Predictive analytics and consumer scoring: How companies use AI, machine learning & big data to create predictive models
Advances in artificial intelligence (AI) and predictive analytics are using consumer scores to automate business decisions to predict things like risk and fraud. But concerns over fairness mean companies need to make scores transparent to consumers.
What is predictive analytics?
Predictive analytics utilizes machine learning and advanced statistical techniques to analyze consumer behavior and make predictions about future actions. Predictive analytics can support applications like scoring risk and preventing fraud, and provide insight into consumer behaviors like lifetime customer value and even affective states, like feelings toward a specific experience.
What is consumer scoring?
Consumer scoring summarizes relevant information about consumers based on past behaviors. It’s influenced by the data collected from the web, mobile, and IoT devices that detail demographics, geographic information, and transaction history. Consumer scoring leverages complex data analytics to evaluate and apply metrics to consumers for automated business decisions.
Predictive analytics & modeling tools
Companies no longer have to employ a statistician or data scientist to use predictive analytics. More vendors are making predictive analytics and machine learning modeling accessible to business users across the organization with platform offerings that feature user-friendly drag-and-drop interfaces and graphical explanations.
These tools address a significant skills gap in AI and data science, but also have the potential to obscure essential auditing functions in a seamless process. Having machine learning and AI run real-time regression and decision tree analysis on big data helps to efficiently develop ‘scores’ for people based on specific goals.
Predictive analytics techniques & consumer scoring examples
With more data, advanced analytics, and machine learning, predictive analytics, and consumer scoring are finding new applications in a variety of business cases across industries. The Predictive Analytics and Consumer Scoring report highlights and breaks down many of these applications, two of which include:
Most US adults are not comfortable with companies using AI to access their personal data.
Fraud, trust, and risk scoring tools identify legitimate users and transactions, detect bots, and automate decisions about transaction safety. Many vendors use machine learning to train models for detecting “normal” and fraudulent behaviors on their clients’ sites. Applications include new account creation, login verification, and e-commerce transaction approval.
Alternative credit scoring uses alternative data—sources of data that weren’t included in traditional credit and loan transaction histories, like rental and mobile phone payments and bank account transactions—to build a more complete view of risk for the underbanked.
Other inputs might include behavioral data like shopping habits, web and social media usage, behaviors, and even device-usage gestures. These inputs aren’t regulated as traditional credit reporting inputs. Alternative credit scoring is practiced by both new disrupting players and incumbent credit reporting agencies.