Sep 18 2008

Why Do We Care About Path Analysis?

Published by at 6:46 am under analytics

Your answer should be, “we don’t.”

Path analysis is bunk. Why do analytics vendors spend so much time on path analysis reports and tools? Taken in context, *some* of the pathing tools can be useful, but giving us a tree diagram of complete paths is completely useless, in my opinion. What does it tell you? Absolutely nothing.

What makes it even more frustrating is how *pretty* the darn tree diagram reports are. Never show this report to anyone! Otherwise you’ll be getting requests for path analysis reports and will end up wasting your valuable time when you could be analyzing your 404 or most popular page reports!

Do you have conversations with your colleagues that go something like this:

Them: “Can we see how people are navigating through the [enter degree name here] section of the website?”
You: “Uh … (knowing full well where this is going) … yes … can you give me more specifics?”

Them: “Like the paths they are taking.”
You: “What exactly do you want to know?”

Them : “If they’re actually reading our application instructions before filling out the application.”
You: “Sure, we can do that.”

And then you say to yourself, “there are other ways to look at that, not path analysis.” Good, good. Go with that!

The trick to using alternative reports to get at your “path analysis” questions is to focus on *exactly* what you want to know. In this case, if the user is reading the application instructions before filling out the application. We’ll get into specifics later, but for now …

Let’s talk path analysis. So what the heck is “path analysis” anyway? It’s basically analyzing the paths your users take throughout your website. The ridiculousness of path analysis might seem obvious to those of us who have been deep into web analytics for a while. For those people just starting out with analytics (or not involved in using them at all), path analysis might still seem like a great idea … at first.

Think of pathing this way, though. How do you make your way through a website? How many times do you use the back button? How many times do you search for a page instead of using link navigation? How many times do you go back and forth between 2 pages?

Now, make believe you are tracking 20 people, just like yourself, through your site. What insights can you possibly get from their exact paths? Not much. Will any of those paths be exactly the same? Probably not. Why? There are simply too many variations. Too many options:

  • User A Path: Landing Page (from search engine) –> Page 1 –> Home –> Page 2 –> Page 3 –> Exit
  • User B Path: Home –> Page 1 –> Page 3 –> Home –> Exit

… and those are the most simplistic I could think of right now.

Another layer. Keep in mind that (depending upon your site, of course) more than half of your visitors land on your site at a page *other* than your home page. This adds another layer of complexity.

What questions might we feel compelled to answer by the path analysis report:

  1. We want to find out if people are taking a particular path through the site.
  2. We want to know what “related degrees or certificates” users are looking at. For instance, are users that are visiting our MBA pages also visiting our Masters in Project Management pages?

There are many ways we can gain insights into two situations without using path analysis. Let’s get specific.

Finding out if people are taking a particular path through the site. Ask yourself this, “what particular path do you want to analyze?” Focus your question to one path. We can do that … and it’s relevant and actionable! It’s called a conversion funnel, goal, or similar depending upon what analytics tool you are using.

For Google Analytics users, here is a tutorial on setting up goals. For Omniture users, use the PathFinder report. For a funnel with more than 3 pages, use the Fallout Report Builder. Remember the example I gave in the beginning of this post? Use the goal/funnel report to see if users are reading your application instructions before starting the application.

Finding out if the users who are visiting our [enter degree name here] pages are also visiting our [enter other degree name here]. Many analytics vendors allow you to set up “sections” of the website. So, logically, you can set up a “section” for all MBA pages and then a “section” for all Masters of Project Management pages. This is fantastic for segmentation purposes. But, don’t get sucked into creating pathing reports for sections. Just like page pathing reports, there are too many options so I can’t see this report having any sort of actionable data to offer.

Instead, how about this … do you have a “request more information” form? Is there a place to select *multiple* programs of interest? How about analyzing that to see the possible “related” programs? Wouldn’t that give you much better data and more actionable data than would trying to figure out pathing options between sections of your website?

Remember that the when you’re feeling compelled to use path analysis and you’re asking yourself questions like, “what paths do users take through my website,” force yourself to give a *specific* example – ” Do users take the specific path from page 1 –> page 5 –> page 6?” You’ll get so much more actionable data out of answering that specific question than wasting your time on a vague, non-actionable question like, “what paths do users take through my website.”

Focus. Don’t waste your time on reports (even if they’re pretty!) that tell you nothing.

Footnote. I must thank Kyle James for (unknowingly) giving me the idea for this post. While reading his great post about why pageviews is not a quality metric (which I totally agree with), it got me thinking about other metrics and reports that are pretty much worthless.

3 responses so far

3 Responses to “Why Do We Care About Path Analysis?”

  1. Kyle Jameson 18 Sep 2008 at 10:26 am

    I think path analysis can be useful if you’re tracking a certain funnel path to a goal. But of course that’s not a complete visit just looks at how someone got into your funnel and to see if they take a certain path from there to complete a goal.

    Then again it looks good on paper, I have yet to actually implement and test it to any degree.

  2. Kenon 18 Sep 2008 at 2:46 pm

    Shelby,

    I probably fall into the data analytics uninitiated category, so please humor me. It would seem to me that a full path analysis could help one focus more relevant or specific questions. That is, do you think that it could provide some useful information as part of some sort of stepwise approach to understanding major user trends in web site use?

    Cheers – Ken

  3. Shelby Thayeron 18 Sep 2008 at 9:26 pm

    @Kyle and @Ken, thanks for your comments!

    @Kyle – Yes, I definitely think that a specific path, or funnel (or goal with Google Analytics) is very actionable. I do think that you have to be careful about how the “funnel” pages are presented on your website, though. For instance, if the funnel is a “mandatory” path, say a shopping cart on an ecommerce site, then a funnel is a no-brainer as it will show you at what specific stage the user “fell out” without any other influences on the pages (I’m assuming the shopping carts are set up correctly – without links elsewhere within the site).

    I think that if you create a goal or funnel with pages with many *other* links on them, you run into the same issue with “too many options” so it’s harder to specifically say, “page 4 of my funnel is where users fell out” if there are other options within that funnel.

    @Ken – I do think that everyone starting out in web analytics is excited about the concept of path analysis. Unfortunately, when trying to put it into any practical action, full path analysis falls short for the specific reason that there are way too many options on a website.

    For example, if my path is 20 pages long and 5 of those steps includes using the back button, I’m just not sure it’s possible to get any trend data out of comparing paths like this.

    You can get actionable data, I think, by creating a funnel (specifying the specific pages through that *funnel*) and then analyzing at what point users get out of the funnel. But, I do think if the pages within the “funnel” have too many links on them, the actionable data from that funnel can diminish the more links are on those pages within that funnel.

    As always, there are exceptions. If you run a short path analysis from an entrance page, you can get a trend on a bounce rate (when a user enters the site on a page and goes no further within the site) or possibly which links on that entrance page are the most popular. You can get those answers easier within other reports, though, in my opinion.

    These are both great comments and I think I’m going to write a follow up post about what exactly “page funnels” are and how to create them.

    Thanks Kyle and Ken!