Expand my Community achievements bar.

Join us for our next community Coffee Break on February 28th! Four of our Analytics Champions will be joining us to discuss Summit tips, best practices, and any of your Analytics questions!

Filtering out "Unspecified" operating systems

Avatar

Level 2

I have a report I'm working on for a client with a relatively small amount of traffic.  But what I've been seeing pretty consistently is their number one referrer is "typed/bookmarked" and I can pretty well guarantee this website is NOT getting that much direct traffic. When you break it down by IP - ALOT of it is clearly bot traffic.  I filtered out the easy ones that were all originating from the same IP block (and filtered the IP in my account).  But I'm still looking at what appears to be alot of junk traffic. 

One thing I noticed is alot of the traffic still falls under "unspecified" for operating system which is typically signs of a bot.  So my question is, would it be a bad practice to segment out the traffic to EXCLUDE any unspecified operating system? Or is there a use case where legitimate traffic could show "unspecified" for operating system?

 

Thanks in advance
Matt

3 Replies

Avatar

Community Advisor

It's okay to ignore/exclude the unspecified value. Unspecified will appear for these scenarios:

 

“Unspecified” is a fairly common line item in reports. It is also frequently referred to as “None”.

  • An event fires without a conversion variable: For example, a user comes to your site and makes a purchase without any eVar1 value. If you view orders using the eVar1 dimension, there is no value to attribute this order to. Therefore, it is automatically attributed to “Unspecified”.
  • Unclassified data in classification reports: When viewing classification data, any value that doesn’t have data associated with that particular classification returns “Unspecified”. To resolve this issue, classify the parent variable value.
  • Breakdown reports where only one variable fired: When you apply a breakdown to a variable, every instance of that variable must be accounted for. If the second variable was not seen or if it persisted from a previous hit, the dimension item is “Unspecified”.
  • Non-mobile hits in mobile reports: Any non-mobile hits in mobile reports are listed as “Unspecified” (“Non Mobile” in Reports and Analytics).

https://experienceleague.adobe.com/docs/analytics/technotes/unspecified.html?lang=en

Avatar

Community Advisor

The other answer is only talking about "Unspecified" in general terms... let me add specific thoughts on your question.

 

I would say "no" it's not bad practice at all... I actually think it's a great idea.... but here are some additional thought that you might want to consider

 

I created a "Clean Data" Segment, and created a Virtual Segment based on "Clean Data"... Clean is where I can exclude data that I believe was inflated by potential bots, or tracking issues where we had double tracking after a bad code push, etc...

 

Using something like this, you can exclude the Unspecified OSs... but if you find that maybe something shouldn't have been removed, you can always adjust the segment and the data will flow back in (both current and historical). But being the main "report suite" that people use, no one has to remember to add a bunch of exclusions... and if you document what is blocked, then there is transparency. It will take a while to transition people to use the Virtual Suite, but then everyone uses that one source of truth.

 

I too do use IP Bot Filters, but as that is all "reaction".. it only applies as of the IPs added to the bot rules (but also, , I plan on using Clean (at some day) to try and also use it to help clean up the days/weeks of data from those Spam IPs But you can't block by OS in Bot rules, so you have to use a few methods to keep your data clean

 

And now I am going to check my data, and maybe apply the OS filter as well if it looks problematic.

Avatar

Adobe Champion

We've actually run into "Unspecified" for OS, and after doing some breakdowns, we found that they mostly fall in two buckets:

  • Feature phones - some really old phones are able to access our websites and they generate no metadata for OS. We were able to discover them by breaking down the Unspecified OS row by Mobile Screen Size.
  • Some referrals from Google using Opera Mini are also registering as Unspecified OS for us. We were able to unearth those by cross-checking Unspecified OS with Referrers.

It might be helpful to cross-check with other dimensions before you filter them out to make sure you don't filter out odd but meaningful use cases like we have.