That post did well. You can see it in the app: the link sticker got clicks, two people replied asking where to buy, someone you have never met shared it to their story. Then you open GA4, sort last week's sessions by channel, and Social is a sliver. Direct is up by a few hundred.
So either the platform is inflating its own numbers, or the analytics is broken, or the traffic was never real. It is usually none of those. Social media traffic showing as direct is the normal behaviour of the web, and once you know why it happens you can stop trying to close the gap and start measuring the thing that actually matters.
Direct is not a channel
This is the part that trips people up. Direct looks like a source in the report, sitting there next to Organic Search and Paid Social, so it reads as "people who typed our address in".
It isn't. Direct is where GA4 files a session when it has nothing else to go on.
Channel assignment works off two inputs: the campaign parameters in the URL, and the referrer the browser hands over when it sends someone to your site. If both are missing, the session has no identifiable source, and the only honest bucket left is Direct. Your typed-address traffic and your untagged social traffic end up in the same pile, which is why the pile grows every time a post lands.
Where the referrer goes missing
A browser is supposed to hand over the page a visitor came from. Plenty of ordinary situations mean it doesn't.
- Links shared in messages. Someone sees your post, copies the URL, and drops it in a WhatsApp group or a DM or a Slack channel. The click comes from a chat window, not from a web page, and there is no referring page to report.
- In-app browsers. Tapping a link inside Instagram or TikTok opens a browser inside the app. Whether a referrer survives that trip varies by platform and by app version, and it changes without announcement.
- The linking page's own settings. A site can tell browsers how much referrer information to pass on, and some pass nothing at all. That is a decision made on their end, not yours.
- Anything that shortens or rewrites the link on the way through.
This is what people mean by dark social: real sharing between real people that leaves no trace in your analytics. It is not a tracking failure. It is what word of mouth has always looked like, arriving through a channel that happens to sit next to a dashboard.
What UTM tags fix, and what they don't
Campaign parameters solve half of this, because they live in the URL rather than in the browser's memory. When the referrer vanishes, ?utm_source=instagram&utm_medium=social is still sitting in the address, and GA4 reads it.
So tag every link you control. Bio links, link stickers, the URL in a scheduled post, the button in your newsletter. It takes an afternoon once and it permanently improves what you can see.
Then be clear about the limit. A tag only survives on the link you published. The moment someone copies your product page URL out of their own address bar and pastes it to a colleague, your parameters may or may not come with it, and the second visitor arrives clean. The strongest signal in your funnel is the one you have least control over.
Two habits worth keeping alongside the tagging:
- Never put UTM parameters on internal links. Tagging a link from your own blog to your own pricing page starts a new session and rewrites its source, which turns a visitor you already acquired into a fresh acquisition from yourself.
- Agree a spelling and stick to it.
Instagram,instagramandigare three different sources in your reports, and nobody will notice until the quarterly review.
Split Direct by landing page
Here is the diagnostic that takes ten minutes and changes the conversation.
Open your Direct sessions and break them down by landing page. Somebody typing your name into a browser lands on the homepage. That is the only page they know how to reach without a link. So homepage-heavy Direct traffic is roughly what the label claims to be.
Direct sessions landing on /products/replacement-filter-kit are a different animal. Nobody types that. Every one of those sessions arrived through a link that lost its identity in transit, which means it came from a message, a shared post, or somewhere else you could not tag.
Watch that second group over a few weeks. If it rises after you post and falls when you go quiet, you have found your social traffic. It is being counted, just under the wrong name.
Measure the outcome, not the attribution
At some point the honest move is to stop trying to perfect the channel report. Two tools measuring the same thing in different ways will disagree, in the same way that a Lighthouse score disagrees with Search Console, and neither one is lying.
Three things worth more than a tidier attribution table:
- A "how did you hear about us" field on your enquiry form, free text, not a dropdown. It is self-reported and imperfect. It is also the only place a customer can tell you about a conversation that happened in a group chat you will never see.
- Total volume against activity. Look at all traffic and all enquiries in the weeks you published consistently, against the weeks you went quiet. It is a blunt instrument. It answers the question the channel report can't.
- A destination that only exists on one channel. A specific landing page, offer or code used nowhere else. Anyone who arrives there came from that channel, whatever the referrer says, because there was no other way to find out about it.
Where to start this week
Tag everything you control, then run the landing page split on your Direct traffic and write down the number. That is your baseline for untagged sharing.
Check it again in a month. If the deep-page share of Direct is growing while you are posting, the work is landing, and the report was the thing that was wrong. If it is flat, that is worth knowing too, and it is a much better prompt for changing the plan than a channel table that was never going to show you the answer.
Reporting on how we run social media starts here rather than with a screenshot of impressions, because the point of the exercise is to find out whether any of it reached someone who buys.
