Question about creating an Average Time to Complete Flow calculated metric

Hello all! Question for the community. I’m looking to figure out how long it takes users to complete one of our online scheduling flows on the site on average (from form 1 to the confirmation screen), to see if mobile or desktop takes longer to complete the flow. I figured out how to create a derived field in CJA to calculate the total time it takes users to complete the flow from start to finish (meaning it calculates the entire total of seconds all users have spent completing the flow). I called that dimension Time To Complete OLS Flow and added it to my data view. Then in workspace, I created a calculated metric with Time To Complete OLS Flow divided by the total number of flow completions (booked appointments) to find the average time per user it takes to complete the flow. However, I'm not sure whether to use events or visits as the persistence in the denominator for that calculated metric. ie, Time To Complete OLS Flow / Booked Appts (visits) OR ie, Time To Complete OLS Flow / Booked Appts (events)? What do you think would be best practice? I feel like the events data will overinflate the clicks to finish the flow, but visits won’t count users who have completed multiple bookings in one session. When I use visits vs events I get very different answers. Thank you in advance for any insight offered!!