Google Analytics has recently changed, introducing a new Unique Events metric calculation. This article has not yet been updated to reflect that change — Mike Sullivan Oct 2016.
As a follow-up to my previous post on Google Analytics events, this article will look at the unique problem of “Unique Events“…the metric that defies understanding by many. Defined in Google Analytics as “The number of times during a date range that a session contained the specific dimension or combination of dimensions“, part of the challenge created by this metric is the way it is presented in the standard reports.
For this example, I sent a number of test events to a test web property to provide a controlled environment. There were 3 sessions in all, with a total of 35 events using a combination of categories, actions and labels. Click the image at right for a larger version.
As per Google’s definition, for each row of the report, which included the Minute and Session Duration to make each event appear on its own row, the number of Unique Events is 1, which is expected.
Lets explore a few query combinations before we look at Google’s standard reports, just to make sure we really understand that definition and how it affects the metric.
Without Any Dimensions: 1 Row
If we make a query (custom report or API call) with no dimensions at all, we see that the number of Unique Events equals the number of Sessions in my test: 3. That is exactly what it should be showing. The calculated metrics Sessions with Event (3) and Events/Session with Event also line up (35/3 = 11.667).
Just the Event Categories
If we now query with just the Event Category dimension, we get a row for each category in our test, and there were only three (category1, category2 and category3)
In my test, category1 existed in the first and third sessions (2 Unique Events), category2 existed in all three sessions (3 Unique Events) and category3 existed in the first and second sessions (2 Unique Events).
In this report, Unique Events equals the number of Sessions that contained each Event Category. This is what you see in the standard Google Analytics Behavior > Events > Top Events report (below).
[sidebar: My previous post about Events explains that metrics calculated from the Sessions metric behave oddly when mixed with hit-level dimensions. All of the Event dimensions (Category, Action and Label) are Hit-level dimensions. You can’t mix Event Category and Sessions just like you can’t mix Page and Sessions. The result: the calculated Sessions with Event and Events/Session with Event numbers are meaningless here.]
Notice that if you add up all the Unique Events in the rows, they will total 7. We know there were only 3 Sessions. The number 7 is nothing more than a total, and the sub-text shows it is 233% of the real total (3). In real data, the percentage would be much smaller because you would have many other non-event sessions, but the number 7 gets top billing for some reason, even though it is meaningless. It even gets repeated in the Events – Overview report.
Drill Down to Event Actions
If you click on one of the categories in that standard report, life starts to get confusing because Google doesn’t explain a key number and you are making (reasonable) assumptions about it should mean. I clicked on ‘category1’ and got this:
I assume that 2 sessions included ‘action2’, 1 session included ‘action1’, and 2 sessions included ‘action3’ (in addition to ‘category1’), and I would be right. That is what those number mean. In this report, Unique Events is the number of Sessions that contained each specific Event Action AND the selected Event Category.
Then there’s that meaningless ‘total’ metric above them: the total of Unique Events…since when does 2 + 1 + 2 = 2 ???
That number actually does have meaning, though: it is the number of sessions that are represented by the report — 2 sessions included a ‘category1’ event, remember? — and that is useful information to know when looking at how the actions are spread around. With a total of 2 sessions represented, seeing ‘action2’ and ‘action3’ with 2’s means the ALL contained those actions! There’s meaning! I wish Google explained it better in the interface.
Oh, and did you notice? The percentages are wrong! They are based on a total of 2+1+2=5 (2/5=40%), NOT on the actual number 2 (2/2=100%). Sigh…
Drill Down to Event Labels
If you then click on an action, you get a similar drill-down report showing labels with a similar metric treatment. That number at the top is the number of sessions that included the action you drilled in to (in this example, ‘category1’ and ‘action2’ existed in 2 sessions). [sidebar: if you go back to the top-level categories report, the number at the top should have been 3 for consistency, since there were 3 sessions that included all the categories]
In this report, Unique Events is the number of Sessions that contained each specific Event Label AND the selected Event Category AND Event Action.
SPECIAL NOTE: there is no ‘(not set)’ for Event Category, Event Action or Event Label, so if you leave them blank, they will NOT be shown as ‘(not set)’….they simply will not be shown, even if there were events and sessions with empty fields. It is not possible to get a report of events with empty labels or empty actions. Events with empty categories won’t appear in the Total Events count.
The Pain of Custom Reporting
People run into problems when they start including other dimensions (we all have unique websites with unique reporting needs). Not all dimensions work together, and Google won’t stop you from mixing incompatible ones.
Here’s my recommendation: Start with the minimal report, just the Event dimensions and metrics. Check the totals as a reference point. Add one of the other dimensions you want to use (or filter with) and compare the totals. If they stay very close to the same, you’re probably good to go. If you find a wild swing, then just accept you can’t get that specific query combination.
I hope this has helped you to understand Unique Events a little better. Google Analytics is an amazing tool with a lot of capability, but it has been evolving quickly, and all the pieces don’t always fit together the way people think they should. Event tracking is one of those pieces.