How I write a case study that someone actually finishes
Most portfolio case studies are written to prove effort. The ones that get read are written to answer a question. Here is the structure I use and why each part earns its place.
Screens, personas, journey maps, a wall of sticky notes photographed from an angle. It is exhausting, and hiring managers skim it, because effort was never the question they arrived with.
Their question is simpler: can this person find the real problem and make a defensible decision about it?
Everything below is arranged to answer that as fast as possible.
Open with the problem, not the brief
A brief tells me what somebody asked for. A problem tells me what was actually wrong. Those are rarely the same thing, and the gap between them is the most interesting part of any project.
Start where the trouble was:
Inspectors found defects reliably. A defect recorded on Monday could still be built over by Thursday, because nobody downstream knew it was open.
Two sentences and a reader already knows what is at stake.
Show one piece of research that changed your mind
Not the whole study. One moment.
The credible detail is the one where you were wrong. If the research only confirmed what you already believed, either you got lucky or you were not really asking.
Name the decision, and what it cost
This is the part almost everybody skips, and it is the part worth reading.
A decision with no trade-off is not a decision, it is a preference. Say what you gave up. Say who disagreed. Say what you argued with, whether that was session recordings, a metric, or a constraint nobody wanted to hear about.
Design the states before the happy path
Empty, loading, error, and the long-running case where somebody has used the thing for a year. Those four change a layout more than any visual pass does, and showing them is the quickest way to signal that you have shipped something rather than mocked it.
End with what you would do differently
Not as false modesty. As evidence that you kept thinking after the work shipped.
The length nobody tells you
If it takes more than about four minutes to read, the structure is doing the work that editing should be doing. Cut the process diagram before you cut the reasoning.