Hear – Listen – Communicate

Creating an analytics enterprise event model
February 27, 2009, 9:56 am
Filed under: Enterprise web analytics adoption, Measuring Performance

Per my previous post, I feel strongly that organizations, and the industry, have let go of the one metric that used to matter: Page Views.  While competition made the web-analytic vendor giants take the stance of “ease of development” rather than “slow and thoughtful planning” all those Web 2.0 events that were able to be measured are now lumped into two categories.  Pages and Events.  In many tools both Pages and Events are treated like page views. 

While this may seem obvious to those analytics guru’s out there who know to speculate and validate the data;
1) Betsy in HR sees the reports at face value.  When Betsy in HR sees her page loads skyrocket, she is just happy as can be and spreads the words.  However, the dev team just launched video tracking so every time someone hits Play, or Pause, or … it counts as a Page View. 
2) Anthony the Video Production Assistant wants to know whether we saw a lift after he split up the videos, but does not see any uptick at all.  Per Thomas’ direction the dev team tagged the events so that they appear in the Events profile, not the Pages one.
While we all try our best to communicate our data as clearly as possible, holes happen.  Before they do create your Enterprise Event Model.
What is the Enterprise event model?
In order to assure your Enterprise has data that can be segmented, cross-examined, aggregated, loaded into a DB, etc it must be “normalized” and “cleansed”.  When you deal with desperate pieces of data, which our “URLS/Events/track-able units” have become, you must categorize the data somehow so you can do something with it.  The Enterprise event model helps start the communication around the company as well as unifies your data into various categories.
Example Interactive Event Model

Example Interactive Event Model


Who needs an Enterprise event model?
Do you work for a multi-national, multi-branded, multi-productline, multi-service company?  You need one.    Do you have a presence on facebook, youtube, twitter, myspace?  You need one.  Do you have multiple games, experiences, mini-sites, blogs that you manage?  You need one.
With all these interactive pieces floating around, it is hard to see from a birds-eye view what is truely working.  We can look from a micro point of view to see how each individual campaign, or subsection, or funnel performs, but what about in comparison to other campaigns?  Beyond the CPA.
Why do I need an Enterprise event model?
All interactions on the web do not equal 1.  You know this, it is why we score things, it is why we get more excited when people convert than simply visit.  That said, neither do all those things you are starting to click.  However, this may be masked in the way you are looking at them.  Without a structure as to how to equally compare different interactive events, we can never truely see how it effects our bottom line. 
What does the Interactive Event Model include?
We have been segmenting our visitors like crazy lately:  Paid, Organic, New, Returning, Bounced, High Revenue, etc.  What about the content itself?
Simply put:  the Interactive Event Model (IEM, cause we need a new acronym) includes all things that can be tracked.  Perhaps you are not tracking them today, but eventually they will make it in your roadmap and you will need to go through these steps.  Get as far as you can now. 
This post is by no means meant to be a catch all solution.  The examples given here should be used more as a framework for your own company.  When working with clients I start here, but normally we think of more appropriate categorizations based upon their geographical, revenue, organizational, political and many other variables.  Who said this was easy?
Step 1) Identify your touch points
A. URLS – Now your site may have thousands even millions of “pages”.  What we are doing here is taking an inventory of assets.  You may have 10,000 knowledge base articles.  Knowledge Base articles are one asset.  That said, if your company has several brands, and each of those brands has a support site.  Then each brand should be listed individually.  Most likely you have also started one or a dozen micro-sites.  List those pages as well.  The Landing page, the about you page, the pitch page, the form page, the thank you page.   
B. Events – as stated, your online brand is much more than the pages on your website.  Before we get too crazy though, start with various actions users make on your site that do not spawn off a new “page”.  This may be downloading a PDF, hitting the play button on an embedded video, commenting on a blog, using dynamic navigation, clicking on the social media buttons (add to twitter, facebook, etc, etc).  This is best done with a whiteboard, projector, beer and pizza.
C. Offsite Events – Now let’s get crazy.  Where I will admit that I was ignorant at one time to say your website is your hub, that is no longer the case.  In fact, in some cases, it behooves you as a brand to branch out and start to own real estate on the various public (or social) properties out on the web.  While it may be a little difficult to embed the webtrends.js file into your facebook page, you know you are counting the number of friends, comments, etc you have.  In only a matter of time that data will be available into WebTrends, Omniture, Google Analytics.  If Google has it their way, you can feed this all into Google Analytics, just ask @anilbatra and use his **Google Analytics URL Tracking Tool**.
D.  Rest for now, but something tells me the list will grow.  Can you think of anything else?
Step 2) Identify your Taxonomies:
You must identify the various taxonomies for the dozens, hundreds, thousands of touch points your users utilize to interact with your brand.  That interaction, is what needs to be modeled.  Try to fit everything into 10 categories.  If this is too challenging, perhaps you need more than one taxonomy. 
While the topic of scoring has not come up in a while (WebTrends, time for some more PR!), it is becoming more and more important.  In order to start scoring though you must go through this normalization stage.  How do you normalize disperate Web Site Pages, Campaigns, Applets, Social Media Interactivity?  Build Taxonomies.
You can start here:
Content Taxonomy:  Some of the more popular web analytics applications out there allow you to create content groups.  Many people simply create groups equivalent to their top navigation.  Wrong approach.  Your navigation was built for users, tracking is for intent.  Your conent taxonomy should be broad based.  Think of this question: 
What content does this asset contribute to?
What organization controls this asset?
Intent Taxonomy:  Beyond the content taxonomy, each of your pages has an intent, otherwise known as the GOAL of the page.  Do not confuse this with what you want people to do FROM the page (that is probably an event) but what THAT page does.  While the goal of your product page is of course for them to engage or convert, the product page itself may still just be an interaction.  Events such as clicking on the zoom, comment, rate on the other hand may be engaging.  Starting to get why this is all importat?
Tip to get started, ask the closest sales guy to hand you their “sales process” PPT slide.  The one that starts at the initial contact, through engagement, into convincing, then conversion, then relationship building …. etc. 
This is a LOT of work, do I really need it?
Perhaps not now, but I know you will later.  Chances are you are already starting to face some of the issues mentioned in this post.  Many times these issues do not come up till the very last moment.  The issue is also hidden till the very last moment … “OK, I have all the campaign data, let’s see how they really measure up”.  Sure we can see which was more profitable, more conversions, more leads.  However, if you look at all the campaigns you launched last year.  Which was most “engaged”?  What is your avg bounce rate according to page intent?  Just like you segment visitors, it is time to segment content too.
Fake Real World Example:
Company:  IAC/InterActiveCorp
(Full Disclosure: I have nothing to do with them.  Mr. Diller, if you like this email me.  I am using this example because it is the most over-exagerated example I could think of.  I mean match.com to hotwire.com, that is brand diversity)
While I am sure the dow jones is a little higher on the scale, think about how Mr Diller’s staff calculates engagement across the portfolio.  Think of how difficult this can be with sites such as: Ask.com (Search/Content), evite.com (Social/Media/Lead Generation), match.com (Dating/Media/Lead Generation), shoebuy.com (ecommerce), zwinky (RIA/Media).

It would look looking nothing like but contain data to the charm of:

IAC properties from compete.com

IAC properties from compete.com

  • Looks like match gets the engagement award this year. 
  • I am fairly surprised that an engaging (or at least meant to be) kids game site only gets 5.7 pages. 
  • I would also be somewhat concerned that my ecommerce company is engaging (or possibly confusing) users more than evite.com. 

The point of that example was really not to tear apart the data, I am not close enough to the data to do so.  I wanted to bring to light how important the definition of “Pages” is on each of your sites.  I am sure IAC has done this.  (if not, call me)  Now my other guess is that compete.com measures Page Views in a manner that gives lots of credit to all those profile views in match.com (also validates how hard it is to find that love interest!) but only one page view upon launch of the zwinky app, losing all insight into each time I change hair color, bikini, or other attributes. This my friends, is the cornerstone of the event model.  


4 Comments so far
Leave a comment

I remember the first time I discovered Page Views could be specifically defined in a WT (log file, 1998) report. And that was a revelation, because I realized how important it was to know precisely what I was reporting on.

This led to creating a “clean” (HTML only) profile and a “dirty” (everything) profile just to see how different they were. The comparison of the same report in each of the two profiles became a standard practice for me when setting up a new client, it revealed “gotchas” you couldn’t even imagine.

With the amount of new interaction types around today, I can’t imagine not having the “enterprise event model” discussion before even looking at a report! What good is it to view a report when you have no idea of what it is you are really looking at?

Comment by Jim Novo

I just came back from a short vacation (in Jim Novo’s neighborhood, hi Jim!) and it’s terrific to have something this meaty as my first catchup item. Events, yes! Taxonomy, yes! Intent, yes! Collapse the data into meaningful groups, yes! Drilldown to infinite detail, no!

I want to come back to this item and say lots more. Keep writing about this kind of thing, Bos.

Comment by Chris

Very well done! I completely agree that preparation in implementing a web analytics solution is the key to success. Otherwise, people merely collect data. Data collection is the easy part…applying it and making it actionable is the useful part!

Comment by Eric

Well done! Enterprise event model takes us from technical/ mechanical site measurements (page views, clicks, referrers, …) to online business analysis. Unfortunately, so much of what is being measured, out-of-the-box, in WA tools today is geared toward the mechanical measurement. It would be great if event identification was part of the tag. That sounds utopia right now.

Comment by Steve

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s

%d bloggers like this: