A Visit-based dimension in Webtrends is an eVar in Adobe Analytics, right? And Visits for a Page in Adobe Analytics will yield the same result as Sessions for a Page in Google Analytics, no? Beware of false friends! This article series shows you the nuts and bolts of Custom Variables, holders of your most precious data – by comparing Google, Adobe, and Webtrends Analytics.
Learning a new Analytics tool is a challenge. While at first sight, many things look similar, under the hood, they are often quite different. One of the most common implementation or interpretation errors stems from Analysts thinking that what they know from the tool they are familiar with will work just the same way in another one.
For example, an Adobe Analytics (formerly “SiteCatalyst”) or Webtrends native will look at you in utter disbelief if you tell him that, in Google Analytics, he should not use understanding Visits (now “Sessions”) as a metric for Pages or other Hit-based Dimensions, and that he needs to use a strange metric called “Unique Pageviews” instead because sessions count for a page only if that page was viewed with the first pageview of a session. Try to teach a Google Analytics native why campaigns are in a completely separate bucket in Adobe and Webtrends and cannot be overridden by normal referrals or organic search visits (which is good because campaigns and referrals can overlap).
Or explain to a Webtrends native why, when collecting data into a Custom Variable in Google Analytics, you have to already know whether you want this data as a Hit- or Visit- or Visitor-based dimension later and that you cannot just collect it in any form and then just use it at all of the three persistence levels (or even turn it into a metric if you wish so).
Learning the Different by Comparing it to the Familiar
It is trivial that we learn similar, but different concepts – think “learning new tools” – better when we can compare them to the concepts we are familiar with. This is why I want to share my learning experiences going from Google to Webtrends to Adobe Analytics in the last years. So my post shall help a Webtrends or Adobe native getting used to Google Analytics, and vice-versa, and maybe avoid some of the headaches I had.
The most useful data usually comes in the form of Custom Variables, i.e. dimensions and metrics that are custom-made to fit your business case. I think one of the main differences in how these three common Web Analytics tools work lies in their different concepts behind Custom Variables.
What is a “Custom Variable”?
First, what is a “Custom Variable” after all? By “Custom Variable”, I mean any non-preconfigured, non-standard dimension or metric that can be filled with any value. To give an example, a pre-configured dimension is usually the Browser name. All Analytics tools automatically extract that name and fill a standard report with it. You can’t do anything about it. A Custom Variable instead could be the login status of a user, the name of a form, the URL of the download link (examples for Custom Dimensions), or the number of logins, the number of form completions, or a revenue counter for a specific product category (Custom Metrics).
Not all Custom Variables are made the same, even in one and the same tool you find different types. They differ in limits like the following ones:
- Flexibility of Scope/Persistence: Can any scope (e.g. Visit-based, Hit-based) be applied to a Custom Variable? Can that scope be changed ad-hoc? Can a parameter used for a dimension be turned into a metric (like in Webtrends)?
- Flexibility of Hit Types: Can any hit (any request) send this Custom Variable along or is the variable tied to specific types of requests? GA’s Event Variables (Event Category, Action, etc.) for example cannot be sent with anything but Event Hits.
- Character Limits: How many characters can go into a variable’s value (very important to know in Adobe Analytics where “props” have a 100 character (more precisely: byte) limit)?
- Reporting Context: Can the variable be used in any context and filled with any content, or is it tied to a certain meaning (e.g. the On-Site Search Term dimension in GA), or is it only possible to use a variable in some reports, but not in others (e.g. not possible to use Adobe “eVars” in Pathing Reports)?
- Availability Limit: How many variables of a certain type can I use (e.g. free GA limits to 20 Custom Dimensions and Metrics)?
From “Pre-configured” to “Free” Variables
I see Analytics variables on a continuum from standard, “pre-configured” (non-custom) to entirely “free” (custom) variables (the ideal that does not exist). Whereas Adobe and Webtrends have always been more at the “free” end of the spectrum, GA has been an advocate of “semi-free” variables coupled with sensible out-of-the-box reports, but GA has recently been moving more and more into the “free” realm with the introduction of its rather free Custom Dimensions and Metrics. The following diagram shows some examples for Custom Variables and where they would be located on such a spectrum (feel free to criticize my placements):
“Semi-Free” Custom Variables
“Semi-Free” Custom Variables are variables that can be filled with almost any value, but they are limited in their flexibility and are often tied to certain pre-configured reports. An example are Google Analytics’ Event Tracking variables. GA allows you to fill three Dimensions (Event Category, Action and Label) and one Metric (Event Value) with custom values, so we could call them “Event Custom Variables” (not an official term!). But they can only be filled with Event Hits, not with Pageview Hits. Their persistence level (also referred to as “expiration” or “scope” as Google calls it (Visit, Visitor, Hit etc.)) can also not be changed. So they are “semi-free”.
Other examples from Google Analytics could be the various Enhanced Ecommerce dimensions, the Social Interaction Dimensions and Metrics, or the “On-Site Search Term” Dimension and the useful out-of-the-box On-Site Search Report it generates, something that other tools handle in sometimes not-so-easy-to-set-up Custom Reports.
“Free” Custom Variables
Now that you know what “semi-free” CVs are, it is much easier to understand “free” variables. “Free” variables are free because they can be configured freely, i.e. their scope/persistence/expiration (Hit, Visit, Visitor or even more expiration types in AA and Webtrends) is up to the Analyst’s will and they can be set with any hit type (Event or Pageview Hit). Google’s Custom Dimensions (and formerly its 5 “Custom Variables”) are an example. In Adobe Analytics, the most flexible example is the “eVar”. In Webtrends, any variable is totally free in its configuration, you just collect a request parameter and decide later what to do with it, but the variable’s flexibility of scope is slightly under that of an Adobe eVar which can be set to persist during any period you choose (a Hit, a minute, a year, until Event x happens etc.). Therefore, the request parameter you collect for a specific eVar can only be used for this eVar with its powers and limitations and nowhere else. If this confuses you, wait for the later parts of this series.
So in short: There is no unlimited freedom. No Custom Variable can be used in any reporting context AND is available endlessly, can be set to any persistence, sent with any hit and offer enough characters to fill it for any reporting purpose. So that’s why in the diagram above, no variable reaches the right end of total freedom entirely.
And one more note: If a tool’s variables are on the “free” end of the spectrum, it does not mean the tool is better. It just means that your custom variable reporting may be more flexible and data collection easier (less important to think about scope, character or availability limits or reporting context when implementing). But of course, too much flexibility can also be overwhelming…
Next part: Google Analytics
So much for the theory behind it. In the next episode of this article series, we will start with the world of Custom Variables in Google Analytics, so we will check on the Event Dimensions, the Custom Dimensions, Metrics, Goal variables, and others.