If I were you, I'd start by running a simple incrementality test, like a geo-holdout or even a short-term blackout for a single channel. It is the quickest way to get a clean, undeniable baseline of a channel's true value before you start layering on complex software and models.
Here’s why I'd start there. So much of the attribution puzzle is just noise. As Constantine Yurevich said on the DTC Podcast, traditional attribution is basically dead. Trying to perfectly track every single touchpoint is a fool's errand. A simple on/off test cuts through the noise and gives you a clear signal about lift. You're not asking
