Can We Be Honest About Analytics? It Is Far From Perfect

Let me start this post by saying that analytics, in particular Google Analytics, does a lot of great things. In fact, since Google Analytics became ubiquitous, many businesses and organizations have certainly benefited from using its data to make better decisions, period. But, somewhere along the way, it has been undeservedly elevated to both omniscient and infallible and that is dangerous for businesses.

I understand how this happened. I’m even guilty of over believing in Google Analytics from time to time too. There are things that GA does really well. For most organizations who are interested in pretty rudimentary site data, GA in its default configuration will probably tell them everything they want/need to know and do it with a reasonable degree of accuracy. Want to know if people are visiting your site? GA can tell you that. Want to know if that blog post you did generated any traffic? GA can tell you that. Want to know which pages are viewed the most on your site? GA can tell you that. Want to know which pages people enter and exit from? GA can tell you that. You get the idea.

It’s when you want to know more than just these very basic stats that GA starts to crack.

Site Stickiness Is Hard To Really Measure

One type of metric that clients are really into is average session duration and average time on page. You can understand why – people want to know if people are hitting their site and quickly leaving or if they are sticking around for a bit. They also want to know if people are staying on pages long enough either for them to load fully (I’m looking at you slow sites or sites with a zillion tags) or long enough to actually have read the content on a page. Sounds like a pretty simple ask of an analytics program, right? Well, not so fast.

How can GA measure data like “session duration” or “time on page”? It’s not like it is using our device cameras to look at what we are doing on screen and time it (yet… ). In order to provide this data, it has to use some type of criteria. Turns out, for the average time on page, it needs two points between which to measure. Ok, what is the big deal about that, you’re maybe thinking? The last page you visit can’t really be measured, at least not without some major custom programming in your GA deployment.  The biggest issue is essentially that GA calculates these times by subtracting the timestamps between pages, so they pretty much can’t measure activity on the last page viewed.

And yet, there are no disclaimers in GA to alert people to this. So, they go on looking at these data points and assume that they are accurate and correct and make decisions based on them.

Referral Traffic Is Still A Mess

I know that Google is trying to stay ahead of referral spam, but like any spam, it is hard to get and stay ahead of it. With referral spam in your traffic mix, either now or historically, data can be skewed. If your site gets a lot of referral spam, the overall traffic numbers could be significantly off. Trying to drop referral spam out of your data is not exactly a one click solution either. It involves sleuthing out offending referrers (there are some compiled lists you can grab online to get started) and then excluding those referrers from your data views – not something that most average business or marketing people could do.

A lot of traffic is still misattributed by source too. We talk a lot about attribution models in PPC to try to give proper credit to our advertising initiatives in converting a person into a customer, but attribution’s wheels can come off for basic site traffic pretty easily. How? Sometimes it is just a technical hiccup at the moment of referral data being passed. Sometimes it is a problems with custom tagging. Sometimes it is a person using incognito mode or an ad blocker. Sometimes it is the result of clicks coming from within a mobile app – the Facebook app is particularly bad for this. Sometimes it has to do with the coding of your site URLs that can, for example, cause the AdWords gclid information to drop some of the time (I have a case like this right now and it is maddening).

Infinite Scroll Web Site Layouts

The latest trend in web site design is for sites with lots of content to use what I call the “infinite scroll” type of layout. In more traditional web site structures/layouts, each page had its own URL and to view a second page, you had to click on a link to get there. In the new layouts, once you scroll to the bottom of a story, another story starts to load just below it, and this will continue as you scroll down the page. In its native configuration, GA is not great at capturing user activity with this type of layout. It is designed to capture data based on the more traditional structure and can have issues with the infinite scroll.

What kind of issues, you wonder? Well, it can really mess up bounce rate. Depending on how URLs change (if they change as new stories load) your bounce rate can suddenly be 100%. This can put clients into a panic. Problem is that what GA considers a bounce is a visitor viewing just a single page on your site. If your infinite scroll has a URL that doesn’t change, all visitors will look like bounces who just read stuff on that infinitely scrolling page. And, that issue I was describing above about time on page or average session duration? Infinite scroll pages can really impact those numbers as well.

You can get around most of these issues with some additional work, most likely using Google Tag Manager (GTM) to add triggers to different parts of the page to get more accurate data passed back to GA on how people are interacting with the page (depth of scroll for instance to see if they reach a certain point on the page).

Lack Of Clear “Successful Conversion” Page To Track

This one is more common than it should be. To search pros, it is obvious that in order to digitally track a conversion, there needs to be a unique page that indicates a successful conversion. It could be a completed shopping cart page or a thank you page after a form fill. Again, without additional coding, out of the box GA needs a unique URL for you to be able to track. So if your lead generation form reloads itself (same URL) with a thank you message when a prospect fills out the form, you can’t really track that in GA without taking the additional step of adding some type of trigger coding in your page to stand in for a unique URL. Again, this is not something most business people or marketing people are capable of doing on their own.

Education Is Key

As with most aspects of digital marketing, education is key. Educating ourselves about how to set up sites to make data capture easier and educating clients about site structure and the limitations of Google Analytics’ default capabilities and problematic data points. Learning to think about sites and behavior in terms of their data points, helps this process too.

I asked on Twitter for recommendations of favorite Google Analytics experts and here were some that were suggested:

Annie Cushing (@AnnieCushing)

Simo Ahava (@SimoAhava)

Optimize Smart (@OptimizeSmart)

Richard Fergie (@RichardFergie)

Measure School (@MeasureSchool)

Brian Clifton (@BrianClifton)

Avinash Kaushik (@Avinash)

Who and what are your favorite Analytics resources? What is your biggest analytics (Google or otherwise) frustrations? As always, sound off in the comments or hit me up on Twitter (@NeptuneMoon).



Speak Your Mind