Here's a tip for getting loads out of a transcript of a client-facilitator session or an interviewer-interviewee recording, especially if you are seeking to model the pattern of facilitator/interviewer decisions.
After you have read the transcript through at least once, go back and read it again, this time think of a unit as a Response-Question-Response triptyct (Question-Response-Question works well too).
My metaphor is a 3-paned window which I can move up or down the transcript to view any three adjacent bits. When I can see the 3 bits of the transcript I muse 
on the relationships
between the three elements, e.g.
- What kind of response has the client given? 
- Where does the facilitator's intervention invite the client's attention to go?
- How did the facilitator take into account the previous answer (or not)?
- How did the client respond to the question?
Then I move on to the next unit and repeat the process. The first pane now contains what I could see in the last pane of the previous window, i.e. (R=client response; Q=facilitator intervention):
1st Unit: R1, Q1, R2
2nd Unit: R2, Q2, R3
3rd Unit: R3, Q3, R4
and so on
When I have looked at a number of windows I add to my musing:
- What patterns of facilitator-client interactions are emerging across units?
I think it is valuable to do it this way because rather of thinking in terms of a simple question-response pairing and therefore the facilitator leading the process, I am trying to see the co-creative aspect, the dance, where both parties contributing to the overall direction of the session and neither is in control of where it goes.
Also, seeing each question in the context of the response before
and the response after
gives me more of a sense of a what the facilitator is picking up on and what the client is choosing to answer about.
Once you get good at modelling in units of 3, you can increase the number of panes to 5, and then 7, and then ...
It is a slow process but, in my not so humble opinion, the rewards are well-worth the time and effort.
I have other ways of modelling transcripts. Perhaps I should write them up some day!
1 See Penny Tompkins and my article: A Model of Musing: The message in a metaphor
2 There are all sorts of ways of classifying the kind of response. One of the most useful Penny and I have founds is to use the REPROCess model. See REPROCess and the First Principle Last modifications 16 Nov 2016.