[kristopolous]

Skincare is an interesting beachhead. How do you test if your results are good? What's the baseline?

I feel like something like movies or video games would be a great way to validate the approach since there's generally agreed upon sentiments regarding these products.

Skin care I'd imagine is fairly complicated. Goal, lifestyle, budget, habit and individual based needs and preferences can lead to different sentiments. How do you calculate say, your loss function?

[qiongzhouh]

I think that's what makes skincare interesting. We want our system to be able to understand your goal, lifestyle, budget, ... and pick out which product is the best for you, given what others who have used the product before said.

With less of this information, the ground truth would probably related to how popular the product is or the average sentiment of people reviewing the product. With this information though, you can compare each one to see which best fits the user's needs. Having compared enough products, you'll eventually figure out which one is the best.

[kristopolous]

it's a recommendation system with a timeline of months and the data to back it up IS your market hypothesis.

you can have whatever risk profile you want, some people are gamblers. Personally I'd like to have greater confidence in the fitness of the implementation before going all in.

Although I say this with the advice adage: unless the business advice comes from someone who is on their own private yacht, it's more opinion than advice. (I've got no yachts)