The tools
Meta Robyn or a simple regression in Python. Data inputs: weekly spend by channel, weekly revenue, external regressors (holidays, promos).
What you'll learn
Saturation curves per channel and diminishing returns thresholds. Answers 'how much can I scale Meta before diminishing?' concretely.
Update quarterly
Run the model quarterly. Monthly is too noisy; annually is too slow. Quarterly catches meaningful shifts in channel efficiency.