The paper examines non-experimental advertising measurement methods, comparing their effectiveness against randomized controlled trials, highlighting significant inaccuracies in estimating causal ad effects despite rich user-level data.