News and case studies on embedded finance, APIs and much more
4 min reading
14 JULY 2016
Network searches and data consumption, telephones, smart watches and bracelets, mobile devices such as tablets and GPS, the Internet of Things, social networks… The number of everyday recipients of personal data have grown exponentially in recent years. Today, users and companies have more unstructured data available than ever in the history of humanity. And they have Data Science, the most efficient, quickest and cheapest way of ordering and analyzing them to extract conclusions that are useful for business.
In the present and future scenario for banks operating as Platforms as a Service (PaaS), their reception, management, structuring and analysis of big data will become a competitive advantage with respect to others. It is not only the revenues from the use of APIs by third-party startups, there are also end products (applications) that accumulate personal data, consumer habits and day-to-day operations that are a paradise of opportunities. To a large extent it will make it possible for banks to reinvent their business.
The customer, the big data center
Big data management has a number of objectives in the case of what is obviously the focus of all banking operations: the customers. The primary target of the mass collection of data, their structuring and analysis is the identification of customer profiles. It is possible to discover what customers consume, their interests, their needs… And with this it is easier to adapt marketing campaigns to the different customer profiles and improve services to the extent that they can be personalized. In the end the banks, like any company, want to improve their brand image.
The other two major objectives are to understand how customers relate to the financial products and what their real commitment is to what their bank offers them, whether it is a personal loan, relations with employees or a cell-phone app. Third, an not least, the aim is to detect who the tired or unhappy customers are, and in the last resort those with a high likelihood of abandoning one bank for another financial institution. The use of machine learning, an effective mix of big data and artificial intelligence, allows banks to prevent customers from leaving them.
Today banks use numerous types of algorithms to predict conduct related to customers, and even to their own employees: decision-making trees, clusters, neuronal networks, text analysis, links and searches or survival analysis are methods used to improve the experience of consumers or retain them. In this case, survival analysis is the method used by banks to establish the moment when a user can leave a bank. It is able to analyze millions of the bank’s user data and establish the customer life cycle (the analysis normally has two elements, one on a scale of 1 to 0 that measures commitment and another that establishes the duration of the relationship between the financial institution and consumer). Some of the elements that survival analysis can provide an answer to are:
● When a specific customer could leave a bank.
● When the customer should be moved to a new segment with new services and benefits.
● Effects that facilitate a better or worse relationship between the bank and users.
Big data and customer segmentation
Customer segmentation is the method by which a financial institution can create groups of consumers who share needs and interests. This is the path towards personalized banking. Trying to adjust financial products to the needs of each individual means chasing a mirage; but segmentation by groups can at least bring the banks closer to the goal of adapting the financial offer.
A normal example of this for banks is the collection of data on the use of credit cards and analysis of consumption habits based on this use. Banks can adapt their offer using this work with big data; but they also establish price scales by financial product according to the type of user, for example for segmentation of applicants for loans (each is prepared to pay the right price for the offer).
The most commonly used big data method in customer segmentation is K-means clustering, a clustering method used in data mining to make subdivisions of a set using different observations, leaving clusters around the nearest mean. It is the most effective form of creating typologies of users or customers around market trends.
Analysis of feeling
Social networks have become an ideal scenario for searching, informing, offering and understanding how consumers feel. But there are millions of users giving millions of opinions at the same time on numerous platforms (Twitter, Facebook, LinkedIn, etc. ), not only in the social networks, but also through comments in forums or news aggregators, among others. Data collection and analysis methods are essential for measuring the temperature of this environment and for taking specific measures to build customer loyalty or solve reputational crises…
There are two perfect algorithms for analysis of feeling:
● Naive Bayes classifier: it is a probabilistic classifier based on the Bayes theorem and simplifying hypotheses. What does this mean? There is a saying that sums up perfectly how the naive Bayes algorithm works: if it looks like a duck, swims like a duck and quacks like a duck, then it probably is a duck. This naive classifier establishes how each of these characteristics contributes independently to the probability of the final premise.
● Support Vector Machines or SVM: a set of supervised learning algorithms developed by Vladimir Vapnik in AT&T. He is now working in the artificial intelligence team in Facebook. It is a very commonly used data mining method in machine learning: based on a set of sample data a support vector machine can be trained to predict the classes of a another set of data. This big data method allows, for example, a forecast of customer defaults within risk management.
If you want to find out more about BBVA’s financial APIs, go to this website.
