Categories
Explainer Public

A Case Study on the Importance of the Referral Exclusion List

Proper configuration of the referral exclusion list is very important, because the referral exclusion list has a huge effect on how visits to you website are categorized. Unfortunately, the default settings for the referral exclusion list are not well-suited virtually all Mason websites. This means that, unless you have updated your referral exclusion list settings from their default, it is highly likely that a large amount of your traffic is being miscategorized.

The default settings make sense for most websites, in which the website is a single domain, like www.example.com. In this case, you can imagine how blog.example.com and store.example.com are not separate websites from example.com. but are just sections of the website.

However, in the case of Mason, most of our websites are subdomains. For tracking purposes, and in many other ways, volgenau.gmu.edu and cvpa.gmu.edu are truly separate websites from gmu.edu.

Let’s take a look at a representative example. Like most Mason websites this one was configured by default to exclude traffic from “gmu.edu” from being counted as referrals. Note that this does not mean that this traffic from other gmu.edu websites was not counted; but it means that traffic from other gmu.edu website was not counted as referral traffic, bu as direct traffic.

If you ever looked at your reports and wondered why your website apparently didn’t receive any traffic from other Mason websites, this is why.

Upon review of the analytics implementation for this website, “gmu.edu” was removed from their referral exclusion list on Oct. 12, 2017. Let’s take a look at the before-and-after to see what effect this configuration change had.

First, let’s take a look at the situation before “gmu.edu” was removed from the referral exclusion list:

Report: Acquisition -> Overview
Date Range: Oct 2, 2017 – Oct 8, 2017

Note that referrals make up a tiny slice of overall traffic, at around 2.5%. Organic search is the largest segment, which is very common with the tendency for users to use search as navigation. But direct traffic seems unusually high.

Let’s take a look at the referral traffic that resulted in visits the this website during this week:

Report: Acquisition -> All Traffic -> Referrals

Top 10 referring websites:

Source Sessions
Total 151
cyberdegrees.org 46
masongrad.hobsonsradius.com 18
duckduckgo.com 9
cgep.virginia.gov 6
m.facebook.com 6
quora.com 6
mail.google.com 5
gmudigcom.teamwork.com 4
datascience.community 3
findengineeringschools.org 3

Note that there are no mason websites listed. No Mason websites sending referrals to this website at all. That seems very odd.

It seems like something is wrong. And something is indeed wrong. The referral exclusion list is set to exclude all gmu.edu websites as referrals. Instead, all gmu.edu websites are being counted as direct traffic, which also explains the high percentage of direct traffic.

As we said above, the referral exclusion list was updated on October 12, 2017 to remove “gmu.edu” from the referral exclusion list. Now let’s look at the situation after “gmu.edu” was removed from the referral exclusion list:

Report: Acquisition -> Overview
Date Range: Oct 16, 2017 – Oct 22, 2017

Note that referrals have increased from ~2.5% of traffic to ~23.9%, and now represent the second highest acquisition channel behind organic search. The direct traffic percentage looks much more reasonable now.

And how do the referrals look now?:

Report: Acquisition -> All Traffic -> Referrals

Source Sessions
Total 1381
www2.gmu.edu 732
advising.gmu.edu 162
cs.gmu.edu 75
ist.gmu.edu 41
cyberdegrees.org 38
ece.gmu.edu 30
telecom.gmu.edu 27
seor.gmu.edu 24
statistics.gmu.edu 24
duckduckgo.com 15

Note that we see data that looks much more in-line with our expectations. There are a large number of referrals from other gmu.edu websites.

The issues with the referral exclusion list have been corrected and the data is now accurate.

Leave a Reply

Your email address will not be published. Required fields are marked *