Its not a crazy business model. My early stage company gets nearly all of its revenue from a product no one in engineering has touched in over a year, which funds some more intensive and risky product development. Similarly, DuoLingo has a foundation in teaching languages while building up a large supervised data set for machine translation. A lot of companies are basically fronts for getting average folks to either curate a label set or provide permission to access existing data. My old company provided a vague statement like, "Improve your experience, integrate your gmail account", providing quite literally nothing to customers except remove the button, yet provided us unlimited access to contact books and emails/attachments, many of which were very private and confidential. Not proud of that one, but thats the reality and for some people, very profitable.