Hetal Gohil
Every few months a client asks us some version of the same question: “which channel should we cut?” Usually they've been looking at last-click attribution in their analytics dashboard, and one or two channels look like they're barely contributing anything. Before we answer, we ask them to hold off — because last-click is often measuring the wrong thing entirely.
Here's the pattern we see constantly: a prospect discovers a brand through a social post, doesn't click, sees a retargeting ad two days later, doesn't click that either, then searches the brand name directly a week later and converts through organic or paid search. Last-click attribution gives 100% of the credit to that final search. The social post and the retargeting ad, which arguably did the actual work of building awareness and consideration, get nothing.
What we look at instead
We're not precious about a single “correct” attribution model — anyone who tells you they've solved attribution perfectly is usually overselling. What we do instead is look at assisted conversions and cross-channel paths, not just the final touch, and we pair that with a blunter but very useful test: turn a channel off for two to four weeks and watch what happens to direct and branded search traffic. If branded search volume drops when you pause social spend, that's a real signal the two are connected, even if last-click never showed it.
We also push clients toward tracking a fuller conversion path where the tooling allows it — first touch, last touch, and the touches in between — rather than collapsing everything into one number. It's more work to read, but it stops the classic mistake of quietly starving the channels that build demand in favor of the channels that just harvest it.
Where this actually changes decisions
The clearest example was a client running both paid search and social awareness campaigns. Paid search looked like the clear winner on last-click ROAS. When we paused the social spend for a month as a test, paid search conversion rates dropped and cost-per-acquisition on search climbed noticeably. The social spend hadn't been showing up as a converting channel — it had been doing the work of making search converters more likely to convert once they searched.
That doesn't mean every underperforming channel is secretly carrying the team. Sometimes a channel really is just underperforming, and the honest answer is to cut it. But we've stopped making that call off a single attribution model's dashboard, and we'd encourage anyone managing their own budget to do the same before reflexively cutting the channel that looks worst in last-click.
Building a rougher, more honest model
In practice, we build clients a simplified view rather than a perfect one: a weekly table tracking spend, assisted conversions, and last-click conversions side by side for each channel, alongside branded search volume as a rough proxy for overall demand generation. It's not a data science model, and we don't pretend it is. But laid out side by side over a few months, patterns show up that a single-touch dashboard hides completely — a channel with weak last-click numbers but a consistent lift in assisted conversions is telling you something useful, even if it never gets the final credit.
The other habit worth adopting is testing changes in isolation where the budget allows it. Pausing one channel for a defined window and watching what happens elsewhere is a blunt instrument, but it's a lot more honest than any attribution model, because it shows you what actually happens in the real world rather than what a weighting formula assumes should happen. We run this kind of test with clients at least once a quarter, specifically because attribution models drift as campaigns and audiences change, and last quarter's conclusion isn't guaranteed to hold this quarter.
Getting buy-in for a messier but more honest model
The hardest part of this work often isn't technical — it's convincing a client's leadership team to accept a messier, less definitive answer than the clean single-number dashboard they're used to. A last-click report gives a satisfying illusion of certainty; a multi-touch view showing three channels all contributing to a conversion path is honest but harder to present in a quarterly review. We've found the framing that lands best is comparing it to a sales team, where crediting only the person who closed the deal and ignoring everyone who built the relationship beforehand would obviously be unfair — marketing channels work the same way, they just don't feel as intuitive.
Once a client accepts that some ambiguity is inherent to the problem rather than a flaw in the reporting, the conversation shifts from “which channel gets the credit” to “which combination of channels is producing the best overall outcome” — a genuinely more useful question, even though it doesn't reduce as cleanly to a single number on a dashboard.
