I think there are two important parts with remodel.
1) Never buy a lot of them (1 is most often the correct number)
2) It is best when there are good cards at all dollar values. This is important, because remodel is best when your hands are flexible.
To expand on point #2:
Lets say a "good" kingdom is:
$2 Pawn, or hamlet, or cellar
$3 Silver is good enough here, but better $3's like fishing village would help.
$4 Caravan (for example)
$5 Laboratory (for example)
$6 Gold is good enough here
$7 Expand (for example)
a) If there is no good $2:
Now your hands of CCCCR are really bad. You'd rather have a copper instead of that remodel so you could buy a lab, and would certainly rather have bought a silver so you could gain a gold here.
a) If there is no good $3:
Unlikely, there is always at least silver
a) If there is no good $4:
This is obviously bad because you cannot remodel estates into a $4. You have to remodel them into a $3 (silver), which is still okay, but slower nonetheless. Possibly the worst thing about not having a good $4 is that you'll fall into the trap of remodelling an estate into a remodel (i.e. the trap of gaining too many remodels). This is exaggerated even further if the good $5 is a terminal action, since you'll clog yourself. If the $5 is a lab, gaining a second remodel isn't the end of the world, but gaining more than 2 remodels is probably a net-negative.
However, this is also vitally important for another reason. Remodel decks will play with 3-card hands on remodel turns, and those 3 cards generally tend to produce $4 quite often (i.e. remodel/estate/copper/copper/copper or remodel, estate, estate, silver, copper).
Having no good $4 to buy (besides remodel itself) is crucial. For this reason, a good $4 is probably the single most important thing, even more than a good $2.
a) If there is no good $5:
This is critical, because of a few reasons. First, you cannot remodel silvers into a solid $5, and remodel decks will buy quite a few silvers with their 3-card hands.
Secondly, your remodel turns that produce $5 with three cards (silver/silver/copper) in the midgame are that much worse.
Imagine a hand like Remodel, Estate, Silver, Silver, Copper. If there was a good 4 and 5, you could remodel the estate into caravan and buy a lab. If there is no good $5, you are forced gaining two caravans, which again, it's not terrible, but it's weaker.
You can always buy duchies here, but that's generally sub-par at this point (midgame) of the game.
a) If there is no good $6:
Again, gold is good enough here.
a) If there is no good $7:
Now you cannot remodel your $5s into Expands. For example, having the ability to take a Gold/Gold/Silver/Lab/Remodel hand and buy a province and expand is pretty huge, especially on a board wihtout +buy. These types of hands will start popping up in the endgame, and your $5 is not going to help you do anything, since you already have $9, so gaining value out of it here is great. That expand you just gained will allow you to gain a virual +buy on this board that could be critical to reaching a 5-3 province split.
If the $7 is a kings court, a king's courted remodel is pretty darn solid for obvious reasons, especially on the last turn of the game where you can potentially gain 3 VP cards in 1 turn by converting your lesser treasure/actions into VP.
Access to a good $7 is nowhere near as important as having a good option at 2, 4 and 5, but it's still a factor.
For colony games, without a good $7, you cannot remodel a $7 into a platinum. But then again, I think that remodel is going to be sub-par in colony games.
In summary, I think the thing that remodel wants the most is to have solid options at each dollar value, so that your hands are as flexible as possible. This is mostly important in the $2, $4 and $5 range (less so for $3, $6 and $7).
And again, this is for two reasons... firstly, because it's important to be flexible on what you remodel something into (to give all your remodels flexibility on both what they gain and in what they remodel). But also secondly, to be flexible on what you spend your 3-card turns buying. If there's no good $4 card, and you draw Remodel, Estate, Copper, Silver, Copper, you have a sub-part turn where you are likely gaining two $3s, whereas if you bought silver instead of your remodel here, you could have gained a gold instead.
Gaining two quality $4s here beats gaining a gold (taking into account not drawing the estate any more), but gaining two $3's is worse (at least I would think, even considering losing the estate from your deck). Regardless, gaining two $4's is maximizing the power of the strategy, so being flexible and having good options, especially at 2, 4, and 5 is super important for remodel.
