What to do about Non-Media Marketing Channels?

[nd-mmm]

I have built a geo level MMM for a sports betting client, however I've ran into an issue. I want to understand the impact of non media marketing channels; two product promotions. I am also using media channel spend share to inform my saturation curves however since I have no spend associated with my product promotion variables mmm.fit won't run since it's trying to apply the spend share list to those channels which are not included. The only work around I could create was to take these two channels out of channel_columns and put them in control_columns but then I can't see the individual impact of these channels? Any suggestion what I might do?

[timbo112711]

@nd-mmm If I am understanding correctly here, you have your media channels with spend associated to them. These media channels are getting added into the MMM via the channel_columns arg. You have two other variables, product A promo & product B promo. What do these two promo variables represent? What are the raw values, units sold on promo?

You'd want to have these added into the model to see their individual impact on your response (sales, etc). Here is what I would recommend doing. Take these two variables, product A promo & product B promo, and add them to the control_columns arg. By doing this we can have two things:

1. Not force these promo variable to go through the adstock and saturation transformations.
2. Have a more accurate model structure. Since media and promo variables are very different.

From this, you'll still be able to see the channel level impacts as well as the individual impact of these promo variables. You can see this by plotting out the waterfall chart via mmm.plot_waterfall_components_decomposition();