First find the kernel of a discovery; you are trying to learn to see them every time, even when they don't appear to lead anyplace. White’s rook masks the path of his bishop toward Black’s bishop—and, behind it, his king. If Black’s bishop could be replaced with his king, White would have the makings of a discovered check. Capturing the bishop would be one way to achieve this, but again White has no way to do it; and yet again, too, there is the alternative of capturing something the bishop protects—the knight on h5. White can take it with QxN+. If Black recaptures BxQ, the bishop is out of the way and White can unmask a check by moving his rook. He looks for damage the rook can inflict with two moves and sees that Black’s queen is within reach. Rd6+ attacks it; RxQ takes it after Black saves his king. White wins a piece.
Well, but wait: how does Black save his king? He can move it (say, to g7); but he also can save it with a threat by playing his bishop from h5 back to g6, where it blocks the check and also throws a counterattack at White’s bishop. Now after White plays RxQ, Black can play BxB. This would mean White traded away a queen and a bishop to win a queen and a knight—but we aren’t quite done yet. After Black’s BxB, White recaptures with NxB and still wins a piece. So the modest-looking knight on e1 is necessary to make the whole sequence work. (If that knight were off the board, the sequence just described still would be worth a pawn, though; after Black plays BxB, White would have Rb5, forking the loose pawns on a5 and c5.)
The lesson of this last note is to take time to consider how your opponent will reply to the checks you make. If it's clear that he will have to move his king, it may not matter too much where it goes (though then again it might!). But if he has interpositions, you need to anticipate what they would make possible for him on offense.