White’s queen can make a protected attack on g7, but the square is guarded by Black’s knight. With his queen placed so aggressively White looks for other ideas and reflects that if he could get a rook behind it, the queen would be able to penetrate to h7 and be secure against capture by the Black king—the only guardian of the square. Observe the weakness of Black’s formation; see what potential for trouble arises when the king has no freedom of movement and is the only guard of a square next to it. Indeed, if White could play Rh3 now—in other words, if there were no bishop on c8 protecting the square—Black would have no way to avoid Qxh7# a move later despite having a free move to use to address the threat. (He could try Nxf6, but it's met with g5xN.)

So Black’s bishop far away on c8 is all that prevents White from finishing the game. The bishop is buried too far within Black’s position to enable White to capture it or drive it off, and it protects nothing else that White can menace. Again, though, when you see a long line between a mating square and its guard, consider whether you can obstruct it. White looks for ways to get a piece or pawn onto the light-squared diagonal and finds Nf5—blocking the line with protection from a pawn. Consider the possible consequences:

(a) If Black plays g6xN, the diagonal now is stopped up with a Black pawn; White plays Rh3 and mates momentarily on h7.

(b) If Black plays BxN, he avoids the mate but suffers other losses. White replies with e4xB. Now Black has to find a new way to prevent White’s rook from moving to h3, as it remains a game-ending threat and no longer is prevented by anything. Since White’s e-pawn has moved, Black can play QxR. But of course then White has BxQ, winning a queen for a rook. That is the true payoff of the sequence for White: not mate, but a gain of material as a byproduct of a threat of mate.