I assume you're talking about computer analysis.
Just a bit of a heads-up; you shouldn't rely upon computer analysis. You should analyze your own games first, then use a computer engine to point out any mistakes you made in your analysis.
All the engines I'm aware of assess the score of a chess position in terms of pawns. Thus, if, at any point in the game, you see +1.00 on the score read-out, then White is winning by a theoretical pawn. This doesn't necessarily mean White is actually up a pawn in material; White might be down in material but have good positional compensation, or White may be higher in material but have a terrible position. But, all else being equal, +1.00 equates to one pawn.
There's no such thing as a good mark at the end of a game. All calculations are measured in terms of the game itself. If the end of the game carries a +0.00, then the position is dead equal (either because the game is completely drawn, due to perpetual check or insufficient mating material or what-have-you). If the end of the game carries a +10.00, then White is crushing (and usually an opponent will resign by this point). If the end of the game carries a score of MX, then there is checkmate in X number of moves (so M3 equates to checkmate in, at most, three moves). If, on the other hand, a game carries a -5.00, then Black is crushing. Negative numbers indicate an advantage for Black, positive numbers indicate an advantage for White.
The way the marks are calculated is quite complex, and is different for each engine. Computer Chess Compendium by David Levy is (probably?) a good resource for learning more about these things (I have it sitting on my shelf, but haven't gotten around to reading it yet).
As for a 'good mark per move,' it's important to note that the engine does not grade individual moves, but rather the positions that result from particular moves. Suppose you have the White pieces, it is your move, and the engine assesses the position at +3.00. Now, suppose you make a move, and the engine assesses the position at +1.00. According to the computer, your move was 'bad,' and resulted in the loss of much of your advantage. But, perhaps your move was actually the second best in the position, and the best move (which would leave the position at roughly +3.00) involved an incredibly complicated tactical play which no one short of a Tal would even think of. In light of this, no one would say that your move was a 'bad one;' it wouldn't get any question marks in the annotation or anything, even though according to the computer you lost most of your advantage.
Don't take anything the computer says too seriously. It's a great tool to check for tactics, though. When I analyze and annotate my games, I go through first by myself (usually taking about an hour per game, if it's a longer match of 40+ or so moves), taking note of the positional and tactical considerations. Then, when finished, I go back through with an engine to check up on the tactics. Sometimes it'll point out that I missed a mate in four, or a simple mechanism to pick up a pawn. Then, I add in these variations, making sure to note at the end that it was the computer that found these things (so that, when I go back over my analyzed games, I can sort out between what I found in the game, and what the computer found, so as to better understand and separate the areas I need to work on). Don't let it be the primary brunt of your analysis, but it is a valuable tool.
I notice during analysis that I am given points for each move. What is considered a good mark at the end of a game ? And how are these marks calculated exactly ?
Whats a good mark per move as well ?