I'm going to say something that will make half of you uncomfortable: session-level attribution is essentially dead in 2026, and no tool is going to bring it back. Not Triple Whale, not HockeyStack, not whatever you saw on LinkedIn yesterday. These tools are useful — we use a couple of them — but they don't solve the underlying problem, which is that user journeys are now genuinely invisible.
iOS 14 started it in 2021. Chrome finally deprecated third-party cookies in 2025. AI overviews are eating clicks. Half of your buyers are researching in ChatGPT and Claude and you'll never see those sessions. A customer discovers you on TikTok, mentions you to a colleague in Slack, gets an email weeks later, types your URL directly, and converts. Your attribution model credits "Direct / None."
So what do we do? Here's the honest version.
Accept that you have three imperfect lenses
- Platform-reported metrics (what Meta, Google, LinkedIn say they drove). Directional, biased toward themselves. Fine for within-channel optimization.
- Server-side CRM attribution (first-touch and last-touch UTMs stored on the lead record). Our working source of truth for channel mix.
- Incrementality (pause tests, geo-holdouts, MMM). The only thing that answers "did this channel cause revenue" — but slow and expensive to run.
None of these is true on its own. Together they triangulate.
Pause tests are underrated
The cheapest honest attribution test is: turn a channel off for two weeks and see if revenue drops. Everyone finds reasons not to do this. Almost everyone who does it is surprised by the result in at least one direction.
We ran a pause test last year on a client's branded Google Ads spend — $32K/month. Conventional wisdom: branded search is always incremental, always keep it on. Our test: 78% of that spend was picking up traffic that would have converted anyway through organic. We reallocated $25K/month and nothing bad happened.
Run one pause test per quarter on your biggest channel. Worst case you learn it's incremental and sleep better. Best case you find the money you're leaving on the table.
What we've stopped telling clients
We no longer claim a specific ROAS on paid social with two decimal places. We give a range. We explain that the low end is probably closer to true. This goes over badly in sales calls and well in quarterly reviews.
We also stopped building multi-touch attribution dashboards for most clients. They created a false sense of precision. A good marketing dashboard in 2026 has: total revenue, blended CAC, channel-level first/last touch from CRM, and a running list of what incrementality tests are planned or in-flight. That's it.
The bottom line
Stop trying to prove exactly which ad drove which dollar. You can't, and the people selling you the dream can't either. Focus on: is total revenue growing, is blended CAC stable or improving, and do we have a good story to tell the board. That story doesn't need to be precise to be honest.