The work that happens before any screen gets drawn: understanding the problem well enough that the right design becomes obvious.
I start most projects with five to eight short user interviews — enough to find the patterns without drowning in noise — and pair that with a walk through the existing product to find where people actually get stuck.
From there it's flow diagrams and low-fidelity wireframes, tested informally before anything reaches high fidelity. Cheap mistakes here save expensive ones later.
The output is usually a clear before/after: the flow as it existed, the flow as redesigned, and the specific friction points each change addresses.