Here are some of the identifiable goals of gems - in other words, here's why Blizzard likes and wants them:
1 - They create a long-term goal for the player. A perfect gear set will include socketed gems of the highest level. Gems take a very long time to acquire, and will therefore keep a player playing. This creates retention. Player retention is important due to the fact that Blizzard makes money from the use of the RMAH. Having a player come back increases the chances that this player may make use of the RMAH.
2 - They create a gold sink. Gold sinks exist to combat inflation. Too much inflation is as dangerous to a game's economy as it is to a real economy. If gold becomes too abundant, the overall price of items will increase. The rich get richer - this would prevent new players from taking part to the game's economy, potentially turning them away from the game, a detriment to the factor of player rentention noted above. Another result of an overflowed economy would be the trivialization of other gold-reliant game mechanics. If repairing no longer feels like a burden, dying will lose its impact. Potions will be purchased without limits, potentially trivializing some content, or creating a dependence, impactful for balance and flow, as players would likely sit around waiting for cooldowns. Artisans would also become proportionately quick to level, with much gold to dump on them, but little materials to actually craft, making the system appear flawed (it is, but we'll be back to that...).
3 - They create additionnal items to feed RMAH. High level gems will be a rare commodity, and should be able to fetch very reasonable prices on the RMAH. This is obviously advantageous to Blizzard, as they profit from a large % from such sales.
For these effects to take place, players must obviously have a desire to use gems. This desire has multiple facets, and skipping on one or many of the following points has very high chances of compromising the success of such a gem system.
I - The gems must exist. Obviously. No gems, no system, no results.
II - The gems must be rare enough so that players are willing to purchase them through either AH, yet they must be common enough to provide an actual supply. With too much supply, the price of gems dwindles, and money may not be generated through the RMAH. With too little supply, sales will be too rare to be impactful. Finding the right balance is extremely difficult; establishing drop rates is one of the few design ideas I have never explored, so my insights in regards to such a subject are limited at best.
III - The gems must be desirable. If players feel like socketing gems has too limited impact on their power, they won't go through the trouble of picking them up, upgrading them or selling them. This point is key.
IV - Upgrading gems must be rightly priced. The irony of a gold sink is that it will not be effective if it is too little, and neither will it be if it is to elevated - the right price must be tagged.
Now, to the meat: What are some of the issues with D3's current gem system?
A - Insane costs and time investments. Currently, the highest gem level to drop is the 8th - the flawless square. There is a total of 14 levels. To jump from one level to the other, you need 3 gems of the inferior quality (3:1 ratio), as well as a certain amount of gold, and of tomes of secrets world drops.
Therefore, in order to craft one level 14 gems, you need to find 729 level 8 gems. Which turn into 243 (9); which then turn into 81 (10); 27 (11); 9 (12); 3 (13); 1 (14). And, whenever you turn 3 gems into one, you must pay a certain fee.
Turning 3 level 8 gems into 1 level 9 gem costs 30,000g and 3 tomes of secrets. So, to merely turn your 729 gems into the next rank, you find yourself paying a whooping and 7,290,000g and 729 tomes of secrets.
Shall we calculate the total cost of merely upgrading to one perfect star gem, just for kicks?
8 => 9: 7,290,000g + 729 tomes
9 => 10: 4,050,000g + 486 tomes
10 => 11: 2,160,000g + 243 tomes
11 => 12: 900,000g + 108 tomes
12 => 13: 600,000g + 48 tomes
13 => 14: 400,000g + 20 tomes
So, for a grand total of 15,400,000g, 1634 tomes and 729 flawless square gems, you can own your very own perfect star gem!
Given that you can have up to 12 sockets in your gear, we're talking about 184,800,000g, 19608 tomes and 8748 flawless square gems to perfectly deck out one of your characters.
And this is without considering the costs of upgrading your jewelcrafter, let alone that of purchasing the top end recipes.
Sure, that sounds like a lot, so far anyway. Things always have to be put into perspective; who knows how much gold we'll have in a year or two? I cannot accurately comment on whether or not such costs are going to scare away players, so we'll just assume they won't - for now.
Even if they did, though, the following point, I feel, is a much bigger factor in terms of raising questions in regards to gem viability.
B - Sockets take up an affix space. What does this mean? Assuming perfect conditions, you'll get to choose between an affix and gem(s) socket(s). In the case of single-socket-possible slots, such as rings, amulet and offhand, the choice isn't very difficult - you probably don't want a socket there, as the bonuses are... say, sadly quite bad, and do not remotely make up for losing on an entire different affix unless you're really looking into specializing. As for pants and chest armor, well... again, the vitality affix can go up to 200 on its own, while a level 14 gem grants a measly 58. So, again, unless you already have the stats, and you're really looking to specialize... probably not worth the 30-45 mil.
The helm and weapon slots are different; the bonuses they provide are considerably better, to the point where you would actually want to replace an entire affix with the bonus provided by a gem. Which *gasp* actually makes sense for such a monetary and time investment.
Why is this ordeal problematic, though? Well, think about it - If a regular affix is better than a socket, who the heck is going to invest 15 mil, let alone 30 or 45 filling just to make an affix almost good? No one, let me tell you! And as explained above (1, 2 and 3), the benefits of having players hunt gems are voided if, well, they don't feel like they should hunt them.
In short, for 10 out of 12 possible sockets, gems don't feel like a bonus at all - in fact, they're a 15-45 mil malus if you get a great item roll with sockets taking the space of another affix.
How can this be solved, so that the benefits of having gems, illustrated at point 1, 2 and 3, can be reaped?
X - Quite obviously, sockets need not take an affix space. As it stands, gems do not feel like a bonus.
Y - Alternatively, the power of gems could be improved so that sockets are not only on par with power but over with regular (good; not improved pick up range) affixes. Why over? Because you need to pay and invest time for them. That simple. As of right now, there's no reason to stress over gems as they are expensive and weaker than other affixes.
Z - Alternatively again, allow gems to bring unique affixes (which may not spawn on items) to the table. This alone could not only bring interest to gems, but also, dictate new interesting builds.
By chatting with other players, I came to remember that the Blacksmith artisan was initially planned to be able to add sockets, as opposed to having them randomly appear as an item affix.
It would appear that in the rush to release the game, they might not have had time to add and / or balance, such a functionallity, opting rather to have sockets as an affix, perhaps foreseeing an uproar at the suggestion of entirely removing gems... so they went ahead with it anyway, without buffing gems to affix levels, without exactly thinking of the consequences of omitting to do so.