According to industry headlines, the answers to many challenges facing credit unions today lie deep within their member data.
However, with volumes of data spanning a credit union’s systems and applications – and multiplying by the minute – bringing it all together under one technological roof is easier said than done.
So how can credit unions better manage their data, implementing the right strategies and infrastructure to transform data into both operational efficiencies and better member experiences?
“Machine learning technology is quickly advancing and promises to benefit credit unions and their members in many important ways – from fraud detection and risk management to member services and marketing,” said Phong Q. Rock, Sr. VP, corporate strategy and business development for Feedzai. “However, leveraging all that machine learning has to offer requires credit unions to first ensure the quality of their data.”
Taking Control of Data
According to Rock, a successful machine learning deployment depends on data integrity, security and accessibility.
“Data integrity is defined as the validity and accuracy of data. Security, of course, is the act of protecting data from being hacked, altered without authorization or otherwise corrupted in some way,” he said. “This can be achieved through a variety of methods, including backup and replication, database integrity constraints and validation processes.”
For machine learning to deliver on its promise, though, data must also be accessible across the enterprise. Toward that end, many financial institutions are building what are known as “data lakes.”
Defined by Computerworld UK as “large, unstructured data sets,” data lakes provide a single data source for the entire computing environment, replacing the “silos” of data that can grow and evolve over time within separate systems and applications.
“Credit unions should embrace the principles behind data lakes to ensure information is easily shared across applications, including new machine learning technologies they wish to deploy for uses such as fraud prevention,” said Rock.
McKinsey analysts agree. The firm’s report, “The Age of Analytics: Competing in a Data-Driven World,” emphasizes that successfully implementing machine learning relies on data “collection or generation capabilities.” These capabilities may, in turn, require organizations to transition from “legacy data systems to a more nimble and flexible architecture to store and harness big data.”
Selecting a Scalable, Flexible Machine Learning Solution
Once data is clean, secure and accessible, Rock advises credit unions to approach their machine learning initiative systematically, starting with defining the project’s objectives.
He points to McKinsey’s methodology as a best practice to follow, recommending credit unions ask the following questions:
- What will data and analytics be used for?
- How will the insights drive value?
- How will the value be measured?
Plus, because machine learning offers an entirely new way of making predictions, Rock advises credit unions to consider only solutions that can integrate across the breadth of their business processes.
“Use cases can increase one after the other, customer expectations change and grow and fraudsters are busy trying to game the system,” said Rock. “This is why credit unions need a flexible and agile machine learning platform rather than one designed to address a singular use case. An agile platform will provide a multiplier effect on your investment – and unlock the full power of artificial intelligence.”
And to be a best-in-class solution, Rock says, the machine learning platform must be agnostic to the data schema. “Since having more data is so important for making good predictions and handling different types of problems, businesses with machine learning platforms must be able to ingest new data sources seamlessly, without lengthy development cycles,” he said.
Putting Data to Work
While the strength and flexibility of a machine learning platform is essential to its return on investment, even the most innovative solutions are only as effective as the individuals who use them.
“Business processes need to be adapted to incorporate machine learning insights into the day-to-day workflow, and this is done by routing the right data to the right employees – and in a usable format,” said Rock. “But it is also important to educate employees on how to apply data-driven insights to the decisions they make.”
He continued, “Unlike in years past, today credit unions have the best of both worlds. They can now deploy machine learning technologies without the need for large teams of data scientists, developers and engineers – and they can realize gains in security, productivity and member engagement in a matter of weeks – not months or years.”
To learn more about advances in artificial intelligence and machine learning technologies – including CO-OP’s new solution set to go live in 2017 – download our newest white paper, “A New Frontier: Machine Learning, Artificial Intelligence and Big Data,” here.
The original article The Machine Learning Dilemma: So Much Data, Where Do We Begin? can be found on Insight Vault.