Every case study says the same thing: there was a problem, we did research, here is a beautiful result, metrics improved. It is a genre, and like all genres it hides more than it reveals.
Read for these instead.
Constraint, not outcome
Look for evidence that something was hard. A budget that was too small, a legacy system that could not be replaced, a brand guideline that could not be broken, a stakeholder who hated the first direction.
Work made without constraint tells you nothing, because your project will have constraints. A portfolio full of frictionless triumphs is either a portfolio of speculative work or a portfolio of unusually pliant clients, and neither predicts how someone behaves in week six of yours.
Ask directly: what did you have to give up on this one? The quality of that answer is the most reliable signal in the whole conversation.
Whether it shipped, and whether it survived
A remarkable amount of portfolio work was never built. Some of it was built and replaced within a year. Neither is a scandal, but both change what the work proves.
Look the sites up. If they still exist and still look like the case study, somebody handed over something maintainable, which is a rarer skill than the design itself.
What they did, specifically
“Led the redesign” covers everything from setting the typographic system to attending a workshop. On agency projects this is genuinely murky and not usually dishonest — it is just how credit works on teams.
Ask which screens they personally drew, and who they handed to. People who did the work answer instantly and in detail. People who did not, generalise.
The second-best project
Everyone’s top project is their best case, often with atypical budget and an unusually good client. It tells you the ceiling.
The third or fourth item down tells you the floor, and your project will almost certainly live nearer the floor. Ask about that one.
Range versus repetition
Ten projects in one sector can mean deep expertise or one solution reapplied. Look at the underlying structure rather than the surface: if the layouts, the interaction patterns and the type scale are identical under different colours, you are looking at a template.
That may be exactly what you want. It is worth knowing which you are buying.