Direct traffic

Direct traffic is the label analytics gives a visit when it cannot tell where that visit came from. No referring website, no campaign tags, nothing attached. The name suggests somebody typed your address into a browser, and a minority of them did. The rest are visits whose origin went missing on the way.

In short

• Direct is the bucket for visits with no detectable source.
• Chat apps, documents and scanned codes all land in it.
• Tag the links you publish and most of the mystery moves out.
• Rising direct alongside a campaign is usually that campaign.

How a source is normally detected

Two mechanisms tell analytics where somebody came from, and direct is what happens when neither one fires.

The first is the referrer, a note the browser passes along saying which page the link was on. The second is UTM parameters, tags you add to a link yourself so the destination is told the source in plain words. A click from a search result carries a referrer. A tagged link in a newsletter carries tags. A link pasted into a private message frequently carries neither.

Nothing about that is a fault in the tracking. The browser was either not told, or was told and did not pass it on, and the analytics tool files the visit in the only place left.

What is hiding inside it

Most of the direct bucket is traffic that had a source until something dropped it.

• Messages and chat. A link shared in Messenger or a group chat, opened inside the app itself.
• Documents and files. Anything clicked from a PDF, a spreadsheet, a presentation or a desktop application.
• Scanned codes. A code on a poster, a receipt or a tarpaulin, where the address behind it was never tagged.
• Untagged email. Sends that carry no campaign tags, especially when opened outside a browser.
• Secure to insecure. A visitor moving from an encrypted page to an unencrypted one, where the referrer is deliberately withheld.

That last one is largely historical now that most sites hold a valid SSL certificate, and it still explains odd gaps on older sites and on links to a supplier who never updated.

Why two reports disagree about the same visit

A visit can appear as direct in one report and as a named source in another without either being wrong, which catches out everybody at least once.

The channel attached to the session describes that session alone. If nothing arrived with it, the session is direct. Attribution reporting asks a different question: who should be credited for the conversion. To answer it, the tool looks back over a conversion window and credits the last known source instead of the empty one, so a person who found you through an ad on Monday and returned untraced on Friday is credited to the ad.

Both numbers are correct answers to different questions, which is a useful thing to remember before deciding that Google Analytics 4 is broken. What you cannot do is add them together or compare one report against the other.

Telling a real direct visit from a lost one

You cannot see the source, and the visit itself still carries clues about which kind it is.

Landing page is the strongest one. Nobody types a long address ending in a product code, so direct visits arriving on deep pages are shared links or untagged campaigns. Direct visits landing on the home page are far more likely to be the genuine article, somebody who knew your name and went looking.

Timing is the second clue. A rise in direct that starts on the day a newsletter went out, a poster went up or a partner mentioned you is that thing, filed in the wrong column. Direct traffic does not spontaneously double.

New against returning visitors is the third. A direct bucket made mostly of first time visitors is not people remembering your address, since they have nothing to remember. It is sharing, and it is usually good news badly labelled. None of this recovers the source, and together the three of them tell you whether to go looking for a missing tag or leave the number alone.

What to do about it

The practical goal is not a smaller direct number. It is knowing that whatever remains in there is genuinely unknowable.

Tag every link you publish or print, using one consistent naming scheme, so that Facebook does not arrive under four spellings. Put tags behind scanned codes, because a poster with an untagged code is a campaign you are paying attention to and cannot measure. Tag the links your staff paste into messages, if that is a route customers actually use. Then watch what direct does over the following month.

If it stays large after that, the remainder is mostly private sharing and returning customers, both of which are good news. What matters is that you can now tell a rise in direct traffic apart from a rise in brand awareness, and those two get confused with each other in more reports than they should.

Words you will hear

• Referrer. The note a browser passes along saying which page the link was on. No referrer, no known source.
• Dark social. Sharing that happens in private, in messages and groups, where nothing can be measured from the outside.
• Channel group. The set of buckets a report sorts visits into. Direct is one of them, and the one used when the others do not apply.
• Unassigned. A separate bucket for data that arrived but matched no definition. Usually a tagging fault, and usually fixable.
• Self referral. Your own site appearing as the source, which means a visit was broken in two somewhere, often at a payment step.