Google Analytics: Attribution Models, Conversions and Traffic Acquisition
We are living in exciting times in the web analytics industry. Why? Because measuring capabilities keep getting better day after day. Google Analytics, the most used web analytics tool worldwide (2012) with its 82% market shares, keeps improving at a very rapid pace year after year.
Measuring Goals and Conversions
This post will cover some aspects of how Google Analytics measures conversions, and its recent evolutions.
First, why do we need conversions? Actually any business has purposes and goals. To achieve these goals you need means. It may be financial, time, human, or any other mean relevant to your business. Knowing your goals helps you better define how you’re going to achieve them.
For web analytics, a conversion or a goal is an action that someone achieves on your site. It is something you can measure, and of which you want to track the evolution.
At the end of the day, it will help you focus your work on how to improve and make your conversions and goals increase. For ecommerce websites, it will often be Sales (Transactions).
Nothing new so far!
Second, how to measure conversion and goals? Google Analytics tracks pageviews from visits from visitors. We can either set up Ecommerce, Events and/or Goals tracking.
Ecommerce tracking allows you to record all of the information regarding ecommerce transactions: product name, unit price, quantity, etc.
Events will be specific actions your visitors may do on your website which cannot be directly measured by tracking a pageview. For instance, someone clicking on a video play button won’t generate a new pageview. Such an action can be measured with Event tracking.
Goals used to measure when someone loaded a specific page. For example, seeing a thank you page means that the visitor made a purchase, and thus converted. You want to consider this as a Sales Goal. For few years already, Google Analytics has allowed us to set up different kinds of Goals such as when a visitor stays more than x time on your site or sees more than y pages (Engagement Goals - end 2009). More recently you have been able to use your events to create Goals based on them (Events Goals - April 2011).
Nothing new so far either!
The last click
Third, let’s talk about money! Because your goals are often monetary, you want to understand how to generate more money. Google Analytics attributes the value of the goal to the last source of traffic. We can call this the last click conversion model. A visitor may first come to your website after clicking on an AdWords ad, then will return by clicking on the link on your Facebook page, and buy something on your website. Google Analytics will attribute the conversion to Facebook.
This last click model is the most widely used model. Web analytics and advertising networks have used this model for years!
Limits of the last click conversion model
There are several limits to this model. One comes from how ad networks measure conversions. Suppose you are advertising simultaneously on two different ad networks. Each one has its own conversion system and conversion tag. A user clicks on your ads on both networks, visits your website twice, and converts during the last visit. Both ad networks will claim the conversion. Nevertheless from a Google Analytics perspective only the ad network that brought the last visit will be the converting source. As a marketer, to which ad network will you attribute the conversion? How much will you invest on the first ad network that was not the last click converting source? Will you?
Now how should you invest and/or optimize on each ad network? Both claiming the converting lead.
One answer is to use an ad server that will tag all your ads across the different ad networks. It will deduplicate the cookies and will only attribute the conversion to the last click. By now we should have the same results in terms of conversion attribution from our ad server than we have in Google Analytics.
But at the end of the day we came back to the starting point! The last click receives the conversion.
Moreover, ad servers tag links on ad campaigns but do not track links on forums, social networks, press articles, etc. You will still have the same question if a visitor clicks first on an online newspaper article and then comes back on an ad and converts.
One visitor, multiple visits
The above example illustrates that a visitor’s journey to a conversion may be much more complex than simply one visit and one conversion. A visitor may make several visits before converting. It will then probably come from different sources of traffic (paid search, organic search, referrals, social networks, online newspapers and so on).
Today web analytics and marketers’ challenge is to start understanding that journey. How people arrived on the website for the first time, how many more visits were needed, ...
At the end of august 2011, Google Analytics announced the Multi-Channel Funnels reporting suite (Google Analytics blog: Multi-Channel Funnels):
“This set of five new reports in Google Analytics gives marketers insight into the full path to conversion over a 30 day period, not simply the last click”.
This is where the future of conversion tracking is heading to. Technology enhancements are slowly allowing marketers to have a better overview of their website visitors journey.
Besides knowing how they interacted with the website, marketers will know more about how many touch points were needed, how some sources of traffic may influence their sales, which traffic sources initiated a sale, etc. This will allow marketers the re-think their budget investments with a more accurate and data oriented approach.
Written by Nicolas Debray
Sign up and receive our newsletter to keep you informed of our news, events, training ...
- Accelerated Mobile Pages
- Is Waze ads the key to local brand presence?
- Blogspot: GTM latest updates
- 5 Bid Adjustments to implement in your AdWords campaigns
- Facebook: Dynamic Product Ads
- Automated rules in Facebook Ads Manager
- Video is hot right now, here’s why!
- Business Case - Creating GTM variables based on existing DataLayers