Not every industry has found their clear and obvious applications of big data analytics. But the clear and present dangers of risk and fraud in financial transactions demand fast predictive modeling. Precisely because we live in the ubiquitous digital era, where most business (and non-business) transactions are rarely (if ever) in analog form and those transactions no longer move at the pace of humans (but at the speed of light), consequently the volume of digital signals as well as lurking dangers is enormous.
Digital signals (from sensors everywhere in our operational business systems) carry transactional information (what happened to what?), as well as metadata (descriptors) and analytics information (data-encoded knowledge and insights). These analytics can be behavioral (providing insights into the interests, intentions, and preferences of the actors in a given transaction) as well as functional (providing insights into the actions or events associated with the transaction).
Behavioral analytics is developing into a major component of digital marketing, as firms seek to sell, cross-sell, and up-sell their products to the right customer at the right time. Behavioral analytics is also critical in risk mitigation of all sorts: financial, cybersecurity, health (individual and population), supply chain, machine performance, and so on.
Here are 10 examples of where fast predictive analytics can play a vital role in most industries (with a focus on financial):
- Predict credit risk and fraud in real-time!
- Use Social Media for deeper understanding (likes and dislikes) of your customers.
- Personalize customer interactions in real-time, across multiple channels.
- Stop improper insurance payments before claims are paid!
- Spot insurance rate evasion tactics during the quote process – before you issue a policy!
- Predict High Health Risk versus Low Health Risk to better manage healthcare decision-making.
- Generate better predictive models of health, car, and home insurance eligibility fraud, underwriting fraud, and improper payments.
- Spot adversarial and anomalous behavior in cyber networks – stop the data breach or illegal funds transfer before it happens!
- Eliminate your Supply Chain hiccups – move the right products to the right locations in the right quantities – and at the right time!
- Make better business decisions regarding merchandising, demand forecasting, and pricing – don’t leave money on the table, or products in your warehouse.
Let us look a little more closely at the financial services industry…
One of the common conditions in traditional financial services (including home, health, and auto insurance) has been the “pay and chase” — i.e., you make the payment to the claimant, and then (after making the payment) you find out that the claim is fraudulent, thus beginning the chase to get your money back.
The new world of predictive modeling and advanced analytics allows for a new mantra in the financial and insurance industries: “Do Not Pay!” — i.e., you do not pay the claim until you have analyzed its likelihood for claim fraud, extraordinary financial risk, or payment anomalies (e.g., duplicate payments).
Predictive analytics modeling delivers a better financial risk posture for your organization than the “pay and chase”. With access to greater and more diverse data sources, it is now possible to develop better models of your customers’ credit risk regardless of the industry. This is certainly true in the financial services industry where there is so much data available: credit scores, credit history, court records, tax records, health records, insurance claims, and more. There is no excuse for not examining as much “public data” as you can in conjunction with other data sources that are available to you internally within your organization. Moderate outlays of your organization’s funds that are incurred in acquiring access to diverse external data sources should be offset by the savings accrued by “not sending your funds out the door” erroneously (either to intentional fraudsters or in unintentional duplicate claims).
An analytics-driven predictive model can predict fraud more efficiently (with fast automatic statistical software packages) and more effectively (with higher precision and higher recall: fewer false positives and false negatives) than traditional business processes. A good predictive analytics model should: (a) detect claims that “smell funny”, (b) prevent the “pay and chase” mode of operations, and (c) stop claims fraud abruptly by empowering a “do not pay” mode of operations. Predictive analytics modeling should aim to satisfy the following business requirements:
- Detect and prevent both opportunistic and professional fraud throughout the claims process.
- Detect underwriting fraud, to prevent premium leakage at the point of sale and renewal.
- Spot rate evasion tactics during the quote process – before you issue a policy.
Many more examples of use cases in the financial services industry (and elsewhere) where fast predictive analytics is important (and what you can do about it) have been expertly enumerated by the fast statistical modeling folks from Soft10, Inc. Check out their fast analytics products (including the Instant Online Overbilling Claims Detector) at http://soft10ware.com/.
Follow Kirk Borne on Twitter @KirkDBorne