Don't worry about the technology risk. People already do this, so it isn't impossible. They probably don't do it well, but you only have to do better than rando

Don't worry about the technology risk. People already do this, so it isn't impossible. They probably don't do it well, but you only have to do better than random to provide benefit. Market risk is more of a concern to me. If you're presenting results directly to users they'll probably care greatly about the quality of your algorithm. If you're just using aggregated data to drive, for example, marketing campaigns then the noise will be washed out given enough data.