The Ecomm Analyst

I had a client last year whose Meta ads looked like they were dying. Last-click revenue was down 40% year over year. The CMO was ready to cut spend in half and reroute it to email.

I pulled the same window in ThoughtMetric, set it to a blended view weighted toward incrementality, and Meta came out as the second-largest revenue driver in the business. Email was getting credit for sending the closing message, but Meta was doing the heavy lifting upstream.

We didn’t cut Meta. We cut the email send frequency, which was eating into margin, and held Meta steady. Revenue grew the next quarter.

This is the problem with last-click. It tells you who closed the deal. It doesn’t tell you who set it up. And in a world where customers see your brand five, ten, fifteen times before buying, that distinction matters enormously.

The case I make to clients now goes like this. Last-click is a useful operational metric. It tells you which message actually pushed someone to checkout, which is genuinely useful for things like email subject line testing or landing page optimization. But it is a terrible strategic metric. It will systematically over-credit anything that touches the customer late (email, retargeting, branded search) and under-credit anything that touches them early (paid social, display, podcasts, influencers).

If you build your media plan around last-click, you will gradually starve the top of your funnel until you wake up one day with no new customers and a great-looking ROAS report.

What I look at instead is a blend. Last-click for the operational stuff. A multi-touch view for understanding the customer journey. And a periodic incrementality test (geo holdouts work well for this if you have the volume) to keep everyone honest.

None of this is new. Marketers have been complaining about last-click since at least 2015. What’s changed is that the tools to look at it differently are finally usable by humans who don’t have a data science team.

The shift I’d encourage anyone running a Shopify store to make is mental, not technical. Stop asking “which channel made this sale.” Start asking “which channels does this customer touch on the way to a sale, and what would happen if I removed one of them.” That second question is harder. It’s also the only one that matters.