webanalytics.cl

Last week there was a very active post and comment thread on Francois Derbaix’s blog (Francois is CEO of www.toprural.com). His post explained the difference he had found between the organic traffic measured  for toprural by XITI and Google Analytics. In just a few days 71 persons commented or linked to the post, some of them just tailgating on the “Google is Evil” theme.

He implied that Google used GA to give a wrong idea on the share of organic traffic over total visits and basically “scooped” the news that GA has a campaign tracking latency with a default value of 6 months. In my opinion Google has been pretty clear on what information is reported on the traffic sources and also regarding the configuration options to modify the campaign tracking latency. So basically, the web analyst using the tool is responsible for the configuration and knowing what GA is measuring and reporting.

To analyze the comments thread I marked each comment as either a trackback or a comment, then I reviewed and matched either the content of the comment or the post linking to the thread to one or more of the following themes:

1. Google is evil or/and Why GA lies to me?
2. Surprised and/or uninformed and/or thankful to know
3. Genuine concern and/or Request for advice/Acknowledge need of specialized support
4. Provide set-up advice and/or additional information
5. Watercooler or miscellaneous conversation and yadayadayada
   

After browsing the 123 comments and posts involved directly in the conversation till the 23/02/2009 I can highlight the following:

• 71 participants (including post author and guest-star  Avinash Kaushnik)
• 123 comments, 27 of them were trackbacks
• 1 every 3 participants was surprised to know Google Analytics track traffic sources in this way and the latency was 6 months
• Only 1 every 10 participants provide set-up advice, solutions or additional information to enrich the conversation
• 12 out of 71 participants at one point of the conversation wrote about the Google is evil theme
• And 6 out of the 27 posts linking the original post were about Google is evil!
• 21 out of 71 participants show genuine concern, requested advice or acknowledge the need for specialized support.

For me the biggest discovery was to learn that so many people use Google Analytics without bothering to learn about its configuration nor the data reported and how quick they place the shame on the tool once they realize they have been getting it wrong.  I’m thinking  “Eyes wide shut”.

Have you ever wondered what the last column of your Top Contet reports showing $ Index stands for?

Well, this is a very short explanation!

If you have monetized your traffic within Google Analytics (either by assigning a value to a goal or by tracking ecommerce transactions with a value) you will see a $ index value that is calculated like this:

For each content:

Sum up the revenue that was generated after visiting that particular content and divide it by the total visits in which that content was viewed at least once.

So imagine yesterday you had only 2 visits that viewed the following content:

• Visit 1:   page 1 > page 2 > goal page > page 3
• Visit 2:    page 2 > page 3
  

Goal page has been assigned a value of $ 2. What is the $ index for each page (for yesterday)?

• $ index page  1 = $2 ($2 divided by 1)
• $ index page 2 = $1 ($2 divided by 2)
  
• $ index page 3 = $0 ($0 divided by 2)
  

As you can see the money index (especially for contents with a lot of unique views) can be understand as the expected revenue of a visit that sees that content. If we are able to increase traffic to these high $ index contents without dropping the figure will be increasing our overall site revenue.

Good treasure hunt!

Most websites have some sort of content organization that group individual pages into directories and subdirectories (and/or subdomains). Although web analysts are not compelled to follow the exact same convention when structuring reports, most of us do it, making Directory Structure one of the most common schemes to organize content reports (an excellent in-depth reading to understand the different content organization options, which include Information Architecture and Business Objectives organization schemes, can be found on Web Analytics Demystified book, chapter 5).

This post deals with how to analyze one or more sections of your site.

Let’s consider Linkara, a Spanish social network for people interested in movies, books and music. They have at least 3 sections, one for each content vertical, structured in subdomains cine.linkara.com, libros.linkara.com and musica.linkara.com.

For simplicity of this post we’ll assume that traffic (both visits and page views) is the only factor considered to weight out the importance of a section (conversion is not considered in this analysis).

Graph 1

Graph 1 shows page views from all contents within the books section (obtained applying a regex filter within the Top Content report).

Graph 2

Graph 2 shows the entrances through contents in the books section.

Graph 3

Graph 3 shows visits in which at least 1 content of the book’s section was viewed (before Advanced Segments you couldn’t really get this information unless you went through a lot of trouble setting up profiles and filters to either group all contents within a section in a single content or map them to a single title).

How would you evaluate the contribution of the books content vertical to the overall site traffic? 

How do you measure the relative success of the section over time?

Although a clear trend pops from both graphs showing a traffic increase, we can not answer the above questions looking at this data. What are we missing?

CONTEXT!

In this case context comes in the form of overall traffic figures and allows us to build 3 metrics:

1. Books section’s page views over total page views or the contribution of the section to the total page views
   
2. Entrances through a page in books section over total entrances or the contribution of the section to traffic attraction
   
3. Visits in which at least 1 page in books section was viewed over total visits or the reach of the section over all visits (this can work for unique visitors too, but I prefer visits)

Graph 4

Now that we have context we can say that the growth in the books section is not only explained by an overall increase in traffic to Linkara (during the same period analyzed) but is also an increase in reach over the visits of Linkara and a greater contribution to site page views. In other words the books section is increasing its importance in generating page views and its relevance during visits. Also very important, book section is gaining attraction over organic traffic attraction (mainly organic in the case of Linkara).

Finally, can we analyze more than one section simultaneously?

A good option will probably be using a motion chart within a google doc. You can visualize absolute metrics like PVs, Entrances and Visits or contextualized like contribution to PVs and Entrances or Visits reach. In this case since you have both visual (size, color,position and several sections simultaneosulsy) and time context in the motion chart is probably better to use absolute figures.

In this post I will show you an example that you might found very useful.

A little bit of background
If you were using Google Analytics utmSetVar() to set a user defined variable in order to segment your traffic and also were working with profiles and applying filters the chances are you experienced some of the following situations on your reports (Look from October 2007 till the day you switched to the new Google Analytics tracking code ga.js).

1. Unexplainable increase in visits and consequent effect on some other metrics

In this report visits increase between a 10% to 15% on October 2nd and keep that way without a logical explanation.

2. Visits with less than one page view

For this profile we had a significant amount of visit with less than 1 page view. Clearly a sign that something was not right.

3. More visits than page views. At this time I don’t have any report to show this case but if you go to the unofficial Google Analytics Support Group on Google Groups and search for “more visits than page views” you’ll find users that have experienced this.

The motivation

Since I had some customers that had experienced this issue for several months. When Advanced Segments were announced I started thinking on how to use this feature to obtain a historic visits trend for 2007-2008 without the presence of the over counted visits that I decided to call ghost visits, that was the easiest name to explain management that those were not real visits (Spooky, I know…).

The solution, in 2 easy steps

1. I created a very simple Advanced Segment, which I named without ghost visits. The condition is quite straightforward, include visits with Page Depth Greater than or equal to 1, this will exclude all visits with less than 1 page view.

2. I applied the Advanced Segment to Reports and enjoyed my reprocessed data!

This figure shows the corrected visits per week graph.

In this case the corrected Average Page views per visit. (It seems there is a small bug in the GUI, for this case the graph doesn’t show the weeks.
 

 
And finally the modified depth of visit report.

As you can see, the Advanced Segments feature gives you a great opportunity to look back into your data in a different way. I’m sure you can think of more scenarios where this look-into-the past and segment your data ability can be really useful.

In the last weeks two things happened that make me think of how we segment visitors, measure engagement and how it contributes to a web’s porpoise or revenue.

1. Eric Peterson wrote an article on audience engagement, explaining his work with comScore, which made me revisit his paper on visitor engagement.
2. Google Analytics released a new feature, Advanced Segments.

The paper proposes a formula for visitor’s engagement that considers several components

As I posted before I wasn’t quite comfortable with the proposal and after some thoughts I finally realized why. The main two reasons are:

1. It puts all in the blender (weighted average of factors)
2. It mixes conversion, “the holy grail of engagement”, with the other terms

I proposed an alternative approach to analyze your visitors engagement that consider three axes Awareness, Amusement & Addiction each of them is build from factors considered in the original paper. Each of these segment can be computed using GA’s Advanced Segments feature, as well as the intersection of any two and the three segments.

Aware visits:

Visits that find your site through brand keywords or write your site’s URL directly on the browser or click on your loyalty campaigns (newsletters, subscribed RSS feeds), shortly visits that already know or are Aware of your site and brand.

Amused visits:

Visits that take their time on your site, that are interested and Amused by your content and product offer. The advanced segment is defined as visits how viewed at least a certain number of page views and spend at least a certain amount of seconds on your site.

Addicted visits:

This segment traces visits from users that come regularly to your site or are Addicted to it by specifying a minimum amount of visits and a threshold on the visit’s recency (the day since the last visit).

Using the advanced segment feature we can know the contribution of each segment (Aware, Amused & Addicted) to visits, page views, conversions and revenue allowing us to understand our audience distribution and which visitors segments are driving revenue.

Additionally since we are using 3 segments we can actually plot the overlaps using venn diagrams for any metrics we choose (visits, page views, conversions & revenue).

Venn diagram for all 3 segments visits and overlapping.

I really like this analysis because it allows you to see which traffic is driving the most conversion (separating conversion from engagement). Additionally you could have three levels and different flavor of engagement. Visitor on visits that match aware and addicted but not amused, visits that match all three criteria and so on.