Remember when you began your career and the prospect of retirement was an event in the distant future? How many of the poor decisions you made over the years could have made for a better retirement outcome had you had a crystal ball to see into the future? With better knowledge about the future, would your decisions have been different?

For many career professionals, the reality of retirement is now significantly closer. While it’s easy to look into the past and wish we’d had a crystal ball, time is better spent improving the quality of the decisions that impact our future.

A personal crystal ball that predicts your days ahead is what financial services firms everywhere want. Every day, these companies pose questions such as:

  • Will this new client provide a good return on investment, relative to the potential risk?
  • Is this existing client a termination risk? What actions can be taken to prevent their churn?
  • Will this next trade return a profit?

Just as looking in the rear-view mirror on your career doesn’t significantly improve the quality of your retirement, simple historical analysis will not answer strategic questions such as those posed above. Just like people, companies are seeking a crystal ball. Fortunately, advances in analytic technology have made the ability to see reliably into the future a reality.

Business applications & the birth of BI

Since the dawn of business applications, the fundamental purpose of these applications has been to increase the efficiency of business processes. Instead of transacting business with only a paper record, enterprise applications recorded transactions in a computer database. This innovation enabled the rapid recording, retrieval and updating of a company’s business transactions. Because the data describing each transaction was in a database, this made it easy to retrieve and summarize multiple transactions together. This data retrieval and summarization capability gave rise to what we now know as the business intelligence industry. Today, the most common usage of business intelligence is for the production of descriptive analytics. 

Descriptive Analytics: Valuable but limited insights into historical behavior

The vast majority of financial services companies use the data within their applications for what is called “Descriptive Analytics.” Descriptive analytics is the process of selecting and often aggregating transactions together for the purpose of understanding historical activities, trends, performance and behavior. Descriptive analytics are useful because this method of analysis enables financial services companies to learn from past behaviors. Descriptive analytics techniques are often used to summarize important business metrics such as account balance growth, average claim amount and year-over-year trade volumes.

Because of the proliferation of application data and business intelligence tools, performing descriptive analysis no longer provides a significant competitive advantage in the financial services industry. As every company is using descriptive analytics with application data, the competitive edge gained by doing historical analysis has essentially disappeared.

This does not mean however that a competitive advantage does not exist in using historical data. What it does means is that a different type of analysis is required to regain the competitive advantage. An analytics alternative that goes beyond descriptive analytics is called “Predictive Analytics.”

Predictive Analytics: Predicting Future Outcomes

While descriptive analytics are focused on historical performance, predictive analytics are about predicting future outcomes. The foundation of predictive analytics is based on probabilities. To generate accurate probabilities of future behavior, predictive analytics combine historical data from any number of applications with statistical algorithms. The output of these algorithms, when used in financial services, can be anything from a customer behavior score to a prediction of future trading trends, to flagging a fraudulent insurance claim. A properly designed predictive analytics algorithm, when combined with a breadth of application data, will deliver a significant competitive advantage in the financial services industry. 

Credit Scoring: A predictive analytics example

One predictive analytics application familiar to many is the credit score. Credit scores are used by financial services companies to determine the probability of customers making future credit payments. Credit scoring algorithms typically combine a number of historical behavioral attributes, from different transaction systems, when generating the score. These attributes often include:

  • Payment history: A number of late payments can lower score.
  • Public records: Bankruptcies, judgments, and collection items can lower score.
  • Length of credit history: A longer credit history improves score.
  • New accounts: Opening multiple new accounts negatively impacts score.
  • Inquiries: A large number of recent inquiries may negatively impact score.
  • Accounts in use. Too many open accounts can have a negative impact on score.

The credit scores generated by the predictive model are then used to approve or deny credit cards or loans to customers. A well-designed credit scoring algorithm will properly predict both the low- and high-risk customers. Because these algorithms are of significant competitive value to each financial services company, the specific attributes, weighting, and logic used within the algorithms is highly guarded.

This credit scoring example illustrates that while the historical activities of a credit applicant are of some interest, the real value of the historical information comes from combining the data with algorithms to produce predictive analytics. Predicting, with a high degree of confidence, how an applicant will behave after credit is granted or rejected is of far higher value than knowing only historical behavior. The result of the synergy of historical data, together with the predictive algorithm can be significant competitive advantage to any financial services company using predictive analytics.

What competitive advantage could your company gain by moving from simple descriptive analytics to predictive analytics? How can you create your own predictive score, using data from your applications, to improve the performance of your organization? Below are the “Seven Steps to Success for Predictive Analytics” to making predictive analytics a reality in any financial services organization.

Seven Steps to Success for Predictive Analytics in Financial Services

1. Identify the metric you want to influence through predictive analytics.

What business metric determines the success of your organization? If, for example, you want to grow your account base but your growth is constrained by account cancelations, the metric you might want to influence is “Number of Accounts Canceled.”

2. Determine which attributes, behaviors or other metrics most influence this metric.

Before a customer terminates business with you, there may be any number of behaviors that the customer exhibits, which may be a precursor to their actual cancelation. Could it be an account balance change? Could it be a number of calls to the call center? Could it be a comment on social media? Consider all customer interactions and their data sources as potential sources for predicting future customer behavior.

3. Integrate the data sources of the various behavioral attributes into a functional data model.

Integrating the various data sources that represent the entire customer experience is essential to building an accurate and reliable predictive model. Sources should include transactions systems, ERPs, CRMs, support ticketing systems, financial systems and social media sites. This may involve integrating different technologies, like cloud sources, on-premise databases, data warehouses and even spreadsheets.

4. Add the predictive logic to the data model.

With the source data now fully integrated into an analytic model, add and test different predictive algorithms. Open source technologies such R and Python can be used to develop predictive scoring algorithms. Once an accurate predictor of future behavior is identified, integrate the scoring measures directly into the data model. The integration of historical data and predictive analytics is key to operationalizing predictive capabilities in large financial services organizations.

5. Create the reports & dashboards needed to visualize the predictions.

With the data for both descriptive and predictive analytics fully integrated, create the reports and dashboards necessary to visualize past and future performance to business users. Using the account cancelation example, create reports for call center agents that easily highlight those current customers that are at risk of canceling, based on your predictive algorithm. By providing proactive customer care to potential at-risk customers, cancelation may be averted.

6. Automate the data processing sequence.

With connectivity, data integration and the predictive algorithm in place, schedule the entire process to update on a daily or more frequent basis. Having the most recent data from all sources ensures the predictive model will generate the most accurate predictions.

7. Enable end users with access to the predictive analytics.

The biggest competitive gains in financial services companies are realized when the solution is deployed to every employee to whom the predictive analytics could be useful. Often, that means deploying the solution to every customer service agent and every sales representative in a company. Delivering these predictive analytics in an easily consumable medium, like web-based dashboards, ensures that their value will positively impact the maximum number of customers.

Summary

While analytics that describe historical performance provided a competitive edge in the financial services industry 20 years ago, with every company using generic business intelligence tools today, this advantage has evaporated. Just as career professionals plan and anticipate a successful retirement by making educated guesses about the future, financial services companies also need information to help their employees make better decisions. Predictive analytics can become the elusive crystal ball that companies are seeking. Through the use of predictive analytics, enabled with multi-source data integration, financial services companies can reduce risk, reliably predict future outcomes and compete successfully in the global market.

About the author:
Richard Reynolds is a Senior Director of Product Strategy, focused on the Financial Services Industry, at Birst. Richard is a veteran of the BI industry, having worked with analytics and data warehousing solutions from Business Objects, SAS, Teradata and SAP.   Richard specializes in dashboards, predictive, and prescriptive analytics for the modern enterprise.  In his BI role, Richard has guided many customers to obtain business value from their enterprise data.