I don’t understand why this is a good feature to predict revenue. Won’t the total revenue for the company just be the sum of revenue for each product? i.e w will just be an vector of ones?
I guess you are thinking in the direction of how we can predict the revenue from some indirect factors such as the shop’s location, opening hours, size and so on, because in this manner, the model can be better generalized to many more shops without knowing the details of orders in each shop. If my guess is right, I think such model will be very useful for evaluating whether I should start a business there.
However, the purpose of the question is just to check how you would engineer the feature and your understanding is perfect, if the revenue is solely from selling products, then I agree that w will be a vector of ones. If there is additional revenue that is not just from selling the products, such as the shop owner will earn extra commissions each time the shop sells some brand A’s product (e.g. 3 packs of milk and 2 packs of cheese) as a rewards to the shop for promoting A’s product, then the vector can be different.
That’s a nice observation, and thank you for sharing that, Huzaifa.
Oh right, I did not consider the fact that the revenue could be from not just product sales. Thanks for the insight Raymond.
No problem! You are welcome Huzaifa!