The Ecomm Analyst

Direct is not a channel. It is a default. When a visit arrives without a referrer and without campaign parameters, the analytics tool has nowhere to file it, so it goes in the bucket labeled direct. That label reads like a finding, which is the problem. People see twenty-eight percent direct in GA4 and conclude they have strong brand recall, when what they actually have is twenty-eight percent of their traffic that the tracking could not resolve.

The distinction matters because those two readings lead to opposite decisions. If direct is brand demand, you fund brand. If direct is measurement failure, you fix tagging. Most stores I look at are in the second situation and spending like they are in the first.

What is actually in the direct bucket

Untagged links are the largest contributor at almost every store. Any link you place anywhere without UTM parameters comes back as direct if the referrer is also missing. That covers a lot of ground: links in email clients that strip referrers, links pasted into Slack and WhatsApp and iMessage, links in podcast show notes, links in a PDF, links in a QR code on packaging.

Dark social is the next chunk. Someone screenshots your product, sends it to a friend, the friend searches your brand name or types the URL. That visit is genuinely direct in the mechanical sense and genuinely caused by something else in the causal sense. Nothing in your stack will connect those two events.

Then there is the technical residue. In-app browsers that do not pass referrers cleanly. Redirect chains that drop parameters on the way through. Link shorteners configured badly. Sites that link to you but suppress the referrer header. And increasingly, traffic arriving from AI assistants, which some tools now classify as its own channel and others still dump into direct.

Why the number is worse in GA4 than in your attribution tool

GA4 resolves a session against what arrives in that session. An attribution platform is doing something different. It is stitching a visitor across multiple sessions, so a visit that looks direct today can be credited back to an ad click eleven days ago because the same visitor is recognizable. That is the actual product difference, and it is why the same store can show thirty percent direct in GA4 and eight percent in a dedicated tool.

Neither number is lying. They answer different questions. GA4 is telling you how this session arrived. The attribution tool is telling you what it thinks started the journey. When a vendor quotes you a low direct percentage as evidence of accuracy, what they are really demonstrating is a longer lookback window and identity stitching, not better tracking of the original click.

How to shrink the bucket

Tag everything you control, which is more than you think. Every email link, every SMS link, every affiliate link, every QR code, every link in a partner’s newsletter. Then check your redirects, because a link that passes through two hops before landing loses its parameters somewhere in the middle more often than not.

For the part you cannot tag, ask. A post-purchase survey is the only instrument that catches dark social, word of mouth, and offline discovery, because it asks the one entity that actually knows. I went through how to read that data without overweighting it in using post-purchase surveys as a real attribution signal. Survey answers are noisy and biased toward whatever the customer remembers last, but noisy data about an unmeasurable channel beats no data.

ThoughtMetric, which sponsors this blog, handles the first half of this by stitching visitors across sessions so returning traffic resolves back to its original source rather than landing in direct. That does not solve dark social, and no pixel will. It does mean the direct number you are left with is closer to the genuinely unmeasurable remainder rather than a pile of tagging mistakes.

Common questions

What percentage of traffic should be direct?

There is no correct number, but there is a useful diagnostic. If direct is above roughly a quarter of sessions in GA4 and your store is under five years old, tagging is the likelier explanation than brand strength. Established brands with real name recognition can legitimately run high. A store that launched eighteen months ago cannot.

Does direct traffic mean people typed my URL?

Rarely. Typed URLs and bookmarks are a small share of the bucket at most stores. The majority is untagged links and stripped referrers. Treating direct as a proxy for brand demand overstates brand demand at almost every store I have looked at.

Can an attribution tool eliminate direct traffic?

It can reduce it substantially by connecting return visits to earlier sessions. It cannot eliminate it. A first-time visitor who arrives with no referrer and no parameters has left no signal to work with, and any tool claiming otherwise is modeling rather than measuring. Ask which of the two you are being sold.