Making gold from the Auction House is all about managing risk.
The risk you take is that your auction will not sell, making you lose not only the deposit fee but also whatever time it took you to list it to begin with. Sometimes your goal is not necessarily to sell the item at all, or perhaps the profit margin is big enough by itself that infrequent sales is not much of a risk as long as one or two make it through. Your risk of a non-sale increases for each competitor in the market with an identical good, and increases further still when you get undercut.
When Not to Undercut
You do not actually have to always undercut the competition, or otherwise cancel your auctions and repost. In a demand-driven high-volume sales market it is entirely possible that buyers will chew through the lower-priced competition's wares and still be hungry for more. The risk you take by not undercutting is that there will not be enough sales within the listing period for your auctions to be on top (and then get sold) - you can also be burned if your competition lists additional auctions throughout the day. Of course, if your competition continues to list goods throughout the day, your undercutting is not going to help anyway.
An example of a demand-driven high-volume market: Mageweave Cloth. The people powerleveling their First Aid are going to need 4-6 stacks while the Tailors are going to need ~20 stacks. Considering that not everyone levels their First Aid as they level their character up, the demand pressure for Mageweave is always there at every level (e.g. a level 40 and a level 80 toon could both need it). While the same could be said of Wool Cloth or Runecloth, Mageweave is special insofar as it is much more difficult to farm as the dungeons in which it drops are out of the way for most max-level characters.
MFCs are also fairly high-volume although they are not entirely demand-driven as they are pushed.
When To Undercut, and How Much
There are generally two approaches to undercutting and using either depends on the circumstances and your expectations.
The first approach is the mechanical undercut. This is when you let addons like Auctionator just undercut the lowest-priced auction by X amount, usually a few silver. The benefit of mechanical undercutting is that it places your auctions at the front of the line while maximizing your potential profits for each sale. The downside? The competition you are undercutting can undercut you by canceling/relisting for the price of the deposit fee + the few silver. In essence, by undercutting by the minimum amount possible, you are taking the risk that you will snag some sales before your competition returns the favor.
The second approach is the savage undercut. You just put up some Bold Inferno Rubies on the AH for 250g because you happened to be the only seller at the moment, and here comes Noob B. Esq. with some of his own Bold cuts for 150g. Why a 100g undercut?! Does he not understand that he could list them for 249g and basically make an extra 99g? Hell, at that price I could buy them up myself and relist them for...
...hook, line, and sinker.
Fundamentally, the savage undercut model is used when you want to guarentee sales - the lower your price, the more of the demand triangle you get on those Supply/Demand charts, which increases the likelihood of a quick sale. The irony here is that if you undercut savagely enough, the buyers actually start to include your competition. If your competition wants to give you 150g for a chance at making 100g, I say let'em. In fact, I insist! Especially when I am getting the uncut gems at 80g or below.
Unless you are cashing in your chips and exiting a market, the key to savage undercutting is to understand your own profit margin independant of what else is going on in the AH. Would you be happy buying uncut gems for 80g and selling them for 150g? That is a 70g margin per gem. If you see someone selling them for 250g on the AH, that potential margin increases to 170g. The key word here is: potential. If you know the gem market and know the market price is 150g, then you also know the likelihood of getting those gems back in your mailbox minus the ~4g deposit if you start pushing 250g apiece with competition afoot is higher.
While the gem deposit fee is big enough to discourage casual cancelling/reposting, the same cannot be said of other goods, like Glyphs for example. If you show up at the AH with a handful of glyphs and start undercutting the 80g ones at 79.99g, the Glyph Barons lose around 3 silver by undercutting you (their time is irrelevant, since they are already committed to the abysmal time sink that is selling Glyphs). If you show up and start savagely undercutting their glyphs, say posting them at 50g, then... well, you will still probably get undercut. The point is that the Glyph Barons aren't losing 3 silver by undercutting you, they are losing 30g. Even if savagely undercutting results in no sale, what you lost was the deposit cost, while they lost nearly half of their margin. In the scheme of things, I'd call that a win.
Have I mentioned I hate the design of Glyphs top-to-bottom?
So What is the TL;DR?
The answer is going to cost you, conservatively, around 10,000g. A month. That is approximately how much you stand to lose by A) never undercutting, B) always undercutting by 1s, or C) always undercutting by 50g+.
If you missed your Ritalin dose for today though, here is the quick and dirty version. Don't undercut in fast-moving markets like Dust/Cloth OR when there are just 1-2 severally discounted auctions compared to the normal, more reasonable market price. Undercut by a minimal amount when it is unlikely your competition is going to bother canceling/relisting, either because they aren't hardcore auctioneers (e.g. normal players who just check their mailbox 48 hours later) or because the deposit fee makes a 1s undercut from you cost them 1g or more. Savagely undercut in slow-moving and/or high margin markets to guarentee sales.