That's because Colville's approach is still the scripted skill challenge which means it puts a lot on the GM to preplan a clear path through or that it feels like you're still rolling even when things should have happened already. If you have a very clear goal, like the one he presents in escaping the collapsing keep, then you can do what he did, which is script in a number of individual challenges and deal with them in a roll. This is less a skill challenge and more just a string of set challenges that you use ability checks to overcome instead of the usual freeform. This structure really starts to show it's issues with anything that doesn't lend itself to a series of scripted challenges.
This is a frustration I have with how skill challenges get presented -- it's like they ignore the easiest approach of fiction first skill challenges (I really need to get that thread up). In a nutshell, the three important things about a skill challenge are the mechanical structure, the goal the PCs are going for, and making sure that each check advances the fiction. The structure is the usual, X successes before 3 failures, but this helps frame and drive the challenge towards the goal. Each check should start with a framed scene that places a challenge to achieving the goal in front of the players. You then get the action and adjudicate it per the normal loop. If the check succeeds, you change the fiction to represent that -- the PCs either advance the current challenge to a new obstacle or they succeed in bypassing it and you move to a new scene with a new challenge. On a failure, you add a consequence that changes the current obstacle, making it more dangerous/harder/costly or you have that line of approach closed out and frame a different approach (in consultation with the PCs). The important thing in both cases is to advance the fiction! Don't leave the situation the same, make it breathe.
Now, you cannot script these things because you don't know what the PCs are going to do to overcome a challenge or how it will come out. If you do try to, you'll end up with all kinds of branching possibilities, most of which won't matter. Instead, to prep, think about the nature of the goal and the current fiction and jot down some notes about possible consequences or players in the challenge. This will aid you when you're moving scenes within the challenge to quickly add new issues. Keep the overall state of the checks in mind, though -- things should be looking dire on 2 failures and no successes! Use the current state, the goal, the PCs themselves, and your prepped ideas to drive the skill challenge. Don't be afraid to make failures hurt and always, always, always honor successes by moving things toward the goal. This way the players feel the situation in the framing, see how successes move towards their goal and how failures cause things to go pear shaped. Mix and match consequences -- make some mechanical, like loss of hitdice or hp or spell slots; make some physical, like loss or damage to gear or property; and make some fictional, like angering an NPC, losing reputation or status. But, always change the fiction and honor the success or failure!
For less complex challenges but still more complex than normal, the 3 ability check combo works. This is because it ties a single challenge to multiple rolls, and where you don't adjudicate the outcome until you see all the rolls. I use something very similar for my expanded downtime activities list for my current 5e game. As an example, Pit fighting is a downtime activity (if that gives you a sense of things) that's handled by a DEX check, a CON check, and an attack roll against a fixed DC. Successes/failures determine the outcome. It's a quick, simple approach that adds some range of outcome, but I'm not sure if it's well suited to most tasks in D&D.
