Too often I think that folks have a tendency toward an unwarranted parallelism as it pertains to application simply because of parallel structure.
Here's a simple example. You open One Spade, and partner makes a splinter response of 4C, 4D, or 4H. What do these calls show?
Well, a parallel structure (often needed for memory reasons) will likely mean that:
1. 4C is a splinter, support in spades with shortness in clubs
2. 4D is a splinter, support in spades with shortness in diamonds
3. 4H is a splinter, support in spades with shortness in hearts
Easy, right? So, what's the problem?
The "problem" is that parallel structure does not equate with parallel application, or at least it should not. In practice, it often does, but that seems to be a mistake.
Consider that within what appears to be a "parallel structure" there is an unparallel circumstance.
1. After 4C, Opener has two bids available below game as slam tries -- 4D and 4H. Two steps of available space actually means three levels of inquiry. 4H as a slam try, 4D as a slam try, and 4D as a slam try with a 4H re-try. Whatever these sequences mean, three levels of inquiry are available.
2. After 4D, one level of inquiry is available -- a 4H slam try.
3. After 4H, there is no level of inquiry below game.
If you consider a 3S splinter in support of hearts, the levels of inquiry atre even higher. 3NT, 4C, and 4D are available. 3NT can be rejected outright, sent back as 4C, sent back as 4C with space for 4D, or sent back as 4D. Hence, seven levels of inquiry exist (I think).
Thus, a one-under splinter leaves no below-game inquiries. Two-under gives one. Three-under gives three. Four-under gives 7. 0, 1, 3, 7.
This should start the thinking that the lower the picture action, the more flexible the picture action can be. Conversely, the higher the picture action, the more pure the picture should be.
This translates, then, into alternative sequences. Suppose, for instance, that partner opts for an alternative 2/1 sequence and ends up revealing a hand that seems to have been potentially suitable for a splinter. If the splinter would have been a one-under splinter, then his range of possible holdings for the alternative 2/1 sequence is higher, as he might be just shy of "pure splinter." If, alternatives, his 2/1 sequence reveals a plausible splinter candidate that would have been a four-under splinter, then his hand likely does not resemble a pure splinter at all, despite the shortness. The difference is probably tactical (meaning, the likely sequence after a four-under splinter could not have revealed the nature of his hand, so he has some feature or feature not otherwise describable via the splinter).
So, when you design a system and opt for a parallel structure, recognize that the goal of parallel structure is self-defeating, in a sense, because you cannot effectively create a true parallel structure when unparallel circumstances necessarily arise from stepped structure, as the degree of step away from a decision point creates necessarily unparallel application. Cater, then, to that unparallel application if you want to design an effective system.