It recently dawned on me what made the crafting system in A Tale in the Desert so compelling.  It wasn't just that there was a huge variety of things to make, though that did help.  It wasn't that the crafting processes for so many different things were so wildly different (including many different mini-games), though that also helped.

The key was that you had some very complex choices to make concerning what you wanted to craft that day.  The game never handed down an edict that you had to craft 50 of something to proceed.  And it never made you craft 50 of some stupid item that no one would ever want.

What you wanted to craft could depend on what you wanted to use, what a potential trading partner wanted to buy from you, which crafting processes you enjoyed, which crafting processes you were good at, which processes you had the infrastructure set up to build efficiently, and so forth.  There was enough complexity that it was basically impossible to turn it into a spreadsheet game and determine that this particular approach was the most efficient crafting that you could possibly do.

And that is the key step at which crafting in most games falls completely flat.  When a game says that you have to level crafting by producing a bunch of something stupid, making the decision of what you will produce is at most one interesting decision, even if you need to craft 100 of the item to grind enough levels.  And it may constitute zero interesting decisions, if it's obvious which item is the most efficient way to grind levels.

Having a minigame doesn't really help that much here in itself.  Vanguard had a minigame for crafting, for example.  But leveling crafting meant that you had to play that one particular minigame endlessly.  And it wasn't a terribly deep minigame, either.  Having to craft whatever crafting leveling quest mechanic some NPC demanded (I forget exactly what the game called it) a bunch of times in a row was not an interesting decision.

It's not that minigames are bad.  A Tale in the Desert got a lot of mileage out of the fact that some people were simply better at some minigames than others, and also liked some minigames better than others.  That meant that the optimal things to craft for one person were often very different from the optimal things for another, even if they wanted the same final products.  But minigames alone aren't enough to make crafting interesting, unless you like playing the minigames for their own sake and don't really care what you craft.

Similarly, if it's glaringly obvious which final crafted products you want, the only possible interesting decision is whether to craft it or not.  If a game offers you one crafted armor set at level 10, another crafted armor set at level 20, and so forth, then it doesn't really matter whether the armors are useful to you or not.  If you craft, then probably you're going to craft them, and if not, then you won't.  Either way, there are no interesting decisions to make.

Some people say that a key part of crafting is that crafted items have to be useful.  And they're correct in the sense that if everything crafted is worthless, then there aren't any interesting decisions to make.  The most efficient crafting choice is to craft nothing.  Making the final crafted item absolutely essential for you to have doesn't make it into a good crafting system; it only makes it into a terrible crafting system in which participation is mandatory.

But merely wanting a lot of crafted items is not, in itself, enough to make a crafting system any good.  If all that crafting consists of is grinding something stupid to finally get the item you wanted with no meaningful decisions to make anywhere along the way, then that's a terrible crafting system.

The key to a good crafting system is that a player should constantly be faced with interesting decisions of what exactly he wants to craft.  It shouldn't be completely obvious that you craft a bunch of this, and then a bunch of that, and so forth.  Ideally, a player should be faced with interesting decisions to make every single time he wants to craft something.  One decision of an item that you want shouldn't dictate the next hundred things that you will craft.  Ideally, it would only ever dictate the next one thing that you will craft.  For technical reasons, that can be hard to do, and it's understandable if a player makes one decision to craft 20 of something in bulk.  What's not understandable is if the player has to craft 100 of something stupid that no one wants, but he needs to do it anyway in order to grind levels.

What is egregiously bad is when a game breaks its crafting materials into tiers.  You've seen how this goes:  you craft copper items up to level 10, then bronze to level 20, then iron to level 30, then steel to 40, then mithril or whatever.  The details of what the particular tiers are called varies by game, but many games have such tiers.  That takes most of the depth out of crafting, as rather than potentially wanting to craft absolutely anything that could be made in the game, the overwhelming majority of the items are automatically ruled out for you as being not in the right level range.

That's not really all that far shy of completely killing any potential that a game had for interesting crafting decisions, as trying to make that interesting means that you need not just one good crafting system, but one good crafting system for each level range that is completely independent of all others.  If most game designers can't come up with one good crafting system, then expecting them to come up with 6 or 8 independent ones that are all good is not likely to end well.

It has been said that fun gameplay is all about giving the player a lot of interesting gameplay decisions.  In combat, we recognize this.  Pressing buttons to attack is not fun in itself.  Deciding which buttons to press in which circumstances in order to use the skills best suited to what we need at the moment can be fun, however.

But the same is true of crafting.  A fun crafting system is all about giving the player a lot of interesting gameplay decisions.  The critical one that players need to be given often--and without also giving the player an obvious, correct answer--is:  what do I want to craft today?

For me it was that ATITD's crafting was a minigame.  It was a game of figuring out how to keep many productions spun up simultaneously to achieve the highest possible rate of production.

A robust item economy certainly helps, but it's the moment-to-moment gameplay of a game that I'm playing.  If that isn't enjoyable, the entire thing falls over.

I think the only exception to this would be super short-session games, which makes me think the design rule of thumb would be "Keep it fun or keep it short."  Sort of like a Travian-style game which didn't bloat into massive overload, but stayed at a manageable Clash of Clans style session length.  So even though the sessions wouldn't involve super interesting gameplay (just tapping on what building upgrade to start next), at least you aren't forced to sit there being bored (EVE mining/traveling.)