I used to compose a list of actions they performed in what situations and then subjected them to a socionics analysis. This usually involved drawing parallels from their actions to the socionics IEs or descriptions. Aside from being very shady and creepy to have a big database of this, it could take hours if it was complicated enough.

 

Nowadays I recognize how totally inconsistent and wrong human memory is, so that database was really just a subconcious projection of my prejudice. I feel as though interactions with people are like dreams; the further in time you are from the experience, the further you will be from understanding. So I believe in speculation in the moment/right after, but not rushed. I've been trying to come up with a system based on the IEs to make this process easier, which involves reducing the IEs to very abstract concepts applied very indirectly by people in conversation; i.e. he's taking the initiative and speaking very confidently about this, what function is he using (how is he explaining it)? Now this other one is responding to him; how do they relate and is this a relation of functions? It's really intense!

Occasionally I picture everything I know about that person in my mind, trying on different types until one 'fits'. I really need to know them well enough to predict their behavior like that though.

 
A list of benchmark types; comparison is a great way to type someone! What bothers me most about it is that you could have been wrong in the first place, making it self-defeating...
Being around the person for a long time; anything you speculate before this is unreliable. I honestly don't think you can reasonably type anyone if you haven't spent extended periods of time with them, i.e. several sittings of several hours +.
Being around socionics for a long time; By seeking out new materials and reviewing/discussing the old ones I feel I have a more fleshed out version of the IEs.
Getting to know the person.. personally; 1on1 time will give you insights to what they're good at/what they like to do and a general overview of their life if you get that far in conversation.

 
Trying to explain a type based on intertype relations; Typing should be clear and not require abstract explanation. You could be wrong on the other guys' types too.
Typing based on personal feelings; your typings will change as often as your feelings!
Taking another's typing for truth; it ain't true till you've gotten your own hands dirty. I've friends who are just bad.

In general; Vibe typing bad. Don't rely on more than one person to type one person.

Objective typing;

I think objective understanding can be achieved to an extent; instead of blowing the IEs out of proportion in order to compare them to people's actions, I try to simplify people's actions until they compare to the fundamental basic functions. This is why I disagree with removing the essential basics from the functions; though a person's actions may not be forceful, once simplified and examined at their most primitive level they could resemble force.
Clearly, using this method, the problem no longer lies in the interpretation of the functions but rather the examination and simplification of another's actions, which leaves much less room for error imo, especially if you understand the person and their motives.

@words; I held your position a couple months ago; that socionics lacks any application and that it may just be imagined. That's why I tried to learn to apply socionics at its very basic level that left no room for interpretation of the functions but still works.

TBH; you'll never be able to get this crap out of your head so best learn to live with it... that's what I told myself anyway.