09 okt2011

Web analytics on “template” and “element” level: that’s what UX needs

  • reageren (7)
  • bookmark
  • e-mail
  • afdrukken
X
X
  • Plaatsen/stemmen op NUjij
  • Geef kudos

I’ve always wanted to answer a few specific questions about the behavior of users on my website. But the answers that analytics tools gave me were always incomplete and unusable, usually coming from an outdated marketing perspective.

From an analytical or UX point of view, a website does not consist of users that interact with pages that have content. Instead, it consists of user segments that interact with templates (page types) that have elements or blocks of content (or content types, interactions, you name it) on them.

So I hope I can explain the types of analysis I want to make. Let me know if something is unclear or if it’s just wrong or already available right in front of me. Here’s my braindump:

Problem part 1: abstraction level in tracking

So I am dreaming of a tool that can do this:

  • Track user segments by letting me segment after tracking based on variables like traffic source, behavior, etc. Google Analytics does a great job with their advanced segments.
    Example: users coming from Google’s organic listings on queries containing the keyword “hotel”
  • Track templates by grouping URL’s (with regexes, after tracking of course!) or by setting templates manually (before tracking) to avoid grouping problems. A site that has good SEO and information architecture almost by definition has a URL structure that does not convey the page type / template in the URL.
    Examples: a category page, or a product page, or maybe a specific type of product page.
  • Track elements, by looking at the HTML structure (DOM) and letting me define the elements / blocks that I want to track.
    Example: the sidebar, or the block in the sidebar that contains the nearby hotels, or the 1st list item in the list of related hotels.
Of course all of the grouping (to abstract) and zooming in (to make things granular), should be done after the tracking. So you can learn from your data, go back and explore new relations.

Problem part 2: useful analysis of  behavior (click paths and correlations!)

Web analytics software is pretty good at segmenting users. But it’s too granular at the page  (URL) level and largely ignores the elements on the pages. More granular things like custom variables and events in Google Analytics are connected to users, not to content and therefore cannot be analyzed in a useful way (i.e. in relation to eachother!).
So I want to analyze the effect templates and elements have on user behavior. Click paths are probably the main tool for this, but the element level analysis makes it a bit more complex. Some example questions I want to answer:

  1. What % of users from Google lands on the product page template instead of the category page or the homepage. Compare metrics on these 3 types of “userlandingpagetypes” ;)
  2. To which templates do these users go next, and which elements do they use to get there?
  3. Of the users that click a “nearby hotel”, how many click the first, second, n’th result? (I want to be able to drill down when I see something interesting)
  4. Of the users that land on a product page, which % used the keyword “hotel” in their query this month in comparison to last month?
  5. Of the people that end up seeing the booking page, how many have used the related hotels feature directly after they’ve landed, and how many used it later?
  6. What’s the conversion rate for each category on my site: how much overlap is there in user behavior (i.e. do they start in 1 and convert in another?)

Number 1 and 2 are already fine fo 80% of the important learning. the others are variations that you can of course broaden with tons of examples.
I think the perfect tool would let me visualize the 3 data types of tracking, and let me drag and drop them to create the relations between them that can then show me the metrics when I want to analyze the visualized “path”. Of course always giving the option to drill down to look deeper and make more specific analysis…
And as a bonus I would like to segment users on their interaction behavior on the templates and elements, but I’d like the software to do this with data mining. So give me examples of the effect a certain page element has on conversion for a certain group of users. And I guess we’re all drooling if we then let the website optimize itself based on these metrics, right?

Possible exising solutions and a mockup of my own

Of course, all of this is already possible. But the tools I have seen were not doing right. So let me know how you think I should use which tools to reach my goal of tracking on segment, template and and element-level. The closest I came was using Google Analytics’ custom variables for setting page templates. But they have several issues like the amount you can set (5, officially) and the fact that you can’t track e-commerce metrics with them.

I made a very low density mockup of what this tool could look like when you’re building reports, I wonder if it’s understandeable for anyone but me…

 

Popularity: 1% [?]

Reacties (7)

Reageren
  • What about BTBuckets for segmentation?

  • Great article Ruben.

    One possible solution is to create an extra GA tracker that only tracks page templates in stead of url’s. On those pages you can use events to track on page element. It’s possible to create a custom report where you first select the dimension “page” and below it the dimension “event action”. That way you can see which events happened on which pages.

    You can use an external tool like paditrack to create funnels and paths based on any segment you want on the fly.

  • Andre, great suggestion. I first thought of just adding an extra pageview on each page to contain template information and then exclude those pageviews with filters. But since we rely so much on Analytics (via the API for reporting to our clients) I really don’t want to risk anything like this….

    But with the async analytics it might indeed be an idea to just add another profile with these special pageviews…

    I’ve found Paditrack unbearably slow so I shy away from them for now (we do use PadiAct for getting more mail subscribers).

    I’ve trie BTbuckets before but no clue why I didnt continue with it, I’ll try again ;)

  • Thanks for posting your braindump. Before reading the comments my first suggestion was: include an extra virtual pageview on template level and then filter all other traffic out in the profile you create for these measurements, but I see you already thought of that. Why are you afraid of doings this? You will not make any changes to existing profiles other than filtering out the virtual stuff. But if you are scared, setting up a second counter like Andre is suggesting sounds like a good idea.

    What I would like to measure on a website has to do with the fact that behind that browser there is a human being who is ready to interact with me. I just need to offer the right dialogue entries (based on my wishes for this person). What needs to be measured is: which dialogue entries (and continuation dialogues) are offered and what succes rate do they have.

    This means that my webanalytics tool not only has to report about element offers and use on certain templates but also about the predicted dialogue reason of (a group of) users (so we can measure dialogue improvement). The combination / integration of profiling / tagging / testen / analytics :-) and yes.. please! with a awesome UI :-)

  • Ton thanks for your thoughts… I missed you last night at UX University at Usabilla’s office, there was a great presentation from the guys of Persuasion API. I think they could fulfill your dialogue-optimization dreams ;)

  • I think they only have a small part of the puzzle. Dialogues are much more than just persuasion… By the way, I much drinking Oktober beers in Limburg :-)

  • Thanks for sharing Ruben! Really great to read this article.

Reageer

(zal niet zichtbaar zijn)