We often hear that a company's LinkedIn page isn't doing anything. A post goes out, it gets a dozen likes, mostly from the team, and it looks as if nobody in the market is reading. So teams either post more or stop altogether.
Some of the market is reading. It's happening in private, where the analytics can't see it. You can measure a good share of it by asking buyers directly where they came across you, and by recording that answer separately from where the conversation started.
What is dark social, and why can't you see it?
Dark social is the research and sharing that happens outside the analytics you control. A Head of Operations reads four of your posts over a month and never reacts to one. Someone forwards your case study to their finance director in a direct message. A peer asks in a private group who they should talk to, and your name comes up. None of it registers as engagement, but it can decide who gets the meeting.
Alexis Madrigal named it in The Atlantic in October 2012, describing traffic that analytics recorded as "direct" because the link had been passed on by email or instant message. The channels are different now and the blind spot is still there. Our dark social explainer covers the wider picture and what it means for demand generation. This article is about one practical thing: getting buyers to tell you.
Two records: where it started, and what influenced it
The single most useful habit is to keep origination and influence as two separate records. Origination is where the identifiable conversation began: a LinkedIn message, a cold call, an email reply, an inbound form. Influence is anything the buyer confirms shaped their decision to talk, such as a post they read or a case study a colleague passed on.
Most CRMs have one "source" field, so teams pick one answer and lose the other. If outbound booked the meeting, the source says outbound and the post the buyer read three weeks earlier disappears. If someone decides the post deserves credit, the source gets changed and now outbound looks weaker than it is. Either way, one channel ends up with credit for the other's work.
We see this most clearly in outbound. When we message a prospect for a client, they'll often look at the company page before they reply. Some mention on the call that they'd already read a post, or that a colleague sent them one. That page visit never appears in a report as the reason the meeting happened. So in our reporting the meeting is credited to the outreach that started it, any influence the prospect confirms is logged next to it, and anything we can't see goes down as unknown. We tell the client that's what the unknown column means.
Log what you can't see as unknown, and say so.
ORRJO reporting rule
The "unknown" column matters more than it looks. It stops anyone, including us, from filling the gap with the answer they'd prefer.
How to ask "how did you hear about us?" so the answer is usable
Ask in two places, in open text, and record the buyer's own words. Most of the value is in the details, so run through this checklist:
- Open text on every inbound form. "How did you first hear about us?" with a hint such as "a person, a post, a podcast, a search: anything helps". A dropdown only offers the channels you expected.
- Use "first". Without it, people tend to name the last thing they did, which is usually "Google" or "your website".
- Ask again on the first call. "What made you take this meeting?" works for outbound replies too, where no form was ever filled in. If the answer is vague, follow up with "where did you first come across us?"
- A dedicated CRM field for each. One for origination, one for confirmed influence, both in the buyer's words. Don't overwrite either.
- Don't lead. "Did you see our LinkedIn posts?" gets a polite yes. Ask the open question and wait.
- Code it monthly. Group answers into a short list of categories: a post or a named person, a colleague, a community or newsletter, a podcast or event, search, an AI assistant, other. Keep "doesn't remember" as its own category.
The playbook has the category worksheet and the first-call script ready to print, so there's no need to build them from scratch.
Which engagement is worth watching?
Check who engaged before you count how many. Ten likes from people at accounts you're trying to reach tell you more than two hundred from other marketers, because the first group can buy from you and the second mostly can't. Go through the names on your best and worst posts each month and mark the ones that match your target list.
The same goes for followers and profile visitors. A small rise in followers who fit your buyer profile is a better sign than a large rise from peers in your own industry. If your page reach goes up but none of the new names are people you sell to, the content has found an audience outside your market. Our piece on demand gen vs lead gen goes further into how the two kinds of work feed each other.
When self-reported attribution misleads you
Self-reported answers are memories, and memories are unreliable. Buyers compress months into one moment, credit whatever they saw last, and sometimes give the answer they think you want. A single answer tells you very little. A quarter of answers, compared with what your tracking says, tells you a lot more.
Small numbers are the other trap. If you have 10 first meetings a quarter, two buyers mentioning a podcast is an anecdote, and moving budget on it would be a mistake. Read a full quarter, look for the same kind of answer repeating, and treat a single vivid response as a lead to follow up rather than a finding.
Dark social also has limits as an explanation. It explains gaps in attribution, and it doesn't excuse a page that never leads to a conversation. If buyers aren't naming your content, and you aren't seeing replies or meetings from the accounts you write for, the problem is probably the content or the targeting. Asking the question will show you that too, which is part of why it's worth asking. If you'd rather run the whole thing with help, it sits inside our Demand Generation service, which we sell, so bear that in mind.
Frequently asked questions
Should "how did you hear about us?" be a dropdown or open text?
Open text. A dropdown only offers the channels you already expected, so it can't tell you about the podcast, the group chat or the colleague you never knew about. Group the free-text answers into categories once a month. It takes an hour, and the exact wording buyers use is often more useful than the category.
Is self-reported attribution accurate?
Not for any single buyer. People forget, compress months into one memory and name the last thing they saw. It becomes useful in aggregate, over a quarter or more, and when you compare it with what your tracking says. Treat each answer as evidence and the monthly pattern as the finding.
Where should we ask the question?
In two places: an open-text field on every inbound form, and out loud on the first call, as "what made you take this meeting?" The form catches inbound. The call catches the buyers who replied to outbound, which is where a lot of dark social influence turns up.
How many answers do we need before we trust the pattern?
More than most small teams get in a month. With 20 first meetings a quarter, two or three answers can swing a category, so read a full quarter before moving budget, and look for the same kind of answer repeating rather than one striking response.
Sources
- Alexis Madrigal, Dark Social: We Have the Whole History of the Web Wrong, The Atlantic, October 2012 (excerpt via Benton Institute) (Origin of the term; dark social shows up as direct or typed traffic.