I've been working on several Keynote presentations in the last year or so, and one very interesting feature of Keynote are its "builds". A build is essentially a way to show/hide/move an object or group of objects. But if your GUI pain threshold is sufficiently high, you can do great things with them.
To create a build, select any object on the current slide, bring up the inspector panel and click the "Build Inspector" icon. Click the "More Options" button at the bottom to see a chronological list of all builds on your slide pop out in a drawer.
There are three kinds of builds, each under their own tab: "Build In" decides when and using what effect your object will appear, "Build Out" when and with what effect it will disappear, and then there are "Actions", which let you move/rotate/scale and change the opacity of an object.
When you specify anything on one of these tabs, a new entry will pop up in the list of builds, where you can drag it around to control in what order it happens relative to the other builds. By default, the next step will only happen when you click, but you can set them to happen automatically after the previous build has happened, at the same time as the previous build happens, or (for the first item) right after you open the slide.
This may not seem like much, but actually is all you need to make very complex animations. I've made dropping domino stones, showed numbers being transferred from one place to the other, replacing an existing number and much, much more (for examples, see any of the movies in my Masters of the Void C Tutorial, for example in the Memory Tutorial). Not to mention that Keynote's seemingly simple user interface actually contains a beautiful and rather powerful vector drawing program.
Another not quite obvious feature is that you can also set up delays when a build happens after a previous build, and you can specify the speed for the various effects, and in the Slide Inspector you can even specify that it should automatically advance to the next slide a certain time after all builds have finished, not just upon a click, with a transition effect (the Magic effects really deserve their names), or without one.
Depending on how complex your animations are, you can keep things manageable by duplicating the current slide and then doing half your animations on one slide, and the other half on another, which reduces the number of overlapping objects you have to fiddle with. You can also group items and animate them as a single unit on one slide, then have them separate on another slide to animate only part of them.
Another way to cheaply get great-looking animations is using effects on text: You can set up many of the effects so they affect each character or each paragraph separately. With a little fiddling with paragraph settings, you can even make things look as if each line was transitioned separately. And of course the "Insert" menu contains Smart Builds that do some really neat effects quickly and easily.
And don't forget that you can use this for more than Keynote presentations: Just use the "black" or "white" theme, set up a custom Slide Size in the Document Inspector, do a "Record Slideshow" and then export to a QuickTime movie without sound, and voila, you have a great-looking animation for a screencast, web site or other opportunity.
The main difficulty with complex builds is that, instead of a timeline, where you can select the current point in time, and people will only see the objects visible at that time in the locations they will be in at that time, we have this darned drawer, and objects are generally always all visible (in some cases an object is even visible in several positions at the same time). This makes it hard to hit the right spot (particularly the red dot that indicates an Action is attached to an object is a bit hard to hit occasionally, and you need to hit it to e.g. see and set the destination for a move).
Do you have any Keynote tricks?