Forecasting Demand of a New Product when there's no previous Sales Data?

Here, we try and develop a machine learning based solution for the problem stated above.

Let's say Ralph Lauren plans on releasing 100 new styles, and they want to forecast the demand for each? How should they do it?

1. What kind of data should they collect?
2. How would you predict the demand of a new style?
3. How to gain credibility?
4. How to tackle explainability of these ML models?
5. How would you evaluate the performance of these models?

All of these questions are answered in the PDF file which contains the solution reached upon employing first principles thinking.

Happy Reading!