Making predictions using credit card data
The biggest surprise to me in this article in Sunday’s Times, about how lenders use meticulous information about purchasing patterns to forecast default, is that the article treats this as a surprise. Data on what we do is valuable to firms, and we should all expect that much of that data is being used for individual-level forecasting.
But I was also interested that I could find only this one source online (unrelated to the Times article) that discusses what Martin did at Canadian Tire, and it and the Times article are comparably vague. I’d be curious for other sources and technical details.
