How OWOX BI Server-Side Tracking Helped to Identify True Acquisition Sources Behind 30%+ (Direct) / (none) Conversions

Tracking Analytics Strategies

A high percentage of direct/none traffic is a headache for any marketer. When you don’t know the real sources of conversions, you can’t understand which channels to invest in, and you also can’t report on ad spend.

OWOX server-side tracking can solve this problem. A joint experiment with a client showed that with OWOX BI, the proportion of transactions attributed to direct/none decreased by over 21%. It was also possible to identify the correct source/medium for over 30% of revenue, sources of which were previously unknown. In addition, by correctly allocating transactions to channels, it was possible to recalculate the CPO and see that for some channels, it was lower than the CPO calculated previously.

In this article, we describe in detail the results of this experiment.

Book a demo

Gain clarity for better decisions without chaos

No switching between platforms. Get the reports you need to focus on campaign optimization

Book a demo

Problem: Significant proportion of traffic comes from unknown sources

You might find that​​ a significant proportion of sessions and conversions come from direct/none. Thus, it is impossible to understand from which sources these sessions and conversions truly come. The most common cause is a limited cookie lifespan, which leads to a situation where each subsequent session by a particular visitor is defined as a new session, and the connection with the very first session is lost. However, that first session contains the true source the user came from.

What is the essence of the problem?

First-party cookies in the Safari browser have a limited lifetime of seven days. The bottom line is that the clientId identifier is used to identify a specific user in Google Analytics. Thus, it is used by analytics tools as a key by which you can understand a user’s actions over a long period: where the user originally came from, what pages they visited, and so on.

The clientId identifier is written to the ga_ cookie and stored on the user’s device when the user visits your website. This is considered a first-party cookie, but since it is written in JavaScript, it is subject to ITP restrictions. So it lives no more than seven days.

This means that if a user visits your website today from a Facebook ad and places an order eight days later, your analytics tool will consider the purchaser a new user, and the order will not be attributed in any way to your Facebook advertising. The marketer becomes blind to this part of the traffic and, not understanding the real source of the order, may disable the allegedly ineffective advertising on Facebook. This could lead to a drop in orders and business profits. Hence, an increase in the proportion of new users in analytics may lead to a loss of income.

How OWOX BI solves these issues by handling direct/none in your reports

With OWOX BI, you can increase the accuracy of your ad campaign estimates and identify the true sources/mediums/campaigns that generate income. OWOX BI server-side tracking monitors any user activity on your website, extends cookie lifespans, and is not affected by ad blockers, allowing you to see the whole conversion path.

Cookieless server-side tracking out of the box

With OWOX BI, you can set up first-party data collection to solve ITP problems. To do this, at the integration stage, we create a separate subdomain on your website on which data collection will take place.

With each hit/event, OWOX BI creates a cookie ouid and renews it with each interaction with the user for 364 days. This cookie will have its own user ID: owox.user_id. Based on it, we may build analytics reports without a large share of fake new users and build a user path for a longer period. This makes it possible to correctly evaluate the effectiveness of advertising campaigns and track the entire user journey.

Joint experiment with an OWOX client

The problem of direct/none traffic was particularly relevant for the client with whom we conducted the experiment, as almost half of their traffic (44%) comes from the Safari browser.

In the experiment, we compared how the company's main metrics (transactions, revenue, CPO) differ when calculated based on data collected using different user identifiers: Google Analytics client_id and owox.user_id.

The key question we wanted to answer was for how many transactions the traffic source would change. Why is this important? Because the effectiveness of advertising channels is assessed based on the number of transactions by source/medium, and based on this effectiveness, decisions are made on redistributing the budget and reports are formed for management.

Results of the experiment

The experiment showed that using owox.user_id reduced the percentage of users incorrectly identified as new by 12%. This means that without using OWOX server-side streaming, the analytics system would have identified these users as new, but thanks to OWOX BI, these users were identified as returning, reducing the percentage of new users. For the experiment, we analyzed data for one month. Over a longer period, the reduction in incorrectly identified users should be even greater.

Share of new users

The next screenshot shows the percentage of users identified as returned (upper graph, by client_id; lower graph, by owox.user_id).

Percentage of users identified as returned

These graphs show the percentage of users recognized as “returned.” We can see that during the first seven days (while the Safari cookie is still active), the percentage of returning users is roughly the same for both methods. However, after seven days, the difference becomes significant. Thanks to owox.user_id, it is possible to recognize twice as many returning users on the eighth day, and six times as many returning users on the thirtieth day. 😎

The next screenshot shows the percentage of transactions for which the traffic source has changed due to the use of owox.user_id (for clarity, the assessment was made using the most popular First Click and Last Non-Direct Click attribution models).

percentage of transactions for which the traffic source has changed

For example, if we look at the data for March 27, we can see that the traffic source changed for 12% of transactions (according to First Click) and 6.8% of transactions (according to LNDC). This means that from the beginning, the source was incorrectly identified for these transactions. Consequently, the channels from which these transactions actually came were undervalued. This led marketers to form incorrect conclusions and inefficiently allocate the budget. Server-side tracking can solve this problem.

Now let’s move to the main part of the experiment and see how changes in the proportion of new/returning users and the transaction source affect the conversion rate, revenue, and CPO.

In the table below, we can see how the use of OWOX BI server-side tracking reduces the number of transactions with a direct/none traffic source. This is done by reallocating these transactions to their true source/medium.

Reduces the number of transactions with a direct/none traffic source

For example, let’s take data for April 6. We can see that the number of transactions with (direct)/none on that day decreased by 33.33%. These transactions were redistributed among other source/medium combinations: google/cpc received +12.5% transactions, +50% transactions, and so on.

The next table shows us how the revenue from transactions, which previously had source/medium as direct/none, is redistributed.

Revenue from transactions

For example, we see that on April 6, revenue from transactions with a source/medium of direct/none decreased by 32.78% points. However, this revenue was distributed among other channels and sources. This feels suspicious, as real channels that brought revenue were undervalued. Now, we can not only feel it intuitively but also justify it with numbers 😎.

We also noticed that for some source/medium, CPO predictably decreased. Why? Because the share of transactions from direct/none flowed into other source/medium categories. The number of transactions (in the denominator) to which expenses for a particular channel need to be divided increased, resulting in a decrease in CPO. To put it in marketing terms, transactions that were not accounted for in Google Analytics were included in these source/medium categories, indicating that their actual effectiveness is higher.

For example, on March 31, the CPO for google/cpc decreased by 8.77%:

CPO for google/cpc

The CPO for bing/cpc decreased by 12.5%:

CPO for bing/cpc

The CPO for Facebook/paid social decreased by 13.33%:

CPO for Facebook/paid social

Brief conclusions

Due to limitations associated with the use of third-party cookies, the share of new users and direct/none traffic is increasing. This significantly complicates the evaluation of advertising channels for marketers.

OWOX server-side tracking helps solve this problem: it reduces the share of direct/none traffic by 21% or more and correctly redistributes 30% or more of transactions and revenue to other sources and channels. Thanks to this, the marketing team can better meet their KPIs and more quickly defend their budget.



Expand all Close all
  • Why is a high percentage of direct/none traffic concerning for marketers?

    It obscures true conversion sources, hindering budget allocation and accurate reporting on ad spend effectiveness.
  • How does OWOX BI server-side tracking address the issue of direct/none traffic?

    It extends cookie lifespans, identifies true conversion paths, and reduces the percentage of incorrectly identified new users.
  • What are the benefits of using OWOX BI server-side tracking for marketers?

    It accurately reallocates transactions to their true sources, reduces CPO, and helps marketers better evaluate ad channel effectiveness.