Hey @lars 

If I’m understanding your use case correctly, this could be a use case for holding properties constant and broken down by feature. Apart from the timestamps, you might need to instrument an invoice_id and invoice_type /document_type event properties if you haven’t already to leverage these features in your use case.

Here are some posts for similar use cases - 

Let me know if this helps or if your use case is a different one.

Amplitude Funnel analysis allows you to see the histogram for time to convert: https://analytics.amplitude.com/demo/chart/new/i8nbpmv

This would give you one starting point.

However, there are two things to note here:

1. The nature of the funnel chart is to track unique users instead of event flows. To get around this, you would need the holding properties constant approach Saish Redkar linked above. For this to work, you’d need to have some kind of id as an event property in all the steps you want to track.

2. It’s still going to be tricky to explain why some invoices take longer than others. This is because “why” questions are generally tricky to answer with analytics. You can find out what, how, and who, but with why, you’ll very easily start to make guesses, which typically means bringing in your own pre-conceived notions and biases.
   But two ways to analyse your data come to mind. First is the broken down by option mentioned above. This is dependent on you having the relevant event properties in all the steps. This would allow you to analyse if certain types of invoices tend to get stuck at some specific step of the process.
   Another option would be to utilize the custom bin sizes in the time to convert view (like this: https://analytics.amplitude.com/demo/chart/new/do1dd52). Set this up so that you are easily able to identify those invoices that take longer than you’d like to get through the process. Then you can click on that bin, create a cohort, and analyse that cohort to see if for example certain types of users or user behaviours pop up.

In addition to Saish and Mikko’s responses, you could also consider instrumenting a duration property on your events to track time spent on each event. 

Then with a custom event of all the approval events you can do Sum of Duration property grouped by invoice_id to get you the sum of all the Duration on the events per invoice_id.

For recommend on instrumenting the duration property, see: