First, I deliberately only plan one week at a time (typically this means 3 days on, 1 day off, 2 days on, 1 day off). I know some people program 4, 6, 8, or even 12 week cycles. Apart from scheduling weeks off at the 6 week mark, (which I always adhere to) I don't find this extensive of a plan beneficial for a hybridized training program because it doesn't allow you the flexibility to respond to your body on a weekly basis. How will I know 7 weeks from now how my body will be responding to the different elements of training that I had programmed? I don't.
Now, if you have decided to train specifically for strength, power, size, or conditioning, then this is a different story. In that case, due to the progressive nature of most of those programs, setting different routines, exercises, and weights for months at a time can be very beneficial. I used to periodize elements of training in this manner for 6 weeks at a time, and I saw a lot of success. However, I found it to be less exciting and less fun, overall, than a hybridized program. The jury is still out on whether one style is better than the other in terms of performance gains, so maybe it's good to try both and alternate between the two.
Currently, I subscribe to the school of conscious variation, always trying to avoid any set pattern from week to week. This is not to say I use a fish bowl method, simply pulling a workout out of a hat each day to adhere to the idea of randomness. Randomness is not necessarily the same thing as variation and it often invites injury. For example, what happens if you pull pistol squat, box jumps, and dips on Monday, max effort thrusters on Tuesday, and squat clean and jerk to burpee on Wednesday? This is not variation, in my opinion, though the method of selection was totally random. And doing a stretch of WODs like this would run the risk of over-stressing your knee and shoulder joints.
Some people try to avoid this problem by designing their programs around muscle groups, much like a traditional bodybuilding program. This is okay, but you quickly run into problems because so much of crossfit (and I think this is a very good thing) is built on multiple joint, multiple muscle movements. For example, if you a do deadlift/push press WOD 0n Monday (back, hamstrings, glutes, quads, calves, shoulders, triceps), what are you going to do Tuesday to avoid working those muscles again? Situps? Biceps curls? The nature of this type of exercise doesn't allow for disintegration of workouts into muscle groups.
So why try? Instead, I think it much better to base training around movement patterns. So, using the example of deadlift/push press from above, the predominate movements are full range hip extension (I make a distinction here because so many lower body movements could be considered hip extension to some degree) and shoulder to overhead. If I were to write a workout for Tuesday in this scenario, it might involve pullups, ghd situps, and box jumps. Many of the same muscles and joints are being used, but the movement patterns are completely different, ostensibly working the muscles in a different, or varied, way.
Furthermore, I also deliberately vary the style of workout based on energy pathway. While I know there is a lot of science dividing the different types of exertion, I try to keep it basic. (I may divide metcons into "interval" versus "just finish", but that's as exotic as I get) More generally, I try not to do strictly heavy strength, strictly metabolic conditioning, or strictly extended endurance work during the course of a given week. This doesn't mean you have to do one of each every week, just that you should avoid over-doing one element, relative to the others. This is a problem I see with a lot of Crossfitters. They only want to do workouts like Fran, Grace, Tabata intervals, and AMRAPs. While this type of programming can easily adhere to the philosophy of alternating movement patterns, it does not significantly alter itself from the perspective of energy pathway. Hence, in my opinion, the program is not varied. (This becomes more like the periodized programming mentioned earlier. I might get really good at metcons after a while, but that's all I'm going to be good at until I change my focus. And, if you believe in the idea of "specific adaptation to imposed demand," the rate at which I improve will most likely diminish over time as my body becomes accustomed to what I ask of it.)
Returning to my earlier example, if Monday's workout looked something like: 21, 15, 9 deadlift/push press, Tuesday might look like: weighted pullup 5 x 1 (rest as needed), box jump for height 7 x 3 (rest as needed), and Tabata rounds of GHD situps. In this example, the only element of the workout that could be considered in the metcon energy system category that was challenged on Monday are the Tabata situps. A Wednesday workout, to play it forward, might be: run 1 mile, 5 rounds of 10 pushup, 10 squat, 10 burpee, run a mile. The emphasis here, obviously, is more on extended endurance.
Having these criteria (movement pattern and energy pathway) to build workouts relative to the previous one is essential. But this is not always the problem people have. Often, the problem is where to start. I think the best way to solve this is by having a cornerstone for your week. For example, the cornerstone in my training right now is heavy Olympic lifting. This is because the totality of the lifts are so beneficial for other elements of fitness (strength, power, coordination, balance, accuracy, flexibility) that they are worth sacrificing a bit of variation and doing more frequently than anything else, in my opinion. Also, the movements are so technical and practiced that if I don't dedicate time to it every week, I simply won't improve.
I set aside at least one day per week to work on Olympic lifting (Typically Monday, now). Having this one element of certainty gives me the base around which I can build the rest of my week. This is so important. If you don't have a cornerstone, the concept of variation can easily spin into randomness or you can find yourself paralyzed in deciding what to do from day to day. I've found that major movements, (the olympic lifts, deadlift, squat, flat/overhead press) as opposed to favorite WODs, tend to work best in this capacity because they represent specific movement patterns.
Now that I have a place to start, I can apply the ideas of varied movement pattern and varied energy pathway to create a complete and kick ass weeklong training program. To avoid replicating weeks, I stay a student of history and of creativity, always writing down what I've already done and always looking for new twists on old movements (Stone deadlift) or new combinations of old exercises (2k row w/ bodyweight bench). Apart from a few benchmark WODs and the major strength lifts, I try to avoid doing any workout twice. Some people think this makes it harder to measure progress, and they're not wrong. But I find that the general benchmarks are enough, for the most part. Besides, as I've said before, improvement is tied to inspiration. And it's far more motivating to tackle the new challenge than to get better at the old one.