The challenge
Everyone could see the traffic. No one could see the revenue behind it. Organic and AI search sat upstream of the pipeline, with broken first-touch tracking and inconsistent UTMs in between, so their contribution disappeared before it reached a dashboard. A channel you can't attribute is a channel you can't defend at budget time.
The diagnosis
First touch was overwritten.
It was stored per-origin in the browser, so when a visitor moved between the company's own subdomains, the internal hop replaced the real acquisition source.
Organic had no label.
Organic search carries no UTM, so in Amplitude ~99.9% of organic sessions read as "(none)" and ~98% of first-touch mediums were empty.
The conversion looked internal.
The referrer on the conversion event was the last internal page, not where the buyer came from.
Nothing derived a channel.
There was no Organic Search, and no AI Assistant, to segment by.
What I built
Persistent first touch.
First touch is now captured once, frozen, and stored domain-wide, so hops between subdomains no longer erase it. Internal traffic between the company's own domains gets its own label instead of posing as acquisition.
GA4-style channels in Amplitude, including AI.
Each session is classified from its referrer and UTMs into source, medium, and channel, and every event carries those properties. Organic Search and AI Assistant (ChatGPT, Perplexity, Gemini, Claude, Copilot) are now real channels you can segment conversions by.
One taxonomy.
A shared naming standard for sources, mediums, and UTMs, so every team speaks the same data language.
Verified, then iterated.
I tested each fix against new-user cohorts before and after release, logged what worked and what didn't, and kept fixing the gaps.
Executive dashboards.
One view from traffic to pipeline, so the conversation moved from "how much traffic?" to "how much pipeline, and at what efficiency?"
Swipe sideways to inspect the full diagram →
The signature insight
The clearest signal came from a number most teams would read as bad news. Search impressions grew +79% year over year while clicks stayed flat, and AI referral traffic rose +76%. That's the zero-click shift: buyers get their answer inside the AI result instead of clicking through. Rather than fight it, I measured it, judging search by citations and pipeline instead of clicks. That turned a threat into a strategy.
The results
First-touch pipeline sourced by organic and AI search in H1 2026.
The company's first closed-won deal attributed to AI search. Proof the channel closes, not just clicks.
Blog-to-contact conversion versus the site average on a single product line (MFT), the model now being repeated.
AI referral traffic growth, now tracked as its own channel.
What carries forward
Attribution breaks in boring places: storage scope, missing labels, internal referrers. Find the root cause, fix it at the tag level, verify it against real cohorts, and give leaders one view from discovery to revenue. That's how a hard-to-measure channel becomes one you can defend.