Thursday, July 24, 2014

3 Advanced Web Analytics Visitor Segments: Non-Flirts, Social, Long Tail

Occam's Razor
by Avinash Kaushik


3 Advanced Web Analytics Visitor Segments: Non-Flirts, Social, Long Tail

SunshineThe last blog post shared custom analytics reports that you can use to find amazing insights faster, enabling you to create a focused, truly data driven organization.
In this blog post I want to continue the let's help make your day-to-day life better path. I'll share three advanced segments that I personally find to be of value in the process of moving from data to actionable insights. I hope you'll download and use these segments, but more than that I hope you'll learn how to create delightful analytics segments with the options you have at your disposal.
I am an unabashed segmentation fan: Web Analytics Segmentation: Do Or Die!
Without segmentation our analysis is focused on unrecognizable blobs of traffic. Total Visits. Average Page Views Per Visitor. Overall Conversion Rate. Yada, yada, yada. Boring. Useless. Life wasting.
With segmentation we focus on groups of people and we focus behavior that has logical connections (everyone who used a particular keyword, group that came via Twitter, people who viewed a TV ad, visitors who saw more than 4 pages on our site etc., etc). That helps us understand data & performance better. It helps us get data-gasms, improve ROI for our web efforts and get our bosses promoted.
How can you not love that?
Below are examples of segments that help us make a lot more sense of all the data we have and the insights that await us. You'll be able to download these segments and import them into your Google Analytics accounts and start using them right away!
Additionally, as I often do, you'll learn lots about the types of delicious analyses you can do with these segments. For good measure there is also a tutorial on regular expressions at the end (no good analyst can live without regex!).
If you use Adobe's Site Catalyst or CoreMetrics or Yahoo! Web Analytics or WebTrends or. . . you'll have enough detail below to create a segment in 5 minutes in those tools as well. Trust me, it takes just 5 minutes and, like with Google Analytics, you won't need to update your JavaScript tags or have to do extra work with IT or buy other expensive versions of their products just to do segmentation.
Let's go. . . three awesome analytics data segments. . .
#1: Non-Flirts, Potential Lovers
Did I get your attention? :)
We all obsess with our bounced traffic because it seems nutty that the person you spent so much time and love attracting to your website bounced! They did not click to see another page. They did not hit play on the video on the landing page. They did not click on a link on your landing page to your corporate site. They just left.
Here's how that segment looks:
web analytics segment bounced traffic
It is tempting to analyze these people. Where did they come from? What campaigns? What landing pages? Etc., etc.
You can find value, but to grow the business it is not prudent to focus on analyzing just the people who flirt with us.
Why not first analyze people who do engage with us?
At this point people switch to analyzing all the non-bounce traffic. This is how that segment looks (bottom right):
web analytics segment non bounced traffic.png
[The above is a standard segment in GA, just look under Default Segments.]
Better. Ignore the flirts. Focus on everyone else.
Unfortunately that is still a "blob." It includes anyone who just had two "hits" in their visit (hits is a technical term for a page view, event, custom variable, etc., etc., more than one hit = non bounce visit).
I want us to be a lot more deliberate.
Look at the Depth of Visit report (standard report in GA in the Visitors section). It shows the distribution of the pages people see on your site (not the "silly" metric, average page views per session).
The distribution will show you the "tipping point," the point at which a core group of people decide to stick with your site after overcoming their initial "fears" (and your perhaps sub optimal pages!).
Segment that.
To use a metaphor. . . look for people who made it with you to a third date. For many sites, but not all, that's people who have seen three pages. It might take 14 pages to buy, but if they stay to three they are giving you a chance. They might read 8 stories on your non-ecommerce content site, but you note that people who see three engage for a longer time.
Here's that segment:
page depth engagement analytics data segment
So simple right?
These 7,610 Visits were ripe with promise. Some people ultimately ended up buying, others just gave you a chance and decided not to consummate.
Rather than focusing on the bounce traffic ("flirts") it is much much more interesting and valuable to initially focus on people who give you a chance.
Where did they come from?  segmentation engaged traffic sources
12.11% from Organic Search via Google. Enough? Not enough?
More questions for you to answer. . . .
What pages did they enter on? What campaigns have a higher percentage of these people? What countries? What keywords? What is the delta between content they consume on your site compared to everyone else?
Look at the row with % of Total. . .
nonflirttrafficcontentconsumption
Helps you find what they are interested in, right?
More questions to answer. . .
Do they all happen to use the comparison chart first? Do they all absolutely read the Sports section? What's so unique about them?
This an astoundingly simple segment to create. Yet analyzing visitor behavior for this segment helps you identify, and perhaps do more of the things you already know are working.
Do this first.
Here's how you can get this sweet and simple segment:
  1. Log into Google Analytics.
  2. Come back here.
  3. Now click on this link: Non-flirt Potential Lovers Segment. It will open in GA.
  4. Click on the Create Segment button to save it in your account.
If you want to share this report with others (say via Twitter / email) you can use this url: http://goo.gl/SuwKp
Have fun.
#2: Social Media, Baby!
Social media is all the rage. Suddenly Marketers have discovered that convincing people to buy their products/services or read their content or apply to university takes just two things:
    A. 140 character missives sent frequently during the day extolling the glories of the company / newspaper / university
    B. Creating a Facebook page, and then proceeding with the glory extolling
So easy. </ever so slight sarcasm>  : )
Our job is to hold the feet of these adventurous people to the warm accountability fire, right?
[Remember everything below is only if you use Twitter, Facebook et al for pimping. If you are participating in those media in the manner in which you are supposed to, conversation and adding value rather than pimping, then I encourage you to read my Social Media Analytics post to learn what the best metrics and tools are.]
The challenge in measuring social media impact on your business is two-fold.
    1. Most content gets consumed in applications (think tweetdeck, my beloved twicca, mobile etc). They don't send referrers allowing us to tie to the source with our analytics tool (any tool, GA or Omniture or CoreMetrics).
    2. Splitting out activity that we caused vs. activity that was caused by others.
My recommendation is, again, two-fold.
First, if you tweet / update / tumble links back to yourself then please for the sake of all that is holy in the world add campaign tracking parameters.
Here's the link I tweet:
http://goo.gl/myisj
It points here:
http://www.kaushik.net/avinash/2010/12/best-downloadable-custom-web-analytics-reports.html?utm_source=social-media&utm_medium=twitterfb&utm_campaign=aktw
See the campaign (utm_) tracking parameters? Trackability, sweet, trackability!! Mobile, apps, html5 pages, bring it on. All visits tracked!
[If you use Site Catalyst or WebTrends or Yahoo! Web Analytics your campaign tracking parameters won't look like the above. Check with your vendor and tag appropriately.]
Second, split your social media value analysis into two different segments. Activity caused by you and activity on your site by all social media visits.
self driven social media traffic segment
You drag over Source and input the Value you are using to tag your SM links, in my case the utm_source is imaginatively titled social-media.
This tracking mechanism (campaign tag) is used both on Twitter and Facebook links. I can, and often do, split out Twitter and Facebook separately by using a different value in the utm_medium value. I can further segment them separately if I want. For now I want to analyze them at a higher level together.
I did pimping. I got 2,486 Visits. So what?
Easy question to answer, go to your outcomes report and apply your newly minted segment:
social media conversion rates
Pretty darn pathetic, right?
Only one of the above goals is connected to a "hard" conversion (leads generated, Goal 2). The rest are "engagement" and videos played and other such goals.
Still pathetic, right?
Do you know how awesome, or not, social media efforts directly initiated by you are? It's not that hard. Go figure it out.
Oh and yes, you don't have to stop here. You can apply this segment to your amazing Page Efficiency Report, to your Visitor Loyalty and Recency reports, to your. . . well any report you have. That allows you to measure a broader view of the success of your social media efforts, rather than my effort to instantly put your feet in the fire! :)
Here's how you can get this social media segment:
  1. Log into Google Analytics.
  2. Come back here.
  3. Now click on this link: My Social Media Traffic Segment. It will open in GA.
  4. Click on the Create Segment button to save it in your account.
If you want to share this report with others (say via Twitter / email) you can use this url: http://goo.gl/fvuXP
Time to create our second, more expansive, social media segment.
As I had mentioned above, this time around we'll look at the social media to our website(s) from our efforts as well as that of all our friends / BFFs / haters.
Before you create this segment you should go checkout your All Traffic Sources report and see how your web analytics tool is capturing various channels people show up from. Based on that review of my site, here's the segment:
analytics segment all social media
A quick explanation.
Values for twitter, facebook, sphinn, stumbleupon are there for people who use web based versions of these social media websites. I can add delicious, digg etc., etc., if I want to. They are simply not that important a source of traffic for me. See why the review of the All Traffic Sources report recommended above was important?
[Some people will obsess and create a ginormous catch-all segment. But remember, you don't need to understand data from the last 10 visitors to make smart decisions.]
The value for "social-media" is there to capture the social media campaigns tagged by me. See our first social media segment above. You'll use your own tracking values.
Value for awe.sm is because for a while I was using awe.sm to auto-tag all my links. There are some latent visits which should get flushed out of the system in the near future (as I have standardized on www.goo.gl and www.bit.ly).
That's *my* All Social Media Segment.
If you are thinking: "Good lord that is messy!"
Welcome to the world of social media tracking. It is messy-ever changing-and you should know that you are going to babysit this constantly. Sorry.  [Also see comment above about needing the last 10 visitors: you don't!]
But after you create the segment, awesomeness follows. . .  analysis!
Step one: Answer: "So What?"
social media conversion ratesall sources
Better, but honestly still pretty pathetic. Remember the goals are a mix of hard and soft conversions (see above)!
By now I am never surprised when I see the above result for Social Media efforts of most outcomes-driven pimping efforts via those channels.
Perhaps you are an exception. Now you know how to measure it!
As mentioned above Analysis Ninjas won't stop at just Outcomes analysis and will dig deeper to see if there is any value that this traffic is adding to our company. My personal favorite place to start is Visitor Loyalty analysis.
Ok so these people are not delivering any hard or soft conversions. Does their loyalty profile look any different?
Here, check it out (standard report in Google Analytics and other tools):
visitor loyalty analysis social media traffic
Hmm. . . a very different profile from other visitors to the site.
Other traffic to the site has much less loyalty than social media traffic. See the delta between 60.98% and 44.35% in the first two rows? Also see the much better, sweeter, distribution for Visitors who visit from 9-14 times through 26-50 times.
For this content website there is value in the social media efforts in that they are delivering an audience that tends to then be much more loyal than all other traffic that ends on their website.
Provable value! From social media!! I know!!! :)
A couple more ideas for our Ninjas to dig deeper, and types of analysis they could do to determine other types of value.
It is trivial to measure the base metrics for your website for your Social Media segment. Visits, Pages/Visit, Average Time on Site, % New Visits, Bounce Rates, Conversion Rates. . . . and so on and so forth. . .
visits pages per visit avg time on site percent new visits bounce rates
You can quickly see at an aggregate level, or a detailed level, if your social media are delivering on the promise outlined by your $150,000 Social Media Consultant.
Here's another bit of analysis that can be useful for certain types of websites.
Say you have a real estate website, or you are responsible for craigslist.com. Both sites are primarily internal site search driven. People come to the site, search, find what they want, do business.
Take your newly beloved Social Media segment and apply it to your delightfully sweet pre-configured Internal Site Search reports. [Left nav -> Content -> Site Search]
Here's what you'll see. . .
internal site search analysis for social media traffic
You'll be able to analyze if people who come to your site from your Social Media campaigns engage with your site more or less (Total Unique Searches per Visitor). Do they exit from the internal site search results faster or slower (% Search Exits)? Do they have a harder time or an easier time finding the right result (Results Pageviews/Search and % Search Refinements)? And other such analysis.
You don't have to just report clicks and visits from social media. In our real estate website we got to the root of what's a deeper engagement (searching) and we got down to measuring real value (or lack thereof).
Ready to do some real social media ROI analysis?
Here's how you can get the all social media traffic segment:
  1. Log into Google Analytics.
  2. Come back here.
  3. Now click on this link: All Social Media Visits Segment. It will open in GA.
  4. Click on the Create Segment button to save it in your account.
If you want to share this report with others (say via Twitter / email) you can use this url: http://goo.gl/YnJON
Good luck!
#3: Search Queries With Multiple Keywords [3, 4, 5, 10, 20]
On this blog and in my keynotes I have bemoaned the obsession Marketers have with brand keywords and their sub optimal strategy of optimizing for keywords, rather than key phrases.
I am a search long tail lover. It is the way to happiness (and finding relevant users!). Hence our first segment focuses on helping you understand the balance between keywords and key phrases in the queries used by Visitors from search engines.
It is not actually a "segment," it is more like using advanced segmentation as a reporting engine in a way you can only do in Google Analytics!
My strategy is simple. Use a regular expression to get GA to segment search queries into various "words this query contains" buckets. Here's what it looks like:
searchlongtailwordssegment
"Magical" part: ^\s*[^\s]+(\s+[^\s]+){2}\s*$
Not that magical actually, just a humble regular expression. It is looking for the number of words in a query (in this case queries visitors typed into Google or Bing or Baidu that contained three words). The second regex counts visits with four word search queries.
[A quick note of thanks to Nick Mihailovski for helping me come up with the perfect regular expression. I was using ^\w*\s\w*\s\w*$. It was good but would not have caught some variations and it would not work for queries in non-English character languages.]
Ok back to using advanced segmentation as a long tail search report.
The final segment I have created, using the method above, has more "or" conditions that contain buckets for counting search queries with 3, 4, 5, 10, 20 and 20+ words. You can of course create any buckets you like; these were ones I find initially interesting.
When you click the Test Segment button (top right) you get this gratifying view (cropped to a small size):
search query words used distribution
Delightful right? It really is.
You get such an immediate sense of the long tail in a way that is hard otherwise in the mass of queries from search engines.
521 Visits from people who typed more than 10 words into Google/Bing! There were 36 visits by people using 20 words in their search query! And 237 people typed in more than 20+ words as their search query!
OMG!
Is your search engine optimization and paid search strategy accommodating for this type of behavior? You still bidding on a word or two?
While the above is not even your complete search universe view, it is a very simple and straightforward way to appreciate how long your search tail is.
And notice you did not even look at a report. You could do all of the above in the advanced segmentation view!
You likely want other buckets than 3, 4, 5, 10 , 20. No problem. Just download the segment below and make the appropriate changes and bam!
Here's how you can get this search long tail segmentation reporting:
  1. Log into Google Analytics.
  2. Come back here.
  3. Now click on this link: Search Query Length Segment. It will open in GA.
  4. Click on the Create Segment button to save it in your account.
If you want to share this report with others (say via Twitter / email) you can use this url: http://goo.gl/v3KbM
Being the Ninja that you are I am sure your thirst of knowledge is not satiated.
Now you are probably wondering how the bounce rate looks for one segment of the long tail traffic (lower usually) or how the conversion rate looks (higher usually) or how many pages do they see (more engagement usually) etc., etc.
The above segment won't help you with that. But all you have to do is create the segment you want.
For example here's the segment for people who see four words exactly:
segmentforfourwordsinthequery
Save the segment, here it is: Visits via Search Queries containing 4 words.
Now apply it to your favorite search report and hello sweet, sweet delicious data!
performancedataforfourwordsinthequery
You know the search queries, you know how many people came and you know their performance ("engagement" or conversions or downloads or leads etc., etc).
Furthermore, you can also segment this data by Paid Search and Organic Search, or Google vs. Bing and start to do very focused analysis that should fundamentally improve your search marketing program.
You can also take another slice at segmenting your search head, mid, and tail. For example you can easily create a segment for Visitors who came to your site via search queries that had more than four words in the query.
Here's that segment: Visits via Search Queries with more than 4 words.
Now go apply it to your search engine or organic search or paid search or goals reports and do really valuable analysis that will earn you the eternal love and adoration of your peers and superiors!
[SIDEBAR]
In case you wanted to do something more sophisticated beyond what's outlined above here are a quick set of instructions, and a tutorial on using regex.
If you want to create a segment for search queries that contain just one word use this regular expression in your advanced segment:
^\s*[^\s]+\s*$
If you want Visits with two words in Google search queries use this:
^\s*[^\s]+\s+[^\s]+\s*$
or
^\s*[^\s]+(\s+[^\s]+){1}\s*$
If you want to identify Visits by people who use three words in their search queries:
^\s*[^\s]+(\s+[^\s]+){2}\s*$
Now you can keep adding to the number in parenthesis and do a happy dance.
Some more cute things.
If you want to query for more than x words, say more than three words use this:
^\s*[^\s]+(\s+[^\s]+){2,}\s*$
Did you see the comma after the number two above? Good.
If you want to identify all search queries where visitors to your site typed 2 or 3 words into the search engine, use this regular expression:
^\s*[^\s]+(\s+[^\s]+){1,2}\s*$
Fun eh?
So what the heck are all those characters in these regular expressions doing? Glad you asked.  Let's consider the regular expression we used to identify 2 word search queries.
The expression is (identified above): ^\s*[^\s]+\s+[^\s]+\s*$
Here's an explanation (as best as I can express in lay terms). . .
^          start at the beginning of the line
\s*        match zero or more white space characters
[^\s]+   match at least one or more non-white space character
\s+       match at least one or more white space character
[^\s]+   match at least one or more non-white space character
\s*        match zero or more white space characters
$          end of string
I hope all this "magic" makes a lot more sense.
[/SIDEBAR]
Isn't advanced segmentation cool? And to think you did all this with your standard javascript tag, all on the fly (including historical data analysis) and without having to buy extensive expensive add-ons!
Ok it's your turn now.
What are your absolutely dearest advanced segments? What's the coolest thing you have done with the advanced segmentation capability in your web analytics tool? Care to share some of your favorites? Perhaps a downloadable link?
It would be incredible to have your wisdom help all of us. Please participate.
Thanks.

No comments:

Post a Comment