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 323 forks source link

Two questions - Plotting decomposition over time, and comparing adstock parameter for Robyn vs. lightweight MMM #1007

Open NataliaVConnolly opened 1 week ago

NataliaVConnolly commented 1 week ago

Hi Team,

I just had two questions about Robyn, and wasn't sure where to post them. Hopefully this is the right place.

  1. Is there a way to get the decomposition over time plot like that from lightweight mmm:

    Screenshot 2024-06-23 at 6 12 26 PM

  2. Is it fair to compare the adstock parameter theta from Robyn with lightweight mmm's 'lag_weight' for a geometric adstock model? The reason I am asking is because I am getting very different values from the two packages. It's true that one is for spends (Robyn) and the other, for impressions/clicks (lightweight MMM), but it's still strange just how different they are. Both packages produce seemingly great fits btw.

    Thank you for your help! Natalia

igorskokan commented 6 days ago

Hi Natalia

  1. Check pareto_alldecomp_matrix.csv
  2. Best to post such questions in the Facebook community group https://www.facebook.com/groups/robynmmm to get opinions from the community esp. when comparing with other packages Another idea would to generate simulated data - for example using siMMMulator - https://github.com/facebookexperimental/siMMMulator and compare the results when you know the ground truth value of parameters?

On Sun, Jun 23, 2024 at 11:18 PM Natalia Connolly @.***> wrote:

Hi Team,

I just had two questions about Robyn, and wasn't sure where to post them. Hopefully this is the right place.

  1. Is there a way to get the decomposition over time plot like that from lightweight mmm:

Screenshot.2024-06-23.at.6.12.26.PM.png (view on web) https://github.com/facebookexperimental/Robyn/assets/25134879/da145c30-6f24-4f19-8750-4f1cb665d167

2.

Is it fair to compare the adstock parameter theta from Robyn with lightweight mmm's 'lag_weight' for a geometric adstock model? The reason I am asking is because I am getting very different values from the two packages. It's true that one is for spends (Robyn) and the other, for impressions/clicks (lightweight MMM), but it's still strange just how different they are. Both packages produce seemingly great fits btw.

Thank you for your help! Natalia

— Reply to this email directly, view it on GitHub https://github.com/facebookexperimental/Robyn/issues/1007, or unsubscribe https://github.com/notifications/unsubscribe-auth/ARF52KGMGWYIQGZYHCJK2BDZI5CTTAVCNFSM6AAAAABJYXLO26VHI2DSMVQWIX3LMV43ASLTON2WKOZSGM3DQOJRGM2TENI . You are receiving this because you are subscribed to this thread.Message ID: @.***>

laresbernardo commented 3 days ago

Re 1, check Robyn::decomp_plot() too.