The word Google Analytics 4 has been thrown around a lot lately. The current Universal Analytics (also called Google Analytics 3) is ending this year on July 1, 2023. It will be replaced by Google Analytics 4 (GA4). All e-commerce store, blog, and other website owners will use this tool from July onwards.
If you use affiliate marketing as one of your marketing channels and look at GA4, you may see completely different traffic and conversion numbers. In this article, we will explain why this is the case.
Note: “dgt” on the screenshots represents one affiliate marketing network.
1. Poor deployment of GA4
A very basic thing that can skew not only affiliate marketing data but all reports. The easiest way to check for accuracy is to compare your data from GA4 and from Universal Analytics or with another data source (e.g., your CRM).
The most common mismatches occur if you have implemented GA4 via a plugin or by embedding the code on the web. The most stable implementation of GA4 is via Google Tag Manager. In addition to flexibility, it allows you to catch flies using Debug mode.
2. Attribution
Attribution in Google Analytics refers to the process of assigning credit for a specific conversion or user action to a specific marketing channel or touchpoint within a conversion path. Attribution models allow us to understand which marketing channels and campaigns are contributing the most to our business goals and optimize our marketing strategies accordingly.
By default, Google Analytics uses a last-click attribution model that assigns 100% of the credit for a conversion to the last point of contact the user interacted with before the conversion. However, there are other attribution models available that take into account multiple touchpoints in the user journey, such as the first-click model, the linear model, and the time decay model.
Google Analytics 4 defaults to cross-channel data-driven attribution, or data-driven attribution that attributes conversions based on how people engage with different ads and make purchase decisions. It uses user behavior data to determine which keywords, ads, and campaigns have the greatest impact on conversions. Google Analytics thus “breaks down” conversions into smaller parts within the conversion path and assigns, e.g., 1 conversion to 3 sources in the ratio 0.4 + 0.4 + 0.2.
Example of a conversion path:
Google search > Facebook remarketing > Affiliate > BUY
Universal Analytics (last click) conversions
Google = 0
Facebook = 0
Affiliate = 1
Google Analytics 4 (data-driven) conversions
Google = 0.3
Facebook = 0.1
Affiliate = 0.6
It was not possible to set up an attribution model in Universal Analytics (UA). The last click was the default. However, in GA4, we can choose from 7 attribution models so far. We can also compare data between models. Therefore it is important which model we choose. The differences can be significant.
3. UTM parameters
UTM parameters are tags you can add to URLs to track the source, medium, and campaign of your website traffic in Google Analytics.
Example:
https://www.domain.com?utm_source=dgt&utm_medium=affiliate&utm_campaign=xyz
If we were sending any source, medium, or campaign data to UA incorrectly due to poorly set UTM parameters, we still had the option to adjust the data via filters. However, filters do not yet exist in GA4. That’s why using UTM parameters correctly to tag the URL in your campaigns is very important. If you don’t do this, a minor problem can be that you will have inconsistent data in your reports, for example like this:
The bigger problem is that not only visits from affiliates (also applies to Facebook, TikTok, Heureka…) will “fall” to the direct source.
4. Source / Medium = not set
Your traffic report may also look like this. It is clear that not set contains data from other sources. We just can’t figure out which ones.
Recommendations:
- Try to look in UA and identify which source you are falling into not set.
- Verify the correct use of UTM parameters. Just omit one “=” and Google Analytics will no longer be able to identify the parameter. Thus, it may end up in not set.
- Test whether some other tag, such as a pixel or cookie bar, is blocking you from overwriting the source.
5. Cookie consent
If that wasn’t enough, we have cookie consent. More and more users (30-50% in our country) refuse cookies. Laws are getting stricter every year, so this number will go up. Users who refuse cookies are “lost” to Google Analytics. However, GA4 has come up with a solution called conversion modeling. Behavioral modeling uses machine learning to model the behavior of users who have rejected analytics cookies based on the behavior of similar users who have accepted analytics cookies.
Conversions are modeled up to 7 days backward. Therefore, it is possible that in GA4, you have 10 conversions from affiliate network today, but 7 days from now, you will have 15. Of those, 5 will be modeled. You can turn off/on the display of modeled conversions in the “Reporting Identity” settings.
Solution
Look at the data from a top view
- Compare attribution models
- Keep track of your e-commerce store’s data (orders, coupon usage) – these are always the most reliable
- Consult discrepancies with your affiliate partner
- Prepare a customized report in GA4 to validate the data. The Explorations section works with raw data without sampling. Watch the video tutorial on how to create such a report.
Follow trends and behavior patterns
- The days when we had all the data about users are gone. We have to make decisions with a smaller percentage of data, but that data has to be of good quality. Based on that, we can identify behavioral trends.
- Follow MER (marketing effectiveness ratio) instead of outdated metrics like ROI or ROAS. Compare investment on a specific campaign vs. total sales and track its evolution over time. This way, we can explicitly determine what change an increase/decrease in investment in a given campaign has made on overall turnover.
Let’s have Google Analytics set up correctly 4
- Attribution model, properly selected attribution window
- Manual tagging
- Correct session timeout length
- Using UTM parameters
- Consent mode
