-time recommendation engines increasing business’ cart size by over 25%, while reducing cart abandonment rates by up to 30%.”  

Clearly, the possibilities associated with recommendation engines are virtually limitless, providing the right amount of data is to hand, and a best-in-class data platform – that manages the entire process in real-time – is implemented.

Providing loyalty programmes

“Loyalty programmes are another example of how analytics can be used to augment the customer experience – rewarding repeat and long-term customers with points, gifts, and other exclusive perks”, said Tarmy.

Credit Suisse’s ‘exclusive club’ is a good example of this – using data analytics to power a personalised loyalty offering on an individual level.

By paying an entry fee, Credit Suisse’s most loyal customers can gain access to high-end benefits and products. Known as the Bonviva Rewards Shop, the loyalty programme gives customers points for every purchase made on their credit card. More exclusive packages – Silver, Gold, Platinum – yield more points, which can be used to purchase an array of exclusive products from the bank’s shop, such as air miles or luxury household devices.

Evidently, when it comes to loyalty programmes, there is plenty to consider. As ever, the key is collecting robust data, and leveraging it imaginatively to enhance the customer’s banking experience in a highly personalised manner.

Unifying siloed data

The more in-house data a bank can gather on any given customer, the more successfully initiatives such as recommendation engines and loyalty programmes will run.

However, this may not be as easy as it seems, argues Tarmy: “While banks already possess a wealth of fraud data, many still need to overcome the challenge of data unification. Often, banks’ data is siloed by business or product line. Pooling this information will allow banks to build a highly detailed picture of every customer, and offer specialise services that each customer truly needs”, said Tarmy. Yet, this will be no easy feat for large, global FIs that are working with fragmented, legacy data infrastructures.

According to a whitepaper from data analytics company, FICO – The Impact of Covid-19 on Fraud and Financial Crime: “Banks are feeling the pain of having fragmented software for managing fraud and financial crime. Even though some 80% of the functions between fraud prevention software and AML software are the same, the systems are nearly always separate, and the teams are usually separate, too.”

In FICO’s survey, 49% of respondents ranked the negative impact of using multiple software systems for fraud management as their first, second or third most significant technological challenge. This is more than any other challenge measured in the paper.

That said, many FIs are beginning to understand the benefits of adopting an integrated approach to fraud analytics, by employing a single data management platform. The FICO survey reveals that 69% of FIs have plans to integrate their fraud management functions or share resources between AML compliance and fraud. Around half of those said they plan to do so within three years.

A digital brain

Arguably, the most effective means to unify all fraud and marketing data within a bank is to build what is increasingly frequently being referred to as a “digital brain”. As defined by SriSatish Ambati, CEO of Open Source and distributed machine learning platform, H2O.ai, a digital brain is “the nexus for continuous, automated learning from data across all business units, departments, product lines and services, that gives the organization higher cognition.”

“In practice, a digital brain collects relevant customer and associated data from within the bank, enriches it with third-party data – such as demographic, behavioural or geolocation data – stores this information in a highly scalable, real-time platform, and utilises artificial intelligence (AI) and/or machine learning (ML) tools to develop a deep understanding of the customer”, said Tarmy. The benefits of this include achieving a 360-degree view of each customer – enabling highly bespoke product offering, without creating a serious drain on human resources.

Talking to Finextra on this concept, Adam Speakman, head of fraud & investigations, Metro Bank, said: “The holy grail of balancing fraud prevention with highly personalised banking services is the establishment of an in-house digital brain”. By pulling together data on each customer, from both the retail and corporate bank, a 360-degree picture of customers can be built.” Indeed, identifying when an individual is both a retail and corporate customer is invaluable to banks, as it maximises cross-sell success rates.

Indeed, the business case for building a digital brain is hard to ignore. Group-wide, harmonsied, data intelligence is becoming invaluable in an increasingly competitive ecommerce landscape.

Yet, this is not to say banks should adopt a Big Bang approach when developing a digital brain. An incremental and concerted culmination of microservices is valuable too – allowing a bank’s architecture to gradually learn from processes, and predict customer behaviours more accurately.

Effective data unification is what has made brands like Wayfair – an ecommerce business that sells furniture and home-goods – successful. Using next generation, NoSQL data platform from Aerospike, Wayfair can connect the dots between its various data points – such as customer activity, product, sales, and event data – in real-time, to create strong sales opportunities. The result is a recommendation engine of the future, and reduced cart abandonment rates.

Big Data: The Digital Brain’s lifeblood

According to Speakman, the lifeblood of any digital brain is data – and plenty of it. Currently, many FIs erroneously view their data as discreet assets that are to remain siloed in various business units and systems. However, the key to building a digital brain is allowing data to flow freely throughout the organisation, so that it can be collected from numerous sources, such as physical branches, mobile, the web, inventory, shipping, and even the complaints department.  

If this is achieved, “banks could easily identify whether an individual is both a retail and corporate banking customer, enabling the sales team to offer everyday banking products, such as savings accounts, overdraft facilities, credit cards or retirement accounts – as well as corporate products, such short-term liquidity lines, loans or accounts receivables financing for their business”, said Tarmy. In turn, vertical FIs can become horizontal ones.

Yet, a next generation digital brain requires vast amounts of data, which must be effectively stored and managed to ensure performance consistency. According to Speakman, one way to access this amount of information is to supplement it with external, third-party data, which must be appropriately standardised and integrated with internal data.

One way to access greater amounts of data is through industry consortia, such as the Economic Crime Strategic Board, of which Metro Bank is a member.

Clearly, the digital brain is the widely sought-after, yet illusive, finish line of digital banking transformation. By leveraging AI, ML, a scalable, high speed platform, and cloud-based technologies, banks have an opportunity to pool Big Data sets, achieve 360-degree views of customers, and, in turn, provide a seamless and personal customer experience.

Balancing customer service and fraud prevention

As FIs increasingly adopt forward-thinking data digital initiatives – such as recommendation engines and loyalty programmes – to enhance their customers’ experience, they will encounter the challenge of balancing customer service with data privacy and data security requirements.

“Due to increased competition and rapidly advancing consumer expectations, banks are under more pressure than ever to enhance the way they interface with customers across multiple channels – from the user experience to serving recommended products”, said Tarmy. “At the same time, however, the regulatory pressure for FIs to ensure customers’ data is secure, is rising.”

Indeed, the General Data Protection Regulation (GDPR), issued in 2016 in Europe, set a foundation for data consent management, and is working to ensure customers’ personal information is safe, handled properly, and not exploited for illegitimate marketing purposes. While GDPR sets a ‘floor’ for privacy compliance – which must be met by all EU countries – participants are able to layer on their own additional requirements, if needed. For example, Germany has more stringent privacy regulations than most other EU countries, and subjects company officers to personal liability if GDPR rules are violated.   

Other regulatory bodies have either implemented, or are in the process of implementing, their own data privacy regulations. The California Consumer Protection Act (CCPA), for example, is closely modelled on GPDR, but has its own nuances. New York State is also expected to introduce its own data privacy laws that will further raise the bar – with the possibility of allowing plaintiffs to initiate class action lawsuits. This is not permitted under current GDPR or CCPA regulations

Yet, to complicate matters further, country and state-level data privacy regulations sometimes work in conflict with other laws. For instance, in Europe, GDPR’s ‘right to be forgotten’ provision – which enables consumers to request that their information be deleted – works in direct conflict with the First Amendment of the US Constitution, which guarantees ‘freedom of the press’, to publish what they see fit.

“These numerous, complex, and sometimes conflicting data privacy regulations place an enormous challenge in front of companies seeking to both handle their data appropriately, and utilise it to enhance customer services”, notes Tarmy.  

Balancing data goals

Indeed, there is a tension here. Designing a seamless customer experience – and delivering it in milliseconds – is key to achieving strong customer satisfaction. However, the quicker the process and the fewer barriers to a loan approval, for instance, the higher the risk of financial fraud, identity theft, and of course, data privacy breaches. If banks become too complacent, and prioritise their customers’ experience, online fraudsters can exploit these systemic weaknesses.  On the other hand, if security measures become too tight, the customer experience degrades – ultimately leading to low levels of retention.

Financial institutions are at loggerheads internally on this subject, too. Marketing teams argue that in today’s highly competitive ecommerce landscape, a smooth customer journey is key to winning new business. The compliance and fraud departments, however, tend to favour a risk averse, safety-first approach, where considerable security measures are taken to ensure customer data is secure, and the bank continues to meet country-specific data privacy rules.

Is it possible to meet both needs satisfactorily? If so, how can this be best achieved?

In an interview with Finextra, Speakman said it is all about using technology to strike the right balance: “The key to balancing fraud prevention with ensuing a smooth customer experience is ensuring the right amount of grit is inserted into the journey. If done correctly, customers do not have to sacrifice a good experience for their data security. The level of grit in the process, however, must be continually monitored against the latest fraud trends. What may have worked several weeks ago, may not work today.”

Nationwide: An efficient, yet safe, payments experience

Nationwide Building Society has been investing tens of millions of pounds on anti-fraud measures – using historic transaction patterns and fraud data to ensure customers’ payments experience is both efficient and secure.

Indeed, payments security has been particularly high on FIs’ agenda during the pandemic, which, according to Otto Benz, payments director, Nationwide Building Society, caused a boom in internet transactions, and in turn, card-not present fraud. This includes smishing scams, whereby fraudsters use a compelling text message to trick targeted recipients into clicking a link and sharing private information.

To combat such fraud, while preserving customers’ payments experience, Nationwide collects and tracks payee data, in order to spot a pending fraudulent transaction – before phoning and informing the customer: “We're looking for clues in payee data,” explained Martin Salter, senior fraud manager, Nationwide Building Society. “If you are illicitly moving money, there's a good chance you're moving it to your own account. So, if the surname doesn't match it’s less likely to be a scam. We also analyse payment references. Fraudsters are creatures of habit – they historically use the same references, such as motorbike, caravan, and car. That's likely fraud, because nobody buys a car and motorbike in a van at the same time.”

“Subtle things, like the specific amount of a transaction also give us a clue that an illegal transaction is occurring,” added Salter. “Fraudsters often send amounts that they think will go under the radar, but they are creatures of habit, often sending similar amounts each time, so this can also give us a clue.”

This historic fraud data is then used by Nationwide in an AI and ML-based scoring system that discerns whether a given card payment is ‘normal’ or not. The score is based on several factors, such as how often the person makes this kind of transaction, the time at which it is being executed, and where it is being executed geographically. “We can tell a lot about the legitimacy of a pending transaction this way,” explained Salter.

Speaking to the impact that Nationwide’s anti-fraud strategy has on customers, Benz said: “We have seen amazing cases where branch colleagues have suspected something illicit is taking place, intervened, and prevented people handing over their life savings. The stakes of failing to fight financial fraud effectively are high. It can mean big costs at an institutional level; customers losing their savings; and even further intervention from regulators – which, if too extensive, can degrade the customer experience.”

“When it comes to fighting fraud,” added Salter, “we’re interested in every kind of non-financial event that happens. For instance, phone calls that were cancelled, attempted sign-ins to online banking portals that didn't go through; anything that might suggest illicit activity.”

Clearly, connecting the dots between otherwise disassociated fraudster activities can be extremely effective in preventing financial crime. “If we can strip out the IP address, we can track fraudster activity across accounts,” concluded Salter. To connect these kinds of disparate datasets, data aggregation and digital brains – as previously discussed – can be extremely useful.

Metro Bank: Using fraud data to streamline call centre experience

Metro Bank, meanwhile, has invested heavily in fraud prevention across its telephony channel, among other areas. By closely monitoring and analysing its historic scam call data, the bank has been able to improve customers’ experience over the phone. “We have revolutionised the way we identify our customers, which has significantly reduced handling times”, said Speakman. “For example, we can quickly detect when a risky call is inbound. As such, we prevent forcing every legitimate customer to jump through hoops just to bank with us.” Metro Bank applies this same filtering principle to all its other channels, such as the mobile banking app.

This is a prime example of how, with the right amount of “grit” inserted into the process, customer service and fraud prevention can be effectively balanced in order to satisfy all stakeholders’ needs.

Once again, cutting-edge technology is the solution to the challenge of balancing customer service and fraud prevention. With nuanced fraud-detection modelling, millions of interactions can be vetted within milliseconds. This is done by comparing numerous datasets – such as customer behaviour, transaction history, geolocation, and so on – in real-time, in order to help separate non-criminal from criminal activity.

Using this approach, a large portion of fraud-management processes can be automated – leaving marketing teams to dedicate more of their time to customer-facing, revenue-driving activities.  

The journey begins now

With an end-goal in sight – the creation of a digital brain – banks have both a direction and an incentive with which to begin their digital transformations. While technologically demanding for some institutions – especially those with legacy data infrastructures to update – the potential rewards are considerable.

As we have demonstrated, by unifying siloed fraud and marketing data, FIs can not only build a robust defence against all manner of rising financial crime trends, but they can also build a 360-degree view of customers – the bedrock of personalisation-as-a-service. Through accurately predicting customers’ wants and needs year-round, highly effective recommendation engines and loyalty programmes can be implemented, which have been shown to turbo-charge customer satisfaction rates, customer purchases, and customer loyalty.

Ultimately, the answer to data analytics is ensuring teams within banks’ sales, marketing, and customer-service departments are fully aligned with finance, compliance, and operations teams, when it comes to effective financial crime detection and delivering a smooth customer journey.

As ever, the lifeblood of this effort is effectively managing large amounts of data, and a best-in-class data management platform is the key to unlocking FIs’ full potential.