News - 18/07/2016

The Competitive Necessity of Transaction Categorization

‘La revolución de las pequeñas cosas’, or in English ‘the revolution of the small things’. That’s the slogan of BBVA’s ad campaign in Spain highlighting the digital transformation of the bank and the design of its services to promote an experience based ‘peace of mind’ and ‘time well spent’. We take an active part in that transformation.

For instance, in the past people spoke of the tedious task of balancing their checkbook with all the enthusiasm they showed about going to dentist to have a cavity drilled. But now, when a BBVA client logs into their bank account through the web or a smartphone, they can see where their money is going, already categorized and displayed in helpful graphics. These features have been given the moniker of Personal Finance Management (PFM) and automatic categorization of financial transactions has been the first step whose success determines the success of all the following features.

46% of banking clients under 30 said the quality of a bank’s Personal Finance Management would influence what bank they choose.

The first generation of these PFMs suffered from single digit adoption rates, one of the reasons being that the client would have to manually classify their transactions, which made working with these PFMs as much fun as balancing a checkbook. Better applications have been around for smartphones for a while through third parties, such as Mint. But while they do a good job categorizing transactions and therefore financial planning, a user cannot execute transactions with these applications. For that, users have to go to the interface their bank provides. After a slow start, banks are waking up to the competitive necessity of providing a PFM with accurate categorization at its heart. According to a survey done by Novantas, 46% of banking clients under 30 said the quality of a bank’s PFM would influence what bank they choose. Of those who use PFMs, 78% use a third-party product, but 40% of these would prefer to use their bank’s software if the quality was acceptable, mostly for perceived security and the ability to carry out transactions. For example, many third-party PFMs require that you input your banking passwords.

As of today, our data engine classifies daily millions of transactions into 15 major categories with 72 subcategories.

At BBVA we also offer a PFM tool in our online banking website. The tool is powered by the categorization engine developed at BBVA Data & Analytics, which uses a set of big data technologies to process daily millions of transactions, classify them and enrich them with category labels. The categorization taxonomy includes 15 major categories with 72 subcategories. Therefore the hundreds of thousands of BBVA Personal Finance Management users have several layers of granularity to examine where the money goes; first on a general scale, then clicking down to a more specific category and finally to the individual transactions. They can visualize the temporal evolution of expenses and compare trends between categories. Our goal is helping customers a more meaningful view of their personal finances.

The classification criteria were first created by subject-matter experts, encoded into a set of algorithms, and have been adjusted with customer feedback. Now, users suggest classification changes and new tags, which generates a much greater amount of feedback that is difficult for experts to quantify and interpret, at least for human experts. The situation is different for machine learning algorithms and the next generation transaction categorization techniques that we are currently developing. These algorithms will automatically learn the classification criteria from user feedback in order to yield even more personalized experiences. More on that soon.