Hey Scott, this is definitely something we're aware of and we're
currently working on a better method for more granular control over the
lookback window. As a short term workaround for your use case set the
reporting date range to start on Dec 31 (as you mentioned), but use a
segment to filter the results to just Jan 1 - Jan X (using the orange
"Day" dimension as a filter). That'll force the report to use Dec in its
lookback window, while still only reporting values that happened in Jan.
Hi, thanks for the suggestion! We actually did an internal POC for this,
but it turns out that regex engines were too slow for us to return
reports in a timely manner. We'll keep looking at it though, thanks for
Thanks for the feedback. For what it's worth, you can reproduce any and
every visualization in the Attribution Panel yourself in a blank panel.
We made panels such that a user could reconstruct them manually if they
wanted to. There's no reason you couldn't look at the same histogram or
Venn, etc. that we prepopulate in the panel.Granted not quite as
convenient, but still doable with a few extra minutes of time.
I like your idea! Certainly not an easy change by any stretch of the
imagination, but I'll take this feedback back to the team. Last question
for you - if we charged not only by hit count, but by size of hit (I
guess that'd be similar to volume or GB pricing) would that help ease
the strange implementation incentives?For example, would you prefer to
send over lots of hits that don't have as much data in them vs not as
many hits but they have lots of data? We experimented with this idea
Great point. The reason for this is because we have no way of knowing
how to apply an attribution model from outside of the calculated metric
builder. For example, if you had revenue / orders and you applied
"Linear" to it - does that mean (linear revenue) / (linear orders) or
(linear revenue) / orders or revenue / (linear orders). Adding more
metrics, segments, calculations, etc. makes it hard for us to know how
to apply the model. Does this make sense?
Ah gotcha - a segment would include both dogs and cats, a breakdown of
"small" by product would only give you "dogs" though. Sub-hit
segmentation is a common feature request - I'll pass this along to the
I should also mention that Attribution IQ will be available to customers
on the new SKUs: Foundation, Select, Prime, and Ultimate. Also fwiw,
participation has always been available in calculated metrics which
surfaces in Workspace.
Hi David, the tricky part there is that the currency conversion rate
changes every day - keeping track of the historical conversion rates and
applying them properly each day isn't as trivial as it sounds!
I'd be careful about sticking IP address into a custom variable -
however, it could technically work as you've described. Have you tried
my suggestion of filtering out Linux, or unknown browsers?Another thing
to try is using customer attributes for this (actually passing the
cookie ID in as an alias ID using setCustomerID). I've seen some
customers do bot analysis offline (or find the bad IP addresses) and
classify the bad visitors as bots. This works retroactively as well.
Hey Jen, thanks for the suggestion! Have you looked at using a virtual
report suite for this? I've found that creating a segment filtering out
the "Linux" or "Not Specified" operating systems or if the browser
contains "unknown version" usually gets 75% to 80% of bot traffic for
most sites.If you need something more serious, you can look at some
solutions in our exchange portal:ShieldSquare PerimeterX Bot Defender
for Adobe Analytics
Great feedback, thanks. It's something we continue to look into - at the
moment these partners are the best options available to any analytics
solution that I know of, but totally understand not wanting to pay
Really love this idea - in fact, it's similar to something we presented
at a "sneak peak" at a session at Adobe's Summit conference last year.
You're right about the technical challenges, this is a tough one to do.
That said, we're developing a few features (like Report Time Processing)
that will make this possible one day.For now, I recommend using segments
to help you filter down your dimensions to just the values you like!
So, looking at this, it's a little more complex than it sounds like -
for example, let's say I have a sequence of hits:a -> b -> a -> c -> a
-> d -> band I create a sequential sequence "a then b",How would you
want "before the sequence", "during the sequence", and "after the
sequence" to work?