No we don't do all that.
First we absorb all the requirements stated, and try to ensure we understand the intent behind those.
THEN, based on our own personal current experience and knowledge of the state-of-the-art and our own experience/ expertise level we choose a variety of approaches (all mentally) which might do the job. There are many parameters against which this is tested (again.. mentally). Like ease of development/ confidence in outcome, cost/ viability, development time, peripheral equipment/resources required for development. Maybe a quick research on some tricky points we are unsure about.
Next we sketch out .. (on paper
Gasp ) a few ideas of the critical elements, and just think about whether they will work, or something else is required. MAybe look online for the latest black-boxes which could do the trick better than we could. Maybe have a few alternate methods outlined, along with the pros and cons of each.
Then we sketch out a fuller method based on our most best feel design selection. Flesh it out with real world components, see if those exist, and if they do then at what price/ spec.
Only as a last stage do we go MatLab, or Spice or whatever, to try understand where the screw-ups could be. And maybe tweak some values.
Choosing a topology and jumping direct into simulations is usually a waste of time without understanding what the limits of a particular design are. It's equivalent to the
million monkeys at typewriters situation.