Community & Dark Social Playbook

Many B2B buying conversations now happen where analytics can’t see them: in Slack and Discord groups, private messages, forwarded emails, subreddits and AI assistants. You can’t track them, but you can be present in them as a person, make your thinking easy to forward, and measure the result by asking buyers where they heard of you.

A founder shows me a dashboard where “direct” is the biggest traffic source and asks why paid search is the only channel that seems to work. The answer is usually the same: the best channel is working, and the dashboard can’t see it. Someone read a post, pasted the link into a Slack channel, three people opened it on their phones, and one of them booked a call a week later. The analytics recorded a stranger arriving from nowhere.

Alexis Madrigal coined the term “dark social” in The Atlantic in October 2012 for sharing that carries no referrer^[1]: a link pasted into a message, an email or a chat arrives at your site with no record of where it came from, so analytics files it under “direct”. In B2B today it covers many of the places a buying conversation happens: Slack and Teams channels, WhatsApp groups, private communities, podcast mentions, screenshots of your post, and questions typed into an AI assistant. None of them leave a trail you can follow back.

Much of the decision is made there, before a buyer talks to you. In Forrester’s 2025 Buyers’ Journey Survey, 68% of B2B buyers already had a front-runner vendor in mind at the start of a purchase^[2]. In Gartner’s survey of 646 B2B buyers, fielded in August and September 2025, 67% said they prefer a buying experience with no sales rep, and 45% had used AI during a recent purchase^[3]. And in a survey of 1,202 US business decision-makers that SurveyMonkey ran with Reddit, peer recommendations were the most trusted source of information, at 73%^[4]. The shortlist is built in conversations you aren’t part of. Your job is to be on it. The dark social article sets out the evidence in more depth; this playbook is the method.

Don’t start a community. Start by listing the ones your buyers already use. Ask your last twenty customers three questions: where do you ask peers for advice, which newsletters or podcasts do you actually finish, and who do you trust on this topic. Add the answers from “How did you hear about us?” on your forms; the Self-Reported Attribution Coder will sort a column of them into channels.

You will end up with a short list: two or three communities, a handful of people, a couple of newsletters. That list is your dark social map. Everything else in this playbook works from it.

Content travels through dark social when someone can send it to a colleague and look good for doing so. That rules out most gated PDFs and most “ultimate guides”. It favours a clear point of view, a framework someone can apply on Monday, a teardown of something real, or a number with its source attached.

The rule I use: ungate ideas, gate tools. Publish the thinking openly, because its job is to travel. Gate the things that fairly warrant an email, such as a workbook or a template, and make the download worth the trade. Every piece should stand alone: if the forwarded link needs the rest of the site to make sense, it won’t be forwarded.

In a community, a company account is a guest and a person is a member. The founder or a senior person should be the one answering questions, and the answer should help even when it doesn’t mention your product. The Founder-Led LinkedIn Playbook covers the public side of this; the private side is the same person, in the same voice, in rooms where a sales pitch would get you removed.

Set a simple rule for the team: in other people’s communities, answer first, link only when asked, and never post the same thing in more than one room. Then read each room’s own rules before the first post. On Reddit, for example, the sitewide rules prohibit spam and content manipulation, and each community’s moderators create and enforce their own rules on top^[5].

If you do run a community of your own, it has to offer something a follower count doesn’t: access to peers who share a problem, answers from people who have solved it, and a place where the vendor is present but not in charge. The communities that last are built around a role or a problem (“heads of marketing at seed-stage SaaS”), not around a product.

Before you open one, write down the one recurring conversation it will host, who will show up every week to keep it alive, and what you will do when it goes quiet in month three. If you can’t answer the second question with a name, don’t open it yet. A dead community costs more than no community.

Decide, too, which number it should move. Proving value is where community teams struggle: in CMX’s 2025 survey of 589 community professionals, 24% said they could confidently quantify their community’s value, and difficulty proving ROI was the most common budget challenge, named by 37%^[6]. CMX’s 2026 report says that confidence has since dropped^[7].

Every market has a few dozen people whose recommendations move buyers: practitioners who write, consultants who advise, the person everyone tags in a thread. Your Step 1 list names them. Don’t pitch them. Send them your best work with a specific note about why it applies to something they said, ask for their criticism, and make it easy for them to quote you. When one of them shares your framework in a private group, that is the most valuable distribution you will get this quarter, and no dashboard will show it.

Where money changes hands, disclose it. In the US, the FTC’s Endorsement Guides, revised in 2023, expect a connection between an endorser and a brand that the audience wouldn’t expect, such as payment, to be disclosed clearly and conspicuously^[8]. The FTC’s rule on fake reviews and testimonials, in effect since October 21, 2024, goes further: courts can impose civil penalties for knowing violations, including undisclosed reviews by company insiders and bought followers^[9]. Beyond the law, a paid mention that looks organic damages the trust that made it valuable. The Authenticity Audit for Creator Marketing covers paid partnerships in detail.

You can’t track dark social, so measure it the way you measure anything you can’t observe directly: ask, triangulate and watch the trend.

• Ask on every form. A free-text “How did you hear about us?” field, coded consistently. The Analytics & Attribution Playbook (Step 5) covers why free text beats a drop-down, how to code the answers and how many you need before the shares mean anything.
• Ask on every first call. One question, logged in the CRM: what made you reach out now?
• Watch branded search. Searches for your company name rise when people are talking about you where you can’t see; in my experience it is the best proxy you have. Google Search Console’s Performance report now has a branded-queries filter that separates them for you, although it isn’t available to sites with few impressions and some queries are misclassified^[10].
• Watch “direct” traffic to deep pages, as a share of all direct traffic. A stranger arriving directly at a specific article did not type the URL. Someone sent it.
• Watch AI answers. Whether assistants name you for your buyers’ questions is now part of the same picture. The AI Visibility Significance Test scores a run.

None of these gives a clean attribution number. Together they tell you whether the dark side of the funnel is growing or shrinking, which is the decision you need.

Dark social and the trackable funnel aren’t rivals. The forwarded article creates the demand; the retargeting ad, the newsletter and the sales call capture it. Once Step 6 tells you which pieces travel and which communities send buyers, point the visible machinery at them: retarget readers of the pieces that get forwarded, invite community members to the things you gate, and brief sales on which conversations to expect.

Name one person who owns community and dark social, usually the founder or whoever leads marketing. Each month they:

• Review the map. Which communities, people and newsletters produced self-reported mentions this month?
• Review the content. Which pieces did people forward, and what do they have in common?
• Review the presence. Where did you answer questions, and where did you go quiet?
• Pick one thing to change next month, and write down why.

Judge it against your own numbers over time. There is no benchmark for dark social, and anyone quoting one is guessing. Then check the quality of the pipeline that arrives this way. In my experience these deals close faster, because the buyer arrived with the decision half made; your win-loss notes will show whether that holds for you. If self-reported mentions rise and pipeline doesn’t, the wrong people are talking about you. Go back to Step 1.

Once a quarter, redo the twenty-customer questions from Step 1. The map moves.

What changes by stage?

What mistakes should I avoid?

• Starting a community to have one. Without a recurring conversation and a named host, it dies quietly, in public (Step 4).
• Pitching in other people’s rooms. The fastest way to lose the audience you were trying to reach (Step 3).
• Gating the thing that should travel. A framework behind a form reaches the people who fill in forms, and no one else (Step 2).
• Killing the channel because it doesn’t attribute. Cutting content because “direct” got the credit is the most common self-inflicted wound I see in reporting (Step 6).
• Faking it. Undisclosed paid mentions, seeded questions and sock-puppet praise are found out, and they poison the well for the real thing (Step 5).
• Measuring nothing. “We can’t track it” is not a reason to skip the question on the form (Step 6).