The One With The Most Data Wins
0 comment(s)
Earlier posts on Bank Fraud Forum have discussed the importance of data in robust, accurate fraud management solutions. Recently I came across two great articles that I think make a strong case for why data is critical and how it influences the analytics used to find fraud.
Data exists in two dimensions. It can either be deep or broad. Let's look at the difference:
Deep data holds a long history of siloed information. In the image above, imagine 2 years worth of data about check transactions. Contrast this with broad data.
Broad data can also be deep but combines multiple information silos to create a much richer, context aware data set. As shown in the image above imagine combining check transaction data with online, ATM data and branch data.
What this means for analytics is crucially important. On his popular blog, Anand Rajaraman makes the statement, "More data usually beats better algorithmsâ€. Why? Even the most complex analytics can't divine fraud from a limited data feed. However, by combining multiple data sources, simple analytics quickly zero in on suspicious behavior.
The popular payment processor PayPal has embraced this concept. A recent article details how PayPal has been combing the internet for data on current and potential customers. PayPal uses extremely advanced analytics but with only 3 seconds to decide to accept or block a sale, PayPal recognizes that every possible data point must be considered. Scott Thompson, the President of PayPal sums it up well. "To do risk management well," Mr. Thompson said, "you have to have more data than anyone else."
In evaluating fraud management solutions the implications are obvious. Analytics are no doubt important but the first question should be "how are you going to take full advantage of all the data I have at my financial institution?". While it's fun to talk about neural nets, artificial intelligence and other complicated analytics, it's often more data that wins the day.