Myth: last-click attribution is good enough to judge a native channel.

Myth: last-click attribution is good enough to judge a native channel.
No. A retailer's last-click model gave native a cost-per-acquisition of $18 — middling, nearly cut from the plan. Before pulling it they ran a national geo holdout.

With native fully dark in test markets, total sales — across all channels — fell 11% versus control. The true incremental cost-per-acquisition was $9, half what last-click claimed. Native was seeding demand that search and direct were closing and getting credit for.

The channel they almost killed was the cheapest real driver they had.

Last-click rewards the closer and starves the opener. If you prune by last-click alone, you'll keep cutting the channels that fill the funnel and wonder why the funnel keeps emptying.