Cake and
This is a pain in the ass and it always seems like networks disagree with tracker databases a lot, it's perverse.
India is a relatively evolving market in terms of internet structure and connectivity so it's possible that IP allocations are changing often, making it a terrible geo for this sort of thing - maybe, just a suggestion.
As for the US traffic, that's really unacceptable. How many clicks were there including dupes (check click report > include dupes, not sure why people don't only report this stat when talking about click loss)? How many were out-of-geo according to your AM?
Hello zeno,
Thank you for your help. Your suggestion makes a lot of sense.
Unfortunately, I have trouble checking how many the dupe clicks are in cake because sometimes it’s really unlikely for the stats to load. Even if it finishes loading and the “loading” pop up disappears it still says “No Items Found”. Sometimes this also happens when I’m actively running campaigns. BTW, is there any possibility that this could affect conversions? (due to something not working in the front end perhaps)
Alternatively, I had a look at
I took a screenshot from the same campaign that was used in the click loss example above:

Unique clicks start to appear from 3rd page so 2/5 of the IP addresses sent me duplicate clicks.
Screenshot of the Indian campaign where the geo-redirection was happening:

IP addresses with unique clicks start to appear from the first page when the scroll bar reaches the red line I drawed. They weren't as many as I expected...
When I complained about this to my AM, she didn't specify with numbers how many the out-of-geo clicks were, but she said that the majority of the clicks were US traffic.
You'll need to sort out the show duplicate clicks bit I think.
Just looking at your data suggests that - at least in the first shot - well over half of your clicks are not unique in terms of IP. It thus makes sense for at least 50% apparent click loss when just looking at the general stats in Cake.
In the click report try setting the time window more narrowly.
Otherwise, export that data from