Did this marketing actually cause the sale? Or would the customer have bought anyway?
Incrementality asks the counterfactual question: if we hadn't run this campaign, would these sales still have happened? Because a lot of what attribution calls 'conversions' would've happened anyway. Retargeting ads 'attributed' to a sale that was already going to close are a classic case. You're just getting credit for inevitability.
Incrementality is the honest question. Attribution often credits channels that aren't actually driving anything. Geo-holdout tests, randomized experiments, and MMM are ways to measure it. The truth is usually: you can cut 20-40% of ad spend without losing revenue.