1. Understand who the listener is, and what they care about. What are their reinforcers, and attractors? What are the things they want to achieve that they haven't been able to so far. What is relevant information for them? What is their state of mind? Are they tired? Do they need to sleep, eat, go to the bathroom?
2. Understand the information content of the speech. A person can say a lot in a little time or a little in a lot of time (with the dependencies being questions listed above in #1. What is a 'little' for someone might be 'a lot' for someone else depending on how much they know).
My concern with an app like this is multi-fold. Firstly, I have some very smart friends who read a lot and talk for hours but I never tire of it because they're always saying something new and interesting. Secondly, sometimes these friends are working out thoughts on the fly that they haven't before. Stopping before the thought is complete risks losing hidden gems, maxims/aphorisms, well-articulated and profound crystallizations of thought. A ticking time-bomb clock adds pressure to end this line of thought that may need time to manifest itself fully.
That being said I totally see the flip side, that it can go on for too long, but I'd point back to all the concerns in #1.
Attendees of a meeting will quickly disengage if someone talks for too long. Attendees of a lecture or presentation have different expectations.
That's the key question that will frame how I schedule my attention and interjections.
The reason I chose the times as seen is because they jive with my personal experience and the content of this article here: https://hbr.org/2015/06/how-to-know-if-you-talk-too-much