Stefanie, just out of curiosity, have you tried to lookup the IPs that generate the most bad traffic to get insights about networks, namespaces, etc.? Sometimes, the traffic can be originated not only by "bad" bots, but also by QA bots your IT may use on purpose to monitor website availability and vulnerability. In my experience, the auto bot filters detection rate for this type of traffic is low. The analysis by IP helps quickly discover that.
Bot detection is really a pain, especially for retail sites. Adobe should come up with a better out of box or engineering solution, current BOT removal process (since 2011-12) is just not seems to be working in case of smart Bots. No solutions mentioned in various threads is seems to be full proof.
just a word about "Virtual Report Suites" (VRS): we changed access and give business only data in the VRS. means until someone is an analyst, they only get access to filtered data. In my opinion there are at least 2 advantages:
1) filter bot traffic: any VRS can have more than one segment filtering bad traffic. Whenever we detect something, we just update our bot segment (or add a new one). what happens is that every workspace project (and scheduled pdf) depending on that VRs is immediately updated! no internal communication needed, no need for updating projects at large ...
2) component curation: in a VRS you can select what dimensions (props/eVars) and events should be visible. this makes it easy for new users to start working with project templates since they don't have a bunch of items in the left row.
(almost) no advantages without a drawback: it takes some time to setup (and maintain), but I believe it's worth in the long run.
I noticed that some people create a virtual report suite. I'm just wondering if this is practical or has disadvantages. Therefore it would be intresting to hear if people generally filter out little or what they do. Our goal is to have a clear view of the conversion rate trend.
Historically grown, our filter contains about 20 old browsers, 10 user agents, 9 uncommon monitor resolutions and 7 orig. referring domains. Most of them cause a relatively small amount of traffic i.e. 2000 visits per month. IPs might be added to the segment in the future.
@Andrey: the IT testing system is filtered out, but thanks for your suggestion because we haven't checked networks, namespaces etc. 🙂