My bank has been storing large volumes of transaction data and would like to use data analytics tools and processes to uncover hidden patterns, unknown correlations, market trends, customer preferences and other useful business information. Our goal is to find new revenue opportunities, improve customer service and operational efficiency, and gain a competitive advantage over rival organizations. I’d like to hear about your experience in this area and understand the marginal value this data can provide over standard predictive analytics. I need to build a business case to invest in the analytics tools and processes. Please advise…
Big Data Admirer
Dear Big Data Admirer,
Big data has been a hot topic these last few years. Like the prospectors of old, many companies want to mine the data for gold nuggets. So put on your forty-niner’s hat and get ready to learn how to dig for gold!
There are a number of advanced analytic tools available to help your prospecting efforts, but you probably still have an enormous volume of data through which you need to sift. Since your question names transaction data specifically we will not discuss unstructured data which is mostly text based (account notes, emails, etc.) and concentrate on structured data (with defined elements and known values, in a specific format).
Credit card transaction data is a good example of this type of data. However, separating the useful information from the useless information can be a daunting task. Once you find the useful information, you need to translate the data into a data story– something that explains how to solve real-world business problems and the issues your business is facing.
Let’s look at an example to illustrate this. Who doesn’t like a good illustration? After all, most of us loved picture books when we were kids. Here’s the question – can the process of transaction authorizations be improved to lower unnecessary declines using transaction analysis? If so, what is the marginal value of this data over the characteristics we already have. That’s a difficult question to answer without building a new model with the old and new (big data) characteristics to evaluate marginal contribution. New techniques, such as adaptive models, can use new predictive elements to build marginal value into a blended model.
Let’s say you find the combination of a number of transactions under a certain amount over a given period of time predicts increased loss. How would you turn that into a decision element to modify your authorization strategy? You have to be able to implement the results of your analysis to get value. Data that looks predictive in analysis has to be available at the time of the decisioning process and when real time calculations need to be completed to drive operational decisions and actions. If you need raw data to perform extensive calculations that turn it into useful characteristics, the calculations need to be done in real time to drive operational decisions and actions.
New cloud based systems that can use more data to drive smarter decisions, such as Hadoop, have addressed some of these issues. While these new systems are very agile, you need to complete a cost-benefit analysis to ensure you get the value you want.
So, you need to start with a decision in mind and only add new data that helps make the decision. It doesn’t do any good to develop decision strategies and analytics before understanding what you want to decide. Data drives analytics and big data can help you improve decisions that impact performance measures. However, without understanding how the data will change decisions, it’s not possible to know what the value of the data is.
Ask Gini Terms
Content provided in this blog is for entertainment purposes only. Ask Gini blogs do not reflect the opinion of BankersLab. BankersLab makes no representations as to the accuracy or completeness of information in this blog. BankersLab is not liable for any errors or omissions in this information nor for the availability of this information. These terms and conditions of use are subject to change at anytime and without notice.