Par in a word ladder is the length of a shortest route between the start and the target, usually computed by breadth-first search over the legal-word graph. That is the whole definition, and the interesting part is the word usually.
Breadth-first search finds a shortest path if one exists, and in a dense graph it is fast. But the result depends entirely on the word list. Allow plurals and obscure forms and par gets shorter. Restrict the list to common words and par gets longer, sometimes by two or three moves. So a par number is only meaningful next to the dictionary it came from.
This matters for players in two ways. First, comparing your move count to par is only fair if the solver used a dictionary at least as generous as yours. Second, a puzzle with a hard word list and a low par is telling you that a rare word is required, which is a hint in itself.
I use par as a weather report rather than a target. A par of four on an awkward pair of words usually means there is a clean central pivot. A par of six with a short start-to-target distance means the neighbourhood is narrow and the route is forced. Both are useful, neither is a score.
If you want to see where a par number comes from, the solver page returns a shortest route for any two words, so you can inspect the pivots rather than just the count. The answer archive shows the published routes for past days if you want to compare the official chain with a computed one.
One practical habit: before you read a par number, guess it. Look at the start and the target, estimate the distance, and write your guess down. Over a few weeks your guesses get close, and that calibration is a much more useful skill than knowing the exact number, because it is the same instinct you use when deciding whether a route is worth trying.