The usual color scan reveals a potential knight fork to be had at d6, but the square is protected. How many times? Twice—by the bishop at f8 and the knight at f7; be sure to account for all the guards, not just the first you notice. Fortunately White has bishops attacking each of the two bothersome pieces, but there still is a complication: when White’s bishops take Black’s bishop and knight, Black will use his king to recapture, and the king is supposed to be one of the pieces in the fork. Will its recaptures ruin the forking opportunity? Not necessarily; so long as the king ends up on f7 it still can be forked. But this means that it is important to perform the exchange on f7 last so that the king ends its travels there. Thus 1. BxB, KxB; 2. BxN, KxB; and now Nd6 forks king and rook and so wins the latter.

Naturally Black might prefer to bow out of this sequence earlier, giving up a piece rather than stepping into the fork. That's often how tactics work, as we will see many times (but won't always point out): the victim can escape final execution of the fork or other idea, but only by making a sacrifice. In that case—which is normal—the tactic still must be counted a success.

The point to take away from this example, apart from the importance of accounting for multiple defenders, is that the order of operations in a tactical sequence can matter a great deal. Here the tactic doesn't work if White takes Black’s knight first and his bishop second. When you consider a sequence that involves more than one exchange, ask whether changes in the order of the moves would make a difference to the outcome.