Honestly, it depends! You need to think about the industry/business context, the company context, and the situationally/problem context. Only then can you create an appropriately tailored cost breakdown.
The 3 ways that you've seen are all "correct" and are situational.
In general, for determining cost issues, you need to break down the problem into a tree/root-cause analysis and ask the highest level (but specific) questions first! In this way, you essentially move down the tree.
How do you identify where to look? Well, you need to look into whichever of the following 5 make the most sense based on where you are:
- What's the biggest? (i.e. largest piece of the pie...most likely to change the end result)
- What's changing the most? (I.e. could be driving the most and most likely to be fixable)
- What's the easiest to answer/eliminate? (i.e. quick win. Yes/No type of question that eliminates a lot of other things)
- What's the most different? (differences between companies, business units, products, geographies etc....difference = oopportunity)
- What's the most likely? (self-explanatory)