A statistical model that relates your historical spend by channel to your revenue, estimating each channel's contribution from the top down - no pixels, no user tracking, so privacy changes cannot break it. MMM is the budget-allocation layer of a mature measurement stack: platform attribution makes the daily in-channel calls, incrementality tests establish ground truth, and MMM splits the budget across channels. The mistake is treating its output as gospel on its own - MMM and platform attribution will disagree, often badly, and without incrementality tests to calibrate the model you are just choosing which fiction to believe.
Benchmark. Possible from roughly $5M revenue with a simple channel mix; standard from $20M+ revenue and $3-5M+ media spend, 3-5+ channels and 18-24 months of clean weekly data. Reallocating spend on MMM output typically lifts media-driven revenue by mid-single-digit percentages.