Banking Fintech

How machine learning brings magic to next-generation PFM

How machine learning brings magic to next-generation PFM. Image: Freepik
Written by Pau Velando

Machine learning embedded into PFM solutions can provide deeper insights and benefits for the bank and its customer. Pau Velando explains more.

Personal financial management (PFM) tools have become a convenient solution for individuals and businesses to get a holistic view of their transactions and overall financial situation. It aggregates information from our transactions and allows us to see where we spend our money, the value of our assets and what liabilities we have, but those are just basic functionalities of first-generation PFM solutions.

Next-generation PFM goes a lot further. When logged into online banking, you’ll see your expenses automatically categorized, your expected cash flow for the coming weeks, and even your financial performance against your peer group, even if you haven’t defined it. So, what’s actually behind all of these next-level insights?

Tangible insights = end-user delights

On the surface, machine learning (ML) algorithms seem to bring a kind of ‘magic’ to analyze and display PFM users’ financial activities and profiles. As these algorithms work in the background, they constantly ‘learn’ patterns that are used to autonomously forecast income and expenses, categorize transactions, simulate financial scenarios or recommend actions that are in the user’s best financial interest.

For the user, the benefits of ML embedded technology outweigh the drawbacks of volunteering their data
Another feature that contributes to the smarter use of PFM is the possibility to aggregate accounts from different institutions, and its transactions, products and services. At this point, the solution shows even deeper insights when it classifies, predicts or recommends based on data coming from another institution.

All these events undoubtedly bring some magic to the experience of the user when interacting with online banking, but also trepidation in wondering how the system knows this or suggests that, particularly if it comes from outside the banking environment. Thankfully, we have enough tools to make ML algorithms work in ways that protect personal information and respect privacy.

For the user, the benefits of ML embedded technology outweigh the drawbacks of volunteering their data (which, of course, is always consensual and highly secure in an online banking context). The algorithms work in the background, transparently, analyzing the data generated by the activities of thousands or millions of users as they use banking services. The end results bring wider and deeper insights that can help end-users manage their finances more intelligently, and with significantly less manual effort.

Driving digital transformation in banking

Not only does ML technology bring magic to the end-user, but also to the bank. ML-powered financial services can uncover a host of powerful insights, revealing exactly how the customer interacts with the bank, from spending patterns to purchase behavior, saving appetite and even investment acumen.

When applied to financial services, machine learning algorithms can uncover new spending patterns among its pool of clients, indicate previously unknown forms of interaction, discover new segments of customers and unlock new insights within the bank’s omni-channel efforts.

Personal data protection is essential and needs to be communicated thoroughly in Terms and Conditions
PFM solutions that embed cutting-edge ML capabilities suddenly become a key driver of digital banking transformation, impacting every channel, from web to mobile, from branch offices and even to the bank’s back office with products such as recommendation engines or card-linked marketing tools. As tantalizing as it sounds, paramount to all of this must be the user’s privacy rights. Personal data protection is essential and needs to be communicated thoroughly in Terms and Conditions, and strictly respected at all times. Yet, how can analyzing customer data, cross-examining spending patterns against peer groups and even generating personalized offers be done in a way that avoids violating customer privacy?

To meet this essential requirement, advanced analytic algorithms must be trained with anonymized data, meaning that the pool of transactions and demographic attributes that are often used to build the learning of the systems need to be pre-processed so that it doesn’t contain any data that can be used to identify individuals.

The machine learning future of PFM

Next-generation PFM solutions have embedded machine learning algorithms that go far beyond the simple reporting of transactional data. Seamlessly integrated within the back-end of core banking systems, ML technology works to bring insights directly to end-users and banks in a secure and transparent way.

For users, it can help them discover spending patterns, compare themselves to groups of people with similar financial profiles. ML solutions can also help banks to better understand how their customers interact with their products and services.

If banks want to leverage the large amounts of data at their disposal, embedding machine learning within their digital offering is an extremely powerful step forward. The reward for banks and users is in obtaining deeper insights and behaviors faster and better than otherwise possible via traditional methods.

– This article is reproduced with kind permission. Some minor changes have been made to reflect BankNXT style considerations. You can read the original article here.

About the author

Pau Velando

Pau Velando is in charge of channel partnerships management at Strands. The majority of his professional career was spent as a Big 4 management consultant between Barcelona and Tokyo. In 2009, he founded a company that develops software solutions to prevent fraud and money laundering. In addition to managing partnerships at Strands, Pau teaches finance at the UAB School of Business and Economics.

1 Comment

  • The biggest issue with this taking off will be compliance rules. The regulators are (currently) going to insist that the bank acted in the customer’s best interest and that the bank can demonstrate that. This means a decent audit trail of why the algorithm made the recommendation it did when it did in time and there are only one or two firms in the world with technology capable of this currently. The other problem is that the more sophisticated the algorithm, the harder it is to reverse engineer (thus explain) why it made the decision it did? The best systems are using 8000-12000 data points including mobile, social media, user behavior for say a loan decision. It’s relatively easy to build an train an algorithm, it is more work to develop a full audit trail so that any individual decision can be looked at in a snapshot of time.

Leave a Comment