RANKX with multiple FILTER

Hi AllisonKennedy 

first of all thanks for your brilliant response. It looks like its working without any doubt and your description.. wow - its exactly what I need to understand what is happend 🙂 Thanks for this!

My rationale of desired output is to calculate how many times user and support write to each other before solving the problem. The total number of messages is also interesting, but this "rotation" is important to me. I imagined that I would then count all 1 per user, 1 per support, 2 per user, 2 per support,.. so then I would know for example that support "returned" to the chat x times.
It is useful to know how long the reactions of one of the communicating sides takes, or then to calculate the total time, the average between the exchanges of messages, etc. This is all for chat statistics and QA.
Then I want to put it in some funnel and see the throughput of all or then filtered problems.


Could I ask you for one more improvement that I canot add to your solution?
In the original data, I also have a column with the content of the message. Can be "Order" count from the time the text appears in the chat? The column is named "message_text" and the text we are looking for is "% someone% connected.." 🙂
The point is that the user can talk to the robot for a while, but the real support will join later.