This depends a bit on of you have set up your funnel. Above the place where you choose your events for the funnel, you have a dropdown that allows you to choose This order, Any order, or Exact order.

The default is This order. In this case the funnel checks if the user performed the listed events in the listed order, but does not mind if they trigger other events between the funnel events. So in this case, drop-off is someone who does not trigger the next event in the funnel at all within the chosen conversion window. Triggering other events is allowed, the only thing that matters is whether you do the funnel events in the chosen order or not.

If you switch to Exact order, you get the behaviour your described: in this case Amplitude expects to see the listed events in the listed order and no other events in-between. So straying from the path in any way is counted as a drop-off.

Also be aware that the funnel has a conversion time, by default 1 day. If it takes longer than that for the user to complete the funnel, they are considered drop-offs.

When you look at the User Paths, does 70% cover only the users who went directly from Login Screen to End Session? At least our data shows users going bit back and forth between different screen before actually dropping off. So if you have users who do for example Login Screen → Welcome Screen (back to the first screen) → Terms and Conditions Screen → End Session, they would still fail to convert in the funnel, but not go directly from Login Screen to End Session.

As for analysing the paths to End Session, would creating a cohort of the users who never signed up and then using Pathfinder with that cohort and “Ending with End Session” be what you are looking for?

In agreeance with MikkoKarvonen!

For more information on the Pathfinder chart, feel free to check out this doc. Please note this chart is available to those only on the Scholarship, Growth, or Enterprise plans. I’m not sure which plan you are on, theo, but if you’d like to learn more about our paid plans, please reach out to Sales here. 

Thanks for your answer!

Indeed, 70% is only users going from Login Screen to End Session. I also noticed some users going back to a previous screen before dropping off so the actual drop off (quit app) is higher. But then I’m not sure what the % next to each path represent? What I thought at first was that among all the users who dropped off at the step I’m inspecting, x% have followed path 1, y% path 2 etc… But adding up the % gives more than 100%, so I guess it’s something else? 

Is this something you could share a screenshot of? It would make this much easier to comment on.

Right, so for example this is the cohort I’m inspecting:
 

And this is what I get when I click “Show User Paths”:

 

Then my two questions are:
​​​​- ​​​​​​Path 3 seems valid. Why is it included here? Isn’t this supposed to be user paths taken by the drop-off cohort = only drop-off paths?
​​​​- What are the % next to each path representing? I first thought it was something like: 88% of users within that cohort took Path 1, 11.3% took Path 2 etc… But when you add up the % of each Path 1 + 2 + 3 you get 108% so I guess it’s meant to represent something else?

This one is difficult to comment without knowing how the chart was set up. Couple options that come to mind are that these users took longer than you had given time to complete the funnel, or that there is some small detail that means that this path does not actually count as completing the funnel.

It is the % of the users in the cohort, i.e. those 9,19% who dropped off. The reason it’s over 100% is that some of the users may have completed more than one of those paths.

For example, the first time they encounter launch screen, they just end the session there. Then they launch the application again / come back to it, initiating a new session where they land on the launch screen again. This time they fiddle around a bit, and trigger the second path instead.

This is the reason there are two versions of the Pathfinder: one for unique users, one for total events.

Got it, thanks a lot for your help Mikko!