One of the best parts about Featuretools is that it’s not constrained by human limits on creativity and time. There’s only so many features we would ever think to build and only so much time we can manually build them in. You’re also right that everyone thinks their ideas are best which can make a establishing a reproducible feature engineering pipeline challenging.

That’s not to say that people don’t think of great features, but only that we often miss out on some potentially useful features because we are limited in what we consider important. One of the benefits of Featuretools is that it can build more features on top of existing manually defined ones to extract potentially more useful information from the dataset.

I’d definitely recommend checking out Featuretools. There are quite a few demos on GitHub and once you learn the technique, you can apply it to any problem! Let us know if you have any problems or if you apply Featuretools successfully to a problem.