Community & Dark Social

Community & Dark Social

A founder shows me a dashboard where "direct" is the biggest source and asks why paid search is the only channel that works. Usually the best channel is working and the dashboard cannot see it: someone read a post, pasted the link into a Slack channel, three people opened it on their phones, and one booked a call a week later. Analytics recorded a stranger arriving from nowhere.

Alexis Madrigal coined "dark social" at The Atlantic in 2012 for sharing that carries no referrer. In B2B it now covers most of the places a buying conversation happens. Gartner's research on B2B buying groups found they spend 17% of their purchase time meeting with potential suppliers; Forrester's 2024 Buyers' Journey Survey found 92% of buyers start with at least one vendor already in mind. The shortlist is built where you cannot see. This skill is how you get on it, and how you know whether you are.

When to use

The user wants to reach buyers through communities, peers and private channels; is deciding whether to start a community; or cannot tell whether word of mouth is growing because nothing attributes to it.

Do not use it to find communities from nothing. The map is built from what customers say: ask your last twenty where they get advice, which newsletters and podcasts they finish, and whom they trust. The Self-Reported Attribution Coder turns a column of "How did you hear about us?" answers into the same list.

Before you start

1. Know your buyer in one sentence (.agents/product-marketing.md). A community is built around a role or a problem, never a product.
2. Ask the twenty-customer questions and count how many named each place.
3. Put a free-text "How did you hear about us?" field on every form and a source question on every first call, coded consistently. Free text beats a dropdown because people name the podcast, the person or the room.
4. Have branded-search impressions and direct traffic to deep pages to hand, by month.

Method

Map by evidence, not by reach. The engine ranks each place by the buyers who named it plus the leads who mentioned it. A place named by two or more buyers where nobody from your team is present is a gap to fix (two is this engine's threshold). A place you are in that nobody named is kept only if it costs nothing.

Show up as a person. In someone else's room a company account is a guest and a person is a member. The engine flags any room where you pitched or posted an unasked link, and any room where the same post went into more than one channel. It also flags presence without answers: lurking is not showing up.

Open your own community only when four things are true. A named recurring conversation; a named host with at least two hours a week (this engine's floor); a written plan for month three, when it goes quiet; and at least ten people who have asked for a room that does not exist (this engine's floor). No host means no, whatever else is ready. A dead community costs more than no community.

Measure by asking, then read the trend. Nothing here attributes cleanly. The engine computes, per period, the share of coded answers that name a dark source (a person, a community, a podcast, a newsletter, or your name directly), branded-search impressions, the share of direct sessions that land on a deep page, and pipeline. It compares the latest period with the average of the earlier ones, treating a move within 10% as flat. Then:

• working: dark signals and pipeline rising together;
• wrong people: dark signals rising, pipeline not. Check the sales cycle first; if it is not that, the wrong people are talking;
• growing: dark signals rising, no pipeline given;
• going quiet: dark signals falling;
• flat: no movement.

Fewer than ten dark-source answers in a period is too few to read as a share, and more than 30% of answers uncoded means the form or the coding needs fixing before the trend means anything. Both are engine rules.

Workflow

Step 1 - Map. Twenty customers, three questions, a count per place. Add the coded form answers.

Step 2 - Show up. Run the map check. Join the gaps as a person, read for two weeks, then answer. Fix any room you pitched in.

Step 3 - Decide on your own room. Run the community check. If it says not yet, keep showing up in other people's rooms and re-run when members are asking.

Step 4 - Measure monthly. Two periods minimum, three is better. Run the measurement read and pick one thing to change next month, written down.

Step 5 - Feed the visible funnel. Retarget readers of the pieces that get forwarded, invite community members to what you gate, and brief sales on the conversations to expect.

Step 6 - Redo the map quarterly. It moves.

Run the tool

node resources/dark-social-read.js --demo

run({ map, community, measure }) returns the ranked map with coverage and findings, the go / not-yet / no on your own community with what is missing, and the per-period reads, trends and verdict.

Present the result

1. The map, ranked, with the places to join, the rooms to stop pitching in, and coverage as "present and answering in X of Y places buyers named".
2. The community call and the gaps, in one line each.
3. The trend table: dark share, branded search, deep-page direct share and pipeline per period, with the direction of each.
4. One change for next month, and why.

Guardrails & common mistakes

• Never start a community to have one. Without a conversation, a host and demand, it dies quietly, in public.
• Never pitch in other people's rooms. Answer first, link only when asked, one room per post.
• Never kill a channel because it does not attribute. Cutting content because "direct" got the credit is the most common self-inflicted wound in reporting.
• Do not quote a dark-social benchmark. There is none; use your own numbers over time.
• Do not seed questions, run sock puppets or pay for undisclosed mentions. They are found out, and they poison the well for the real thing.

Sources

• Wikipedia, Dark social media - Alexis Madrigal coined the term in The Atlantic, October 2012, for sharing of URLs with no referrer.
• Gartner, Win more B2B sales deals - buying groups spend 17% of their purchase time meeting with potential suppliers.
• Forrester's 2024 Buyers' Journey Survey, reported by Digital Commerce 360 - 92% of B2B buyers start with at least one vendor in mind, 41% with a single preferred vendor.
• Gartner, 67% of B2B Buyers Prefer a Rep-Free Experience - 9 March 2026, survey of 646 B2B buyers.

Related AAJ resources

• The eight-step method this skill runs: https://aajconsult.com/playbooks/community-dark-social-playbook
• Coding "How did you hear about us?" into the map: https://aajconsult.com/tools/self-reported-attribution-coder
• The public half of showing up as a person: https://aajconsult.com/playbooks/founder-led-linkedin-playbook
• The argument: https://aajconsult.com/blog/dark-social-b2b-demand-generation

Related skills

• founder-led-content - the founder's public channel; this skill is the private one
• creator-partnerships - when the trusted person is paid
• aaj-marketing-loops - which loop the forwarded piece feeds
• marketing-report - putting the trend read in front of the board