Last-click attribution has been making decisions on our behalf for as long as I’ve worked in this, and mostly making them badly. The way most analytics have counted a sale is simple and wrong: whichever channel the visitor arrived from immediately before buying gets all the credit. Someone buys after a Google search for your brand name, and search takes the medal, even though the customer first found you through a display ad, then a blog post, then an email, and only searched your name because they already knew it.
Google added Multi-Channel Funnels to Analytics last year, and I’ve finally had the time this autumn to sit with the reports properly. They’re worth the effort. The Assisted Conversions report shows which channels turned up earlier in the journey and helped, even though they didn’t get the last click. The Top Conversion Paths report lays out the actual sequences, and confronts you with the fact that a fair chunk of your sales involved three or four touches across a fortnight, not the single visit last-click implies.
What this tends to reveal, and it’s an awkward conversation to have with a client, is that the channels people want to cut are often the ones doing the early work. Display and social rarely win on last-click; they make the introduction, they don’t close. Paid brand search, meanwhile, looks like a hero because it mops up demand that other channels created. Judge purely on last-click and you’ll defund the top of your funnel and pour money into the channel that was only finishing a job the others started.
None of the attribution models on offer are correct, and I want to be careful here. First-click just moves the bias to the other end and over-credits the introduction; the linear model spreads credit evenly whether or not that reflects anything real; time-decay bakes in assumptions of its own. So I don’t treat any of them as the truth. What they’re good for is comparison: run the same sales through two or three of them, watch how differently the credit falls out, and that alone is enough to shake you off the single dumb number you’ve been running the budget on.
I’m not about to rebuild anyone’s budget on the strength of a fortnight in these reports. But I’ve started putting the Assisted Conversions figures next to the last-click ones in the monthly report, side by side, and the conversation about where the money should go has already got a good deal more interesting.
