search

facebookexperimental / Robyn

Robyn is an experimental, AI/ML-powered and open sourced Marketing Mix Modeling (MMM) package from Meta Marketing Science. Our mission is to democratise modeling knowledge, inspire the industry through innovation, reduce human bias in the modeling process & build a strong open source marketing science community.
https://facebookexperimental.github.io/Robyn/
MIT License
1.07k stars 322 forks source link

Allocator Error #955

Open AlessioPallotti opened 2 months ago

AlessioPallotti commented 2 months ago

Hello,

I have a problem regarding the allocator. Can you tell me why I have a higher "total response increase" value if I set a variation of +-50% for each variable than if I put +-100% as variation limits? Conceptually I should expect a greater or at most equal value in the case where I have +-100% variation. Also, there are some special cases where in the output of the allocator sometimes it does not print me the results for all the variables but only for some of them.

Thank you very much!

gufengzhou commented 2 months ago

Hi, it's quite difficult for me to debug "blindly". Can you share some visuals? In general, you're right that more freedom means higher response. I can only think of one exception: when you have S-shape curves, and allow -100% for lower constraints, the algo might prefer a lower than a higher spend point. The reason is that every S-curve will always have 2 symmetrical points that have the same slope, located on the lower and upper part of the curve. That's why -100% might not be the most practical.