AI And Alternative Data
Join The Discussion | Ai4 Finance 2019
Within financial services, the term Alternative Data refers to data that is collected through non-conventional means in order to obtain some investment insight.
Examples include IoT sensor data, satellite images, mobile device data, or really any form of data that can be obtained and is not traditionally analyzed for purposes of investment or lending.
While more and more “alternative” data is being generated daily, only those groups with the resources to access and analyze that data will be able to benefit. In this way, many financial firms looking to gain a competitive advantage have successfully developed artificial intelligence and machine learning technology capable of analyzing these new datasets to help augment investment decisions.
Here are some examples of how artificial intelligence and machine learning technology can be used in conjunction with alternative data to inform investment & lending decisions:
1. Counting Cars With Satellite Imagery: An investment manager will purchase large datasets containing satellite photos from the past 3 months. Once they have the dataset, they will use computer vision AI to automatically count the number of cars in all the parking lots of a particular retailer, let’s say Walmart. Once they know the car data from the past three months, they can estimate the number of customers and predict revenues from the last quarter before Walmart reports their quarterly figures. Needless to say, an investor can use this revenue prediction to decide whether or not to buy Walmart shares before the rest of the market who is relying on more traditional data sources.
At Ai4 Finance, Orbital Insight, an alternative data company performing geospatial analytics, will moderate a panel discussion entitled “Alternative Data’s Impact On Investing and Beyond.”
2. Visitor Traffic Through Smart Phone Data: An investment manager can purchase foot traffic that is sourced from millions of GPS, WiFi, and Bluetooth signals in order to know exactly how many people visited a particular location. You could imagine using this type of data to count how many visitors a network of malls received over the past quarter in order to predict revenues.
3. Determining Credit Worthiness Through Online Activity: When a consumer seeks a bank loan, the bank begins the underwriting process. A common data point is the borrower’s credit score. However, many people do not have a solid credit history. Approximately 1.7 billion globally don’t hold an account with a bank or credit card company. In order to lend to one of these 1.7 billion people, a lender can use alternative data points such as the borrower’s online purchase history, social media posts, and browsing activity. Lenders have built AI models that can predict whether or not a burrower with certain online activity characteristics will repay their loan.