Use cases
One class of problem, four faces of it: the exposure happened somewhere tracking cannot reach, so every one of these people lands in the same bucket as someone typing your domain from memory. Asking them is the only signal that survives all four.
“TikTok” is not an answer — six ambassador accounts collapse into one string and every conclusion drawn from it is noise. Each walkthrough below covers the second question that gets you to which account, which room, which show, and where to place the form so the answer sits next to revenue.
AI assistants
No referrer leaves an assistant, and most of the time there is no click to carry one: the person reads your name in an answer, searches it, and lands on you from Google. So the option has to say “ChatGPT” — not “AI assistant” — and sit beside Google rather than under it. The follow-up can ask what they were asking about, which gets you a real question from someone who then paid.
Read the walkthrough →Communities and word of mouth
Discord, Slack groups, subreddits, a friend forwarding a message. The exposure happens inside somebody else’s room, and by the time the link is clicked it has been re-pasted twice and stripped of everything. The follow-up gets you which room — “a community” is not something you can go and do more of.
Read the walkthrough →Launch day
Product Hunt, Hacker News, X, three newsletters, a dozen DMs. In-app browsers and forwarded links drop the referrer, so the traffic that decides whether the launch worked is exactly the traffic you cannot attribute. Ask at signup and you learn which post; ask again at payment and you learn which post sent people who pay.
Read the walkthrough →Podcasts and events
A sponsored episode, a conference talk, a booth conversation. Nobody clicks anything — they hear a name in a car and type it in a week later. The follow-up asks which show or which event, because “a podcast” is not a media-buying decision.
Read the walkthrough →Not on this list
Channels with their own reporting console — LinkedIn Ads, Google Ads — already have ground truth, and are not what this measures. Their role here is the reverse: a calibration anchor you can compare self-report against, to see how much self-report under-counts.
View this page as markdown — for agent context / LLM readers.