Trending a step in a fall-out report

[MarkVictoria]

I wanted to track the fallout rate between two pages:  a homepage and a conversion page.  Easy peasy.  Then I decided to dig a bit deeper and trend the touchpoint.  It doesn't work.  
If I do a fall out for any given day in the range, it is lower than the one reported by about 10%.  Not a trivial difference.

What am I missing, or is there a bug? 

Which data point is the reliable one?  Note that if i break down a day by hour, there is a similar impact.  Averaging the elements manually, does not equal the average displayed. 

[MarkVictoria]

Interestingly, I switched the container to visit (vs. default visitor), trended my fall out point, then 'show data', then manually averaged the revealed data - that was within a single percent.  I can live with that. 

[KumarRishii]

@MarkVictoria You  can use raw counts (Fallthrough and Fallout) and create a calculated metric for accurate trended fallout.

Trust the overall fallout % (e.g., 60.8%), not the average of trended percentages.

[pradnya_balvir]

Hi @MarkVictoria ,

You're encountering a common discrepancy in Adobe Analytics related to how fallout visualization and segmented or trended fallout data are calculated. Let's break this down to identify what might be causing the inconsistency and which data point you can trust

The overall fallout rate (60.8%) at the top of the trend chart is the correct one when you ask:

“What percentage of visitors dropped off from homepage to conversion page over the entire period?”

This value is not the average of daily fallout rates. It is instead calculated from aggregated totals over the selected time range:

Fallout % = (Total users who viewed Homepage but not Conversion page)
÷ (Total users who viewed Homepage)

Meanwhile, the daily percentages reflect:

Daily Fallout % = (Homepage viewers who didn’t go to Conversion page that day)
÷ (Homepage viewers that day)

And averaging these daily rates is not equivalent to computing the rate from summed totals. This is due to differences in user counts and the effect of smaller/larger volume days on percentages.

If you want to accurately trend fallout over time, here’s how:

1. Use calculated metrics instead of default fallout charts:

   • Create a custom metric like:

     (Homepage Instances - Conversion Page Instances) ÷ Homepage Instances

2. Use Segments to isolate users:

   • Segment 1: Users who saw homepage only.

   • Segment 2: Users who reached conversion.

   • Create a fallout comparison based on unique visitors or visits for each.

Try replicating the fallout manually using segment-based counts.

Thanks.

Pradnya