How many unique values is included in one dimension? If you have possibility to collect data for e.g. unique logged-in visitors with anonymous ID, you might want to see how many different ID numbers you had last week aka how many real visitors you had logged-in. Out ouf the box metric “unique visitors” is really just “unique browsers” and nothing more. Nowadays, when there are so many browsers and devices available should we try to change that metric to “unique browsers”? Well, that’s another story. Back to this new count function in Adobe Analytics.
Before this new function you could check unique values by just going to the last page of the ID report and see how many rows there is. You could also use “row count” function and get unique values reported in the first page of ID report. So there has been many ways to find out unique values before, so what is this new “approximate count distinct function”? With this one you can add this as a new metric to any other (dimension) report, this wasn’t possible before. Let’s take example and you will get the idea:
Oh yes, now you get the idea. You can really start to investigate by real customers or logged-in users or whatever you want to measure based on different values. Again, there is great detailed video by Adobe and you can see how exactly you can build this kind of calculated metric. There is also offical help documentation available for approximate count distinct (dimension).
Note: Like Count() and RowCount(), Approximate Count Distinct() is subject to “uniques exceeded” limits. If the “uniques exceeded” limit is reached within a particular month for a dimension, the value is counted as 1 dimension item.
Oh wait, that’s not all folks!
This is like a TV Shop, you get lots more with count distinct function.
How about doing some quality checking with this one? Let’s say you have eVar1 variable for different sections on your site. You have stated for coders that this value should be only “home”, “search” or “profile”. Somebody makes a mistake and you get 4 different values for this eVar1 variable. With the help of this function you could do some kind of quality report and see are there unexpected amount of values for different custom variables and you could fix those ASAP. (Thanks for this idea to Urs!) Btw, it would be great if you could do alert for this one e.g. “if eVar1 is lower or higher than 3 then send email alert”. And I guess we can, since you can use calculated metrics (or segments with functions) in alerts too. However, this isn’t so easy as it sounds. Will have to investigate more and will give some examples later on if I find the time…
With older row count it was hard or probably impossible to do trended view for unique values, but now with this new magical calculated metric function even that is possible now. Bada boom!
I’m sure there are some other cool calculations you could do, share if you know.