There are many considerations:
[1] Space: Space is the #1 most important key. Even if you're working in the outdoors (i.e. have unlimited space) your design will still take up space and the design should account for that. If your factory is too large, it might become to tedious to find errors/tweek your system. In my case, where I have limited room, space becomes an even bigger problem. It is more of an issue of trying to design a system that its space meets its purpose. If your system is needed to attach to other system, you may want to locate it in a central location such that it is easily accessible. Space includes (1) the efficiency in which it takes up space as per design (design it in a way that it isn't wasteful of space) and (2) is externally and internally accessible, such that you can access the innards should you need to make small changes and to allow for easy integration to outside systems.
Compact Design:
Underground Base:
[2] Cost: This should be obvious, but optimizing a design such that it would give the "best bang for buck" is more involved than you might think. Depending on how you acquire resources the value of a given resource may vary. Including the time-varied acquisition of the resource as per your strategy (quarry, turtles, etc.). Examples of such considerations would include whether you wish to integrate your AE system into the system you are considering. AE materials aren't excessively valuable, but require time to accumulate. Projects that utilize a high volume of iridium should be optimized such that they utilize the blocks using these resources to its maximum. This also means designing the system such that you can achieve such optimization through testing/tweeking. Synonym: Non-Wasteful.
Very Efficient Utilization of Lasers:
[3] Integration: A very important aspect of a system is whether it can be successfully integrated into your base. This goes back to accessibility as well. If it is located in a poor location, getting resources to that system might be cumbersome/expensive. Having items floating in transport pipes should be minimized for any case, as to keep the load on your computer as low as possible. If your system doesn't really provide much to enhance other parts of your base, its existence would need to be justified that much more. Usually I build my systems to synergize with one another. Synonym: Versatile/Purposeful.
[4] Scaling: A system should be able to grow as your needs grow. Whether it be increasing your energy production or your item processing, your systems should be designed in such a way that it can be expanded. This means leaving extra space or incorporating the ability to change your system such that it is more efficient. The best way to achieve this is to modulate your system. Every time you wish to expand your system you build another module. You know exactly the resources required to build each module and exactly its output. This is extremely useful in the process of design. Synonym: Upgradable.
Version 1.0 Lava Gen Module: (40 Combustion Engine -> 4 Magma Crucible)
(These are side-by-side modules of the first picture, two layers)
Version 2.0 Lava Gen Module: (4 - 36HP Boiler -> 72 Industrial Steam Engines -> 20 Magma Crucible)
8boiler
12boiler
sideview
crucible1
crucible2
thermalgenerators
These are a few of the things I might consider when building a system. Main key points are its accessibility, upgrade-ability and its cost. Space is more of an issue for me because my base is inside of a mountain.