Dark social is eating your attribution

Ask a new customer how they heard about you and they'll say "Google." Look at your analytics and it agrees: organic search, branded query, clean conversion. Case closed.

Except the branded query is the end of the story. Nobody wakes up knowing your product's name. Somewhere upstream, in a place your pixel has never been, someone said your name to them. A Slack community. A group chat. A Reddit thread from 2024. A Discord server for people who hate their current vendor. That upstream place is what marketers started calling dark social, and in B2B it is quietly where most decisions actually form.

Why it went dark

Dark social is the shorthand, and the buying committee moved into private and pseudonymous spaces for a simple reason: that's where the honest information is. Vendor websites all claim the same three benefits. Review sites are gamed. So buyers ask peers, and peers answer candidly precisely because no vendor is watching.

The irony is that a lot of "dark" social isn't private at all. Reddit is public. Hacker News is public. X and Bluesky and Threads are public. They're only dark in the sense that your attribution software can't connect a thread read to a signup. The conversation itself is sitting there in plain text.

Stop trying to track it. Start attending it.

The instinct of every marketing team meeting dark social for the first time is to try to measure it: better attribution surveys, self-reported "how did you hear about us" fields, UTM archaeology. Do the survey field, it's cheap and directionally useful. But measurement is the smaller half of the opportunity.

The bigger half: the public part of dark social is a place you can simply show up. When someone asks "what do people use for X" in a public thread, three outcomes are possible. A competitor answers. A stranger answers with whatever they remember. Or you answer, helpfully and with a disclosure, and the thread that shapes the next two hundred lurkers' shortlists now includes you.

That's not attribution. It's presence. And presence in the threads that matter compounds in a way ad spend doesn't, because threads persist. They rank in search. They get cited by AI assistants answering the same question next year (more on that in share of model).

What "attending" looks like in practice

Three habits, roughly in order of effort:

• Listen where your category is discussed, not just where your brand is mentioned. Brand mentions are the trailing indicator. Category questions are the leading one.
• Answer the question first, sell second or never. The lurkers are the real audience, and they punish selling.
• Treat speed as the metric. Public threads have half-lives measured in hours. A response system that surfaces the right thread on Tuesday for a conversation that peaked Sunday is a history lesson.

What to measure when you cannot measure the source

Give up on the source and measure the behaviour instead. Count the threads you showed up in this month. Count the ones where somebody replied to you. Count the times your name appeared in a thread you were not in, which is the only real evidence that the presence compounded. None of those numbers land in an attribution report, and all of them move before revenue does.

The one measurement that does survive is the self reported field at signup. Keep it open ended and read the answers yourself for a quarter. People will type the name of a subreddit, a newsletter, a person. That list is your real channel map, and it usually looks nothing like the paid dashboard. Pair it with social listening on the places that keep showing up and you have a loop that improves without a single new pixel.

Whether you do this with a person on patrol, or with a tool that watches several platforms and pings you when a thread qualifies, is an execution detail. The strategic shift is accepting that a growing share of your funnel will never appear in your dashboard, and deciding to be in the room anyway.