Why there are differences between same data (screen included)

[kermits6]

Hello I am asking about why are there different time when I am selecting exact same data on freeform table on Average Time on Site in Analytics Panel as seen on attached screen.

Thanks for explanation.

[igupta]

Interesting! I can see same behavior so it must really be how sequences are being used with month dimension. Sharing few observations and assumptions-

First Average Time on Site is calculated based on the total seconds spent divided by number of sequences (of that dimension). Detailed info here - Average time on site

As we can see, Time Spent Per visit is same for both reports, so it's only the number of sequences that is different for both tables. My assumption is, for table 1 - the sequence is counted specific for only the selected dimension value (using it like segment) and that's why the count is higher. For table 2, it looks like the number of sequences is calculated aggregated of entire dimension values i.e. aggregated based on all the values the dimension holds, hence the count of sequences is lower resulting in higher averages. Furthermore, if you will break it down by pages, you will see exactly same values per page and at aggregate level. My other assumption is that for table 1, it is counted based on number of days in the month and in table two, as dimension value, taking it as 1 per dimension value.

Since, I couldn't find anything in documentation, I am just sharing my assumption. Maybe, other experts in community have better answer for you.

Best,

Isha

[Jennifer_Dungan]

I also see the same thing, and even more interesting is when I take the version that has the metric as the breakdown, and break that down again by Day, then the "average" on the Day row matches the first table.

I think @igupta might be right, that the way the calculation is being done may have something to do with the number of days vs the month as a whole... Sadly we don't have visibility into how these calculations are really performed...

[igupta]

Interestingly that's probably because for table 1, the Total Time Spent (in seconds) value seems to have been divided by 1 whereas, for table 2, all values seems to have been divided by 4.7..Now it could be 4.7 weeks in average per month?

e.g. =305/4.7 = 65

[Jennifer_Dungan]

That could be? Or by the average number of Days per each user? Mine numbers don't line up with 4.7 (I have 8.3 in Sept, and 8.0 in Oct), which makes me think it's not specifically days or weeks in the month, but maybe the days for each user?

[igupta]

Hmm...makes sense. I think only someone from Adobe Engineering team can confirm the actual reason! But it's fun sleuthing

[Jennifer_Dungan]

I wonder if we could reach out to someone like Ben Gaines?

[igupta]

No harm in trying, I guess