Blog analytics
Dark social: what it is and what you can measure
Dark social describes visits to a website from links shared through private channels when analytics does not retain the referrer. The term points to an attribution limit: you can see the visit without knowing who shared the link or which channel brought it. This article is about referral information, not a post or ad format.
A link sent by message, email or an app may open without passing a source analytics can recognise. Google’s documentation on understanding (direct)/(none) also lists directly typed URLs, untagged links, redirects, shorteners, offline documents and blockers among possible causes. The label therefore means “no clear source in this record”; it does not mean WhatsApp.
Alexis C. Madrigal popularised the term in his 2012 article on dark social and links shared privately. It describes email, messaging and some app links that did not pass a referrer. The analysis used Chartbeat data from publishers at that time; that historical estimate does not provide a current percentage for a shop, a brand or your website.
Read a report without turning a clue into attribution
Imagine a camera shop publishes a guide to choosing a travel lens. In April, its report records 620 sessions on that page. Links from a controlled campaign use UTM tags, and an optional question asks how people found the guide. Here is a fictional report:
| Record | What was observed | Reading limit |
|---|---|---|
| Guide sessions | 620 sessions during April. | The total reported sessions on that page in that period. |
| Tagged link | 210 sessions carry utm_source=instagram and utm_campaign=lente_abril. | The tag identifies the campaign written into the link, not the person who last shared it. |
| No clear source | 95 sessions appear as (direct)/(none). | The report cannot distinguish typed URLs, untagged links, messaging, redirects or other causes. |
| Other recorded sources | 315 sessions have another reported source. | These are sessions classified by the system; they do not explain the 95 visits without a clear source. |
| Optional question | 19 people answer; 8 say someone sent them the link privately: 5 name WhatsApp, 2 email and 1 another app. | These are self-selected answers, not a representative sample or sessions that can be assigned to the 95 visits. |
Swipe the table to see all columns.
In this report, the 210 tagged sessions, 95 sessions without a clear source and 315 sessions with other sources are disjoint groups under the same session definition and reporting period: 210 + 95 + 315 = 620. The 19 survey answers are people, not sessions, and stay separate. The eight people who remember a private share could be in any of the three session groups; their answers do not identify a specific visit. Do not add answers to sessions or present five WhatsApp mentions as part of direct traffic. The report can say that some links were tagged and some people remember a private share. The source of many visits remains unverified.
What you can check
Start with links you control. Use an agreed UTM convention for each social campaign that leads to your website, then test the destination after any shortener or redirect. Google’s documentation on campaigns and traffic-source dimensions explains how analytics collects these parameters when configured. If someone forwards the link, its tags may remain; they identify the encoded campaign, not the last sender. Our social media UTM guide explains how to name and check those links.
Keep (direct)/(none) as a no-clear-source category. You can check whether those visits reach a deep page, compare periods using the same definition and see whether a link change happened around the same time. These observations help you investigate. They do not establish that one app caused the visit or that a post generated a sale.
A short, optional question can add context: “How did you find this guide?” Offer answers such as search, someone sent it to me, email or another channel. Record how many people saw the question and how many chose to answer. Show the answers separately; people who opt in may have different experiences from everyone else. You do not need to read private messages or ask anyone to expose a conversation.
For a practical report, keep four lines separate: tagged campaign links, sources recorded by analytics, visits without a clear referrer and voluntary answers. Note the period, page and tool. If the direct category grows, first check whether links still have their tags and redirects preserve the parameters. Where you cannot observe the origin, leave it unknown.
What this means for measurement
Dark social is not a figure you can reconstruct by adding all direct traffic. It names visits that may come from private sharing when the referrer was lost. An origin may stay hidden, overlap with other causes or appear under a tag that remained on a forwarded link. Your report should preserve those distinctions.
The definition, case and downloadable check are original editorial work; the counts are fictional and do not estimate the size of the phenomenon. Use the dark social reporting checklist to review your next report. The cover is an AI-generated illustration.
Plan your next post in HeyMark.
Keep the idea, review the draft with your team, and see how it performed in the accounts you connected.