Practical Analytics by DigiKnow

Google Analytics: Accounts, Domains, and Profiles (oh my)

Over the past week, I’ve had several clients ask me about the best way to configure their overall Google Analytics setup to cover multiple sites and domains.  I wanted to take a few minutes today to outline the options, what factors determine their use and how to setup the right solution.

Within Google Analytics there are three tiers for grouping your data: Accounts, Domains and Profiles.

Accounts – analytics accounts are the top-level container for all other Google Analytics data.  For the most part, unless you are an agency, like DigiKnow, you will only have access to one account.  Account = Company.

Domains – a domain is generally an individual website and more specifically is a unique Google Analytics tracking code.  For the most part every domain or sub-domain your company has, will be its own domain in Google Analytics. Domain = Website.

Profiles – a profile is an individual view on a domain’s data.  This one is a little more tricky.  For the most part, if you don’t have someone on staff or your agency that is intimately knowledgeable about Google Analytics, you will just have one profile that has all your data.  There are nearly infinite ways to create views on your data and I’ll cover a few of them in a minute.

One of the big challenges when you have multiple websites is whether it is better to setup multiple domains or multiple profiles within a single domain.  Generally, I recommend creating a unique GA domain for each website you have.  The one exception (which I ran into this week) was where it was technically difficult to use a different tracking code on each sub-domain.  My client didn’t have an easy way to put a unique code in each footer, so having a single domain split into profiles for each sub-domain was the better solution.

So how do you use profiles effectively?  There are some best practices, but ultimately it comes down to business needs.  For everyone, I recommend two profiles be created right from the start.  The first is an unfiltered profile that has all the collected data and the second is a profile that has your own traffic removed via IP filtering.  This lets you separate out what you are doing from what “real” customers are doing, but still gives you access to everything should you need it.  From there, the possibilities are nearly endless, but let me give some examples of what we’ve done.

• For a local public institution, we setup profiles with and without visitors on the local wifi and publicly available computers.  This let us see how visitors used the website differently in the building, the workplace or at home.
• For a national CPG company, we setup a profile to treat a section of the site as it’s own website.  This let us see the interactions with the site in a narrow view. Data like page views, bounce rate and referring sites give a greater accuracy than with an advanced segment.
• For a state university, profiles were setup for each of the individual colleges to allow access only to data for their area of the site. 

Profiles and filters can create a near infinite views on your site data.  Hopefully this serves as a good primer for the setup of a set of domains and profiles within Google Analytics.  I’ll continue to expand on this in the future.

Questions from the field

DigiKnow continues to offer our analytics seminars that we started late last year.  Since we just had our latest one yesterday, I figured I’d take a few minutes today and blog about the questions that came from our attendees.

Question 1: We moved from AWStats to Google Analytics and our data doesn’t line up to compare year-over year.  Can you explain this?

Matching data between various analytics systems is always a tricky thing.  Moving from a log based system (AWStats) to a tagging based system (Google Analytics) is even more confusing.  Log based systems track every interaction on the website, including image views, PDF downloads, etc.  Tagging based systems only track page views, unless specific code is in place to track downloads.  If one system is considering PDFs as page views and the other is ignoring them, you will have a discrepancy in your data. 

A major area of discrepancy for this attendee was in their traffic sources.  Although I don’t remember the exact numbers, AWStats was reporting something like 30% search, 30% direct and 40% referrals whereas Google Analytics was reporting 60% search, 20% direct and 20% referrals and 2% other (I realized these don’t add up).  This is significantly different and as it turned out was due to two main factors: First, AWStats was reporting alternative domain names (ex: site is xyz123.com but abc987.com also gets you to the site) as referring sites, dramatically boosting the amount of traffic from referrals (I think it was 15K vs. 5K between systems).  Secondly, Google Analytics was tracking banner and paid search campaigns in the “Other” category, whereas AWStats was putting that either within direct traffic, referral or search.

 

Question 2: On our Ecommerce site, my development team said we can’t use Google Analytics because of the way the pages are structured.  Is that true?

This one is a little less obvious since we weren’t looking at the attendee's website, but given what she was describing, it sounded like they were able to get the data into Google Analytics, but it wasn’t displaying anything useful.  I’m speculating that, because this was Ecommerce, they were including a transaction ID value in the URL string, so that each visitor’s pages were unique.  Visitor #1’s visit to www.xyz123.com/product?id=4564&transid=1 and Visitor #2’s visit to the same page looked like www.xyz123.com/product?id=4564&transid=2.  They are the same page, but to Google Analytics, looked completely different.  The solution I suggested was to apply a Google Analytics filter to the profile that would strip off the Transaction ID (“transid” in my example) and then the two pages would look the same.

Question 3: To install a tagging based analytics packages, to I have to edit every page on the website?

This one is a little easier and the answer is yes.  But what marketers not familiar with HTML and server side scripting don’t know, is how includes work and how easy that can make update every page on the website.  An include is a sub-set of the full page HTML that is stored in a separate file for reference by a page.  A single include is often used as a footer for the entire website, so a single update can change the entire site, making adding a tracking script just a few minutes of work. 

 

If you have any questions that you’d like to see answered on this blog, please email analytics@digiknow.com and we’d be glad to help.  If you are in the Cleveland area and want to attend one of our seminars, please feel free to email about that too.

Custom Variables – Tracking more than just visits

Analytics has come a long way over the past ten years.  We started with simple hits and visits from daily log analysis, then moved on to real-time tagging and more “modern” tracking methods.  Along the way we’ve gained the ability to segment, test various ideas, and filter out the noise from our website data.  Today, we have the ability to add data to website visitor activities to gain another level of insight.

When web analysis started, the only data we had access to was that coming from the web browser stored in server log files.  This included the pages/files requested, computer information and visitor IP address.  Even if the visitor had logged in and we had a complete profile on them, we couldn’t do anything with that data.  Today, thankfully, we can insert these additional data points, such as demographics (age, gender, etc.), preferences (heavy user, light user) or even memberships levels (non-member, basic, premium), into the data stream and use this data in our analysis.

Each analytics package has it’s own way of handling custom variables, both in the number allowed and the code-level configuration.  Below is a video from Google about how to handle this in Google Analytics (fast forward to 11:21 in the video – audio quality is not great).

One custom variable implementation that DigiKnow did that I think is a little outside the box, is using the page-level variable to track 404 (page not found) errors on a website.  This may be a little more technical than I usually get here, but to explain what we did, the line of code in red was added to the 404.html page for the website.  The parameters on the “setCustomVar” call are indicating: slot/key #1, Category of variable, Value for category and that the variable is page-level (3).

<script type="text/javascript">
var gaJsHost = (("https:" == document.location.protocol) ? "https://ssl." : "http://www.");
document.write(unescape("%3Cscript src='" + gaJsHost + "google-analytics.com/ga.js' type='text/javascript'%3E%3C/script%3E"));
</script>
<script type="text/javascript">
try {
var pageTracker = _gat._getTracker("UA-XXXXXXXX-1");
pageTracker._setCustomVar(1, "Error", "404", 3);
pageTracker._trackPageview();
} catch(err) {}</script>

What results from this is new content in the Custom Variables section (under Visitors)

More importantly, now I can setup custom report “Pages with 404 Errors” and a segment “Visits with 404 Errors” and see how and why visitors are getting to pages that don’t exist. 

If the technical details here are daunting, don’t worry, there are experts to help with that.  Just knowing that you can bring customer profiles or other custom variables (like the 404 example) into your analytics data should get your mind going about what else you might want to track.

Google keeps the new features coming

Google delivered an early Christmas present on Monday announcing some new features that will be coming out soon.  You can read the full text at Google’s Analytics blog, but I wanted to both make you aware and give you my take on the new features.

Annotations

As I said in the comments on Google’s post “Hallelujah”.  Because “why” is such an important part of analytics, being able to record the cause of a spike (or dip) is very important.  Until now, I’ve had to manage that in Excel or in a screenshot with annotations in PowerPoint or Word.  Having those in separate sources meant that as time goes on, my records got messy and it was hard to overlay my info on the GA data.  Annotations fixes that.  You can now go into GA and on the timeline, make a note of what happened that day.  Because we have multiple agencies working together and multiple analysts even within DigiKnow, this is a great feature for sharing knowledge.

Although none of my logins have access to this new feature yet to try it first hand, from the video, I really like how they are handling sharing.  Notes can be public to the profile or private to the user.  For every other GA customization (reports, segments, intelligence) you have to share the information via link to other users, but with this, it’s just there for everyone.  I hope that they roll out this system to those other customizations AND leave in the cross-profile sharing too.

 

Custom Variables in Segments & Custom Reports

While I think annotations are for the masses, custom variables (launched in October) are definitely for the more advanced GA users.  I haven’t really talked about this yet (or played with it), but until recently, you could only have one custom variable per domain.  So for a client that I wanted to track two things, I had to jam it into a single field.  With unlimited custom variables, you can track anything without having to be cautions of “wasting” the variable.  Some great things to track are registered vs. non-registered users, previous purchasers, visitor profiles established through questions/surveys, etc.  This is very powerful stuff.  Having the ability to segment by and report on these data is an obvious next step.

Code Setup Wizard

Google keeps adding features that makes configuration of complicated things simpler, putting me out of a job.  Good.  Although setting up multiple domains, sub-domains, etc. is definitely an art, I’m all for a wizard that just makes it happen.  There are parts of analytics that should and do require deep thought, but configuration shouldn’t be one of them.  Just like the GA URL Builder (for tracking campaigns), having a form or system to make setup easier saves time that can be better spend on more meaningful tasks.

 

There is also an announcement coming this week about API updates.  Probably to give access to these new data points.  I’ll update this post with a link to it when the announcement is made, but I don’t think any of you will probably be lining up at the Googleplex to hear that announcement.

UPDATE: Here’s the link to the Google Analytics API updates.

 

So what’s still missing?  We’ll plenty could be added, but the one simple thing that I’m dying for them to add is an import/export or sharing feature for Filters.  Given that we are now able to share custom reports, custom insights and advanced segments, it seems like only a matter of time. 

Google Analytics Goal's Expansion

This week Google made another expansion in Google Analytics allowing additional goals to be created within a profile.  This is a tremendous change because previously there were only 4 goals per profile, now you can set up to 20.  Before this week, if you had 5 goals you wanted to measure, you had to setup two profiles, which caused a lot of switching between profiles and created somewhat of a nightmare in keeping the filtering all the same, etc.

So now that you have 20 goals to play with, what should you be measuring?  Here’s my top 10 list of general purpose website goals to measure (in no particular order).

1. Newsletter registration
2. Contact Us form completion
3. Quote request
4. Online order submission (for Ecommerce)
5. Long visits [time] (over X minutes)
6. Long visits [pages] (over X pages per visit)
7. Content Upload/Submission
8. Printing a page (for PDFs or Printer-friendly HTML pages)
9. Downloading a Coupon
10. Visiting a target content page (product page, employee bio, etc.)

I’ve heard this feature is not rolled out to all accounts yet, so if you don’t see these expanded goals in your account today, keep checking and I’m sure they will appear in the next week or two.

Google Analytics – Turning an existing source code into a campaign

Measurability is one of the best aspects of today’s digital world.  This is probably most true about online campaigns.  We now have the ability to know which half of our marketing is working.  One of the problems we face is that we have multiple systems tracking campaigns, for some clients, even multiple web analytics packages. 

To get around this challenge, we’ve devised a set of filters to take a generic campaign source code (in this case for HBX) and format it for Google Analytics. 

Google analytics needs three components to make a complete campaign entry: a name, a medium and ad content.  So, our filters take the one piece of data we have, a campaign ad code, and create a properly formed campaign in Google Analytics.

To do this, we need a filter for each campaign field we Google requires:

Campaign Name Filter: Field A: Request URI = (.*)hbx=(.*) Field B: (blank) Output: Campaign Name = Unknown Campaign This filters looks for any URLs that contain the HBX source code variable and writes the word “Unknown Campaign” to the campaign name field.
Campaign Medium Filter: Field A: Request URI = (.*)hbx=(.*) Field B: (blank) Output: Campaign Content = Unknown Medium This filters looks for any URLs that contain the HBX source code variable and writes the word “Unknown Medium” to the campaign medium field.
Campaign Content Filter: Field A: Request URI = (.*)hbx=(.*) Field B: (blank) Output: Campaign Content = $A2 This filters writes the HBX source code to the campaign content field.  The $A2 variable represents the second grouping within Field A, in this case the second (.*) content of the URI.

 

This same process will work for any source variable that you want to put into a campaign.  Give it a try and let me know if this works for you.

 

Update: It should be noted that the Request URI matching is assuming that the source code (my hbx=XXXXX) assumes that the code is always the last variable in the query string.  I’ll write a more robust expression and update this again when that is done.

 

Technorati Tags: Google Analytics,filter,regular expression,campaign,source code,HBX,hitbox

Filtering Incoming Data to Improve Accuracy

The best part of digital marketing is the data.  One of the worst parts of digital marketing is… the data.  Any kind of digital analytics collects more data than can possibly be used and often the data coming in is not in the ideal format.  Fortunately, the web analytics tools from Google, Omniture, Webtrends, etc. have gotten much better in the last few years and they do a lot of the heavy lifting of data clean up for us.  But there will always be website specific tweaks that need to be made.

To parse through the mountain of data that comes through analytics, there are two “tools” that we can use to limit the data collected and/or our view of the data: filters and segments.  Filters are server-side tools that limit the data that is loaded into the analytics tools.  They are typically used to remove test traffic or clean-up extraneous data being collected in URLs.  Generally these are only setup by administrators, but with the ease of the interfaces today, anyone can do it.  Filters usually make permanent changes to the recorded data.  Segments on the other hand, are client-side filters that can be setup in a profile and can be turned and off to compare different groups within your visitor audience.  Typical segments include visitors from search, visits that completed a goal or visits from a campaign.  Segments are very powerful analysis tools and I’ll be talking about them a lot more in the future.

Today, I wanted to briefly talk about filtering and some effective filters I use on a regular basis.  My example code is for Google Analytics, but these filters can work on any analytics solution.  Before we talk about the filters themselves, I should probably back up and talk about profiles.  Since filters are destructive (they change the actual stored data), I always recommend creating multiple profiles for a specific domain.  In Google, this is done with the “+ Add new profile” link on the Settings screen.  At a minimum, I create a ‘Full Data’ profile with no filters and then a ‘Clean’ profile with all my basic filters applied.  So with that in place, these are the filters I apply to my ‘Clean’ profile.

1. Filter yourself out One of the most important things to do in your clean profile is to make sure that you (or your agency/partners) are not skewing the data.  The easiest way to prevent this is to setup an IP filter for your company and anyone you regularly work with that aren’t true customers/visitors. You can start by going to www.whatismyip.com and seeing what IP you are displaying to the world.  It may be as simple as loading that into Google or you might have to talk with your IT group if you find that your filter isn’t working.	This filter covers a large range of IP addresses, but yours may be as simple as the example on the page (63\.212\.171\.123)
2. Removing the Query String This filter is not for everyone, and this is another case where I might setup a extra profile, but if your site has a lot of extra data in the URL after the question mark that is making the same page look like multiple pages in analytics, then you might want to try this one. Field A: Request URL = (.*)\?(.*) Output: Request URI = $A1 This splits the page URL into two parts, the file name and the query string, then overwrites the full URL with just the first part (the file name/path)
3. Clean up Email Provider sub-domains If you do any amount of email marketing, you’ll see that in your Referrers to the site, a lot of different domain names all from the same provider (us.mc317.mail.yahoo.com, us.mg3.mail.yahoo.com, us.mg201.mail.yahoo.com, etc.).  These are all Yahoo mail, but they aren’t nicely combined into one line.  I use this simple filter to strip off the extra sub-domain information for the top email providers. Field A: Campaign Source =  (.*)(mail\.yahoo\.com|webmail\.aol\.com|mail\.live\.com) Output: Campaign Source = $A2 This spits the referring URL into two parts, and drops off the sub-sub-domain that varies for each email provider.

 

There are many other filters that I use, but these are some of the basics.  I’ll post additional recommendations on more advanced filters in future posts.

 

Technorati Tags: Google Analytics,filter,query string,regular expression,email service provider,IP address