Avinash Answers My Questions About His Book: Part 2 Of 4
As I explained in Part I of this series, after I read Web Analytics: An Hour a Day, I had a lot of questions (and even some things I didn’t agree with.) So I wrote the author, and he sent me back nine pages of thoughts. That’s why I’m chunking my interview with Avinash into sections. Unlike the first part of this series, this one is very down in the weeds; I asked about some very specific best practices. You can see my questions below in boldface and his answers in quiet type, perfectly matching our personalities.
Moving from the very general to the very specific: On page 33, I scribbled, “First party cookies don’t talk to each other, and third party cookies get deleted.” Do you have any recommendations on choice of (vendor? technology) that deals with both of these issues? A first party cookie solution where the various sites in the enterprise talk to each other without lots of manual coding? A solution that you love?
In my prior role we had implemented first party cookies and “first party third party” cookies to overcome this challenge somewhat.
Now I also set a “first party third party” cookie on both, let’s call it tracking.zqinsight.com. The latter cookie I can use if I was looking at both sites as one monolith (to for example get true unique visitors).
It is less likely that this cookie will get blown away by spyware (because it is not being set from known domains of web analytics vendors), though high security settings will still be a issue.
One last point, you have to have a web analytics tool that allows you to create “local” (site specific) and “global” (all sites) datasets with ease and mix and merge sites. ClickTracks is one such tool.
I hope I have answered your question (and the answer is not clear as mud).
Question: On page 37, you stress the importance of having the analytics code at the bottom of the page (“Customer’s first.”) But what’s an analyst to do when the other fancy things on your page don’t work unless the tracking code loads before them?
Let me share some context.
The reason for the tracking code to last is simple: Nothing should interfere with the customer experience.
The page that the customer has requested has to go back as fast as possible so that they can get on with their life (and convert for example). Just in case you have something “funny” going on in the code, just in case your analytics providers servers are under heavy load, or just in case….. we want the customer to get the page first and us to get the data second.
There are always exceptions to any rule. I would set the bar really high to ensure that decisions to load the tag first pass rigorous scrutiny.
I think you do novice analysts a disservice by focusing so strongly on bounce rate. To you, bounce rate is about time on page. But most bounce rates are calculated as (visits entering and exiting on the same page without looking at another page) divided by (visits starting on that page.) Someone can bounce after spending 15 minutes reading the home page of this blog. So when you associate bounce with ways your website is failing (p. 145), the new analyst will be confused. I am not sure I have a question there, but you are welcome to respond.
For blogs my recommendation is that analysts should not measure either bounce rate or time on site. Both metrics will paint the wrong picture, precisely for the reasons you have so correctly identified.
Regardless of how it is computed for most types of websites Bounce Rate is a excellent metric that helps identify opportunities for improvement in acquisition strategies or website entry points.
On P. 274, you wrote that one of the questions you should be asking of your clickstream tool is, “What is the most influential content on the site? How do we know what convinced people to buy?” In general, how do most analysts figure that out, and specifically, how do you like to get that answer with Google Analytics?
I refer to a specific example of using the ClickTracks â€œfunnelâ€ report to identify influential content on your website. I am not aware of any other web analytics tool that can do that (or as easily as ClickTracks does), even if they all have â€œfunnelâ€ reports. It is something unique, and built into, ClickTracks.
If you have access to Discover2 or MarketingLab or your own data warehouse environment I suppose you can construct a complex query to replicate the ClickTracks logic. If you want to understand content influence you should so that, it is amazing what youâ€™ll learn.
You can also use page level surveys (described in detail in the book) to understand value and influence of individual pieces of content on your page (and do it at scale).