This goes back to
Rob Donoghue's analysis of D&D's strength being combat, because combat combines the swinginess of the d20 (which makes the rolls exciting) with rolls that individually are low stakes ("Do I deal some hit points damage or not?") even if they add up to something important in aggregate. But when you make something important into a simple pass/fail check ("Can I talk my way past the guard or not?"), the game doesn't work anywhere near as well.
The funny thing is that for many combat isn't very exciting in the attack roll (the d20) by itself. It is fairly well established that most PC will hit 65% of the time (+/-5%). It is actually somewhat predictable, which is why creating encounters to a particular difficulty isn't too hard. What makes combat exciting is when those chances drop significantly and hitting becomes rarer OR when the low stakes pile up enough that further loss (or gain) becomes more meaningful. When you have 50 hp, 5 damage isn't a big deal; but when you are down to 10 hp, 5 damage suddenly becomes much more meaningful.
Because the stakes are (at least potentially) life or death, a single roll would be more exciting simply because the point of failure would likely be ultimate--ending the game for that PC in all likelihood.
How important does "talking your way past the guard" become? Is it life or death? Not likely. If talking doesn't work, intimidating might, failing that direct force or finding some other elusive way (sneaking or another path) works typically.
So, in essence, it can be like combat. First "attack" is Charisma (Persuasion) to "talk your way by the guard." If that fails (a miss), you attack with Charisma (Intimidation), then perhaps (failing that as well) leave and try Dexterity (Stealth). Most combat in 5E is often resolved in 2-4 rounds, so those would be your three "rounds" for the social challenge. The greatest difference is at any point, success can "end the social combat" with victory for the PC. Failure at all stages would be tantamount to defeat in the social combat (with the PC being arrested perhaps?).
Many 5E threads have talked about the problem with the swinginess of the d20 when it comes to ability checks because of it's single-roll resolution system, but what I think often people fail to understand is that "failure" in one roll doesn't necessarily mean failure in the social encounter. We know failing an ability check could represent "success at a cost or with a setback" or simply failure to make progress, in which case you can try again.
[
In my own game, when you fail at an ability check (or repeated saving throw), the DC increases by 1. If the DC gets to a point where even a natural 20 fails, you fail. There are no further attempts and no success at a failure possible. If at any point, you fail by 1 or 2, then it could be success at a cost or with a setback.]
You can elaborate further on the first "attack" of Charisma (Persuasion) in the example above by implementing a 3-strikes policy, similar to death saves. Something like if you win three contested checks against the guard, you talk your way past. If the guard wins, you have to try a new tactic (such as Intimidation), with the chance of another win by getting three successes before three failures. You could even implement a natural 20 counts as "critical" and is an automatic win, while a 1 could count as two failures (a la the
death save concept).
BUT the issue (IMO anyway) with things like this is it is just a bunch of dice rolls. I would prefer players role-play the social encounters, and leave the "roll-playing" to the combat. Yes, you could expand features to enhance these rolls just like we do in combat if you wanted to, and then you can spend an hour on a social challenge instead of a combat encounter and award XP for it as well.
