Last-click attribution rewards the channel that opened the door, not the one that brought the customer.
Someone sees your ad on Instagram, reads an article a week later, then searches for your brand name and buys. With last click, Google gets all the credit and Instagram looks worthless. Cut back on that, and you'll see the rest drop too.
No model gives you the truth. They divide the same revenue according to different rules, and nobody knows which split is right. Anyone who presents a model as fact is selling certainty that isn't there.
Switching a channel off for a period and seeing what happens to your total revenue. It's crude and costs you for a while, but it's the only way to separate cause from correlation.
A question or project on this topic?