I'm trying hard to figure out an odd behaviour but I cannot yet grasp the explanation.
I'm currently testing 2 segments ( that are almost exactly the same ) , one has a "Hit" container (let's call it segment "A"), the other one has a "Visit" container (let's call it segment "B") on the same custom traffic variable ( lets call it propX).
I was predicting that I would have the same number of visits if i segmented the visits metric with segment A or B. This showed already alright on both of the segments previews. But on the workspace, I found out that there were more visits using the B segment.
I found out what was causing this - I remembered that I had already another segment (let's call it "C", also on a custom trafic variable)with a Hit container on the panel header, based on another prop ( let's call it prop Y).
If I take out this segment the number of visits is the same regardless of the segment (A or B).
So, how is it possible, that
(C+B)visits > (C+A)visits
(B)visits = (A)visits ?
This almost looks like a math topology problem ! Sorry for the odd formalism on the 2 previous wannabe equations...
segment A only contains hits where propX is set, no other hits. in segment B there are all hits where at least one hit has propX. you should see a difference in occurances or page views if you only look at those two segments.
now adding sehment C:
A+C: only Hits (and corresponding Visits) where both PropX ANd propY are set on the same Hit
B+C: shows Hits (and corresponding Visits) where propY is set on every hit BUT only needs one hit of the Visit to have propX. Because B is a visit segment, the propX must not be on the same hit ad the hit with propY. this means, that the number of visits with B+C should be higher than A+B.
does this makes sense? is it possible that a visit can have propY but never propX on the same hit (but elsewhere in the visit)?