During the competency model development process, you start by identifying the categories or big buckets of things that someone in a job role does, and then you unpack each category to identify the tasks within it. This is “WHAT” someone in that role needs to do. Once you have the tasks, you identify what it looks like to perform that task at various levels of proficiency. This is “HOW” someone would do it. Writing the behavioral examples is the longest part of the process.
Lastly, you identify the target level of proficiency, which is how you know whether a skill gap exists.
If you are performing the process for a job family, you may have some jobs that don't perform a particular task at all, and others that overlap with varying levels of proficiency.
For example, a senior software developer and an entry level developer may share the need to “program in some language”. However, the entry level person needs to be a Level 3 and the senior person needs to be a Level 4. Additionally, the senior software developer may have some team management responsibilities that an entry level person wouldn't have.
Here's a link to a free ATD webinar next month (October 18) with all the detail on that process.