What is direct traffic?

In order to understand how a certain session is defined as Direct, we need to go through one part of the concept of source precedence in Google Analytics.

Demystifying (direct) / (none)

The official definition of direct traffic from Google is “users that typed your URL directly to their browser, or who had bookmarked your site”. Even though this is true, it is somewhat of a simplification. In order to understand how a certain session is defined as direct, we need to go through one part of the concept of source precedence in Google Analytics. The concept of source precedence outlines how users and sessions will be reported in relation to traffic sources in Google Analytics (nerd mode here). A user or session will be reported as direct if the user goes directly to the website without having another source associated with its client ID within the Campaign Session Timeout period that is defined at property level for the specific Google Analytics view. The out of the box Campaign Session Timeout period in Google Analytics is 6 months, which means that we get the following:

User A comes to a website through an email marketing campaign. 2 months after that the user is going directly to the website by typing the URL in the same browser. This results in two sessions from one user:

• Session 1: source = email
• Session 2: source = email

User B comes to a website through a referral. 7 months after that visit, the user is going directly to the website by typing the URL in the same browser This results in two sessions from one user:

• Session 1: source = referral
• Session 2: source = direct

The raw data of these interaction paths, will be accessible in the Multi Channel Funnel (MCF) section of Google Analytics, but with a maximum lookback window of 90 days. Full raw data can be stored if using Google Analytics 360. This means that the MCF report is not affected by the Campaign Session Timeout. In that report these two scenarios would give us:

• Scenario 1: One interaction path with email and direct
• Scenario 2: Referral and direct in two separate interaction paths

I will not elaborate further on Campaign Session Timeout in this blog post, but a hint is that it might be a good idea to shorten or lengthen the Campaign Session Timeout depending on how you view the length of a specific consumer journey.

But what is direct traffic?

“Hey digital marketing agency, I would like to buy some more direct traffic. It’s really good for our business since it has a very high conversion rate and it feels really long term to invest in that channel.”

Direct traffic is not a channel and we shouldn’t define conversion rates based on sessions anymore (you know, mobile revolution, cross-device being a problem etc.). What is the conversion rate of a user who first visits the website through mobile and then through desktop followed by a conversion? The true conversion rate of the user is 50% even though desktop would report 100% on a direct visit. Let’s move away from session-based definitions!

If you think about it, it’s logical to say that direct traffic is not a channel. It’s simply a matter of navigation. If we simplify it a bit more, it’s fair to say that direct traffic can be divided into two segments:

• Direct traffic as a result of tracking problems
• True Direct traffic (e.g. brand awareness)

Direct traffic as a result of tracking problems can come in many shapes and forms, and the fact is that it’s a growing “problem”. Back in the good old “beginning of the internet” days, users were operating the internet through one device and the landscape of channels was limited. This created a landscape that, from a cookie-based measurement perspective, was rather simple. Today, we have several devices and a mature landscape of channels, which creates a bigger tracking problem. A clear example would be to isolate the effect of the cross-device problem in relation to direct traffic. If a user first visits a website through mobile and then comes back via desktop, which will result in the desktop visit being reported as a path with one direct interaction, which is not the case. This gives us the effect that the increase in cross-device behaviour is causing an increase in direct visits, but that’s simply a measurement error. Sorry, some parts of your growing direct traffic are not due to your great job in strengthening brand awareness.

How about the true direct traffic?

This is a bit more complicated to go through. You could argue that no direct traffic is true direct traffic. A simple mind game would be to use the (fair) assumption that no user is born with information about a company’s URL. This means that the action that a user takes in going directly to a website must be affected by something else. This could be word of mouth, display ad impressions, tv ads etc. But remember that most of those things should be in the category of “tracking problems”. If a user first sees a TV ad and then go directly to your website, that is not true direct – again it’s just a tracking problem. However, we should argue for repetitive direct visits over time being a true direct visit, since the effect now no longer could be argued to attribute the direct values back to the initial TV ad interaction.

What can we do about it?

Now that we understand the tracking definition of direct traffic, and that it’s actually made up from one part “true direct traffic” and one part “tracking problems”, we need to elaborate a bit on what to do with this information. The end goal should be fairly simple: if you could get an understanding of how much of the direct traffic is true and how big a portion is due to tracking problems you could reach two valuable insights:

• Having a clear understanding of the current brand awareness value of your business (true direct traffic)
• Re-allocate the part of the tracking problem to the right channels (TV, offline, online and other paid activities) in order to evaluate paid marketing initiatives more precisely

How to do this is of course the million-dollar question and it depends a lot from case to case, but at Precis we use two different methods to assess this problem:

• Marketing Mix Modeling: Through the use of Marketing Mix Modeling, we can create a statistical model using direct visits as the target variable, and reach conclusions on the effect from independent variables (paid activities, etc.).
• Data-driven Attribution Modeling: Through our proprietary data-driven attribution algorithm, we can modify the interaction paths and merge them, in order to run data-driven attribution towards direct visits. This analysis results in us showing what marketing investments that generate direct visits.