[inputcoffee]

If you could find someone offshore that is smart enough to improve a predictive algorithm by even 0.01%, you should hire them as well.

Consider that the holy grail is finding out if customers actually want a product.

Break that product down into features, and run a regression. Historically, how well does that do?

But these are new products. They have new features. How do you solve that?

In a sense this is "the" problem. Anyone who is genuinely additive would be fought over by any company that produces a new product.