Metrics Average time spent on site

[RamyaSheela]

Hi ,

I’m noticing a discrepancy in the average time on site. When I view the time for a specific segment, the total average time is shown a minute, but when I break it down by day or week, the average time increases significantly around 6 minutes.

Could you help clarify why this difference occurs?

But when I breakdown by custom date segment it showing "around 1 mins"

If I need to show the break down by month is there any other way I could present data.

Thanks

Ramya

[Martinej]

Hi RamyaSheela

You could check if there are outliers or high-variance days affecting the averages but also I recall this might be related to how Adobe Analytics calculates this metric at different levels. 

The metric "average time on site" is calculated by taking the timestamp differences between consecutive hits for a specific dimension item. Then AA sums these times and divides by the number of unbroken sequences ( continuous sessions without missing data points). When there's no subsequent hit to a dimension item (such as the final hit in a session), that last hit isn’t included in the calculation, which can lead to variations between totals and breakdowns. At least that is how I understand it and would make sense in this case here
(There is an example in the documentation:  https://experienceleague.adobe.com/en/docs/analytics/components/metrics/average-time-on-site )

[MandyGeorge]

The time calculation metrics can be quite tricky. I attempted to replicate what you're seeing, and I got essentially the same result. 

I decided to look at it both ways, breakdown the segment by the day, and then breakdown the day by the segment. In the top of the screenshot, I get a very short time on site for the segment, but then the day breakdown is much larger. In the bottom half, the day has a larger time spent on site, and then the breakdown gives the segment row a very small number again. 

What I think might be happening is that in the row with the segment, the time spent on site is looking at just the conditions that are in the segment (so only specific pages), which looking at a specific page or two is definitely going to be shorter than an entire visit. Then using the day breakdown I think is bringing in the whole visit again, not just the part in the segment.

I also looked at what happens with the date range. The date range for my panel is last week, so the header row of the second table below is the same time period as the date range for the first table. I had to block out visits, but they're the same in both tables. You can see that time spent per visit (seconds) is the same for both. 679 seconds divided by 60 gives 11.32 minutes, which is pretty close to what the bottom table has. So even with the date range it's doing something weird like it is with the segment. 

I'm not entirely sure what is causing it. I think @Jennifer_Dungan is going to try and figure this out too.

But I think my recommendation would be to put your segment either at the top of your panel (if you want it to apply to the whole panel) or above the metrics in your table (if you only want it for a particular table/column). Both of these solutions will result in no single line with a time against the segment itself, it'll result in having your dimension in the table and seeing the values associated with them. As for time periods, I would try and use panel date ranges instead of the component ones to avoid any miscalculations like this. If you do have to use the component date ranges, put them on top of your metric and use a dimension for the rows. I tested putting the segment and the date range on top of the metric and the totals are unaffected.

So to summarize - don't put the segment or date range directly against 'average time on site', use a dimension with the metric and apply the segments/date ranges you need above the metric or in the segment drop down at the top of the panel to avoid this issue.

[MandyGeorge]

Ok, so I was just reading through some documentation because I still didn't get why this was working the way it was, and I came across this. 

https://experienceleague.adobe.com/en/docs/analytics/components/metrics/time-spent-per-visit

The important part here is "average time on site uses the sequences that includes a dimension item as it's denominator"

Meaning when you don't have a dimension item against it, it doesn't have a proper denominator to use, so it's likely taking each hit individually (basically giving a time spent on page, not a time spent per visit). But when you put a dimension item against it, then it does the calculation correctly. 

So my solution from the comment above still stands - don't put segments or date ranges in the row against the metric, use them at the top of a table/panel and make sure to include a dimension against the metric.

[kayawalton]

@MandyGeorge's suggestion is spot-on. If you want to see the breakdown of Average Time on Site for entire visits by day/week for a particular segment, it is best to apply that segment to the metric in the freeform table rather than treat the segment as a dimension. That is because Average Time on Site is calculated based on the number of unbroken sequences associated with a dimension. When you add the segment into the dimension, Average Time on Site is then calculated based on the dimensions included in that segment. An entire visit could consist of several unbroken sequences depending on what dimension is considered, which is why the Average Time on Site is lower with the segment as a dimension.

However, since it looks like you want to know how long an average (entire) session is for anyone who applies to that segment, you will want Adobe Analytics to consider these sessions as whole as possible. Applying time dimensions (except for minutes or hours) are more effective in that case, but the likelihood of an entire session being broken up across days/weeks/months is very low.

I provided an example of the effects of Average Time Spent as a dimension vs. as a segment to a metric here.

https://experienceleaguecommunities.adobe.com/t5/adobe-analytics-questions/total-time-spent-on-site-...