Attribution looks weird

[charlesdethibault]

Hello, does anyone have an idea about why the attribution looks like this? I have not included seasonal features, just weekly spending by program and sales number. Thank you so much for your help!

[charlesdethibault]

So I found out two things. First, the data is getting better if we add monthly seasonality—the second problem. I have the paid search causality bias that is explained in this Google paper: https://research.google/pubs/pub46861/ . There is a big surge of demand for Christmas which increases both paid search and revenue and the model, which relies on correlation thinks that an increase in Paid Search spending leads to an increase in revenue while the true confounder is Christmas. Has anybody talked about the problem with the Leigthweigth model already?