Agree. The problem comes first. And: expect change.
If you strive for Quality Without A Name (Christopher Alexander allusion), you cannot just copy any old cookbook pattern. You have to get involved and understand what is going on; there are no shortcuts to qualified design decisions.
Once you understand what's in play, what people involved are up to, you will often find that a composition of old cookbook patterns describe the situation well. And sometimes you will have the joy of discovering a new situation, a new pattern.
Obviously, our understanding of what is going on will be limited and incomplete. As an engineer, it is my profession to make informed decisions given noisy and incomplete information, c'est la vie. It is also my profession to do myself and my colleagues the favour to name assumptions, to offer handles to the design decisions made, to present models of the design: to expect change.
Unnamed assumptions create a global state that makes change risky: there is no way to explore the impact of a change except in full scale. Also, it gets costly to identify bottlenecks and boat anchors, because you have to benchmark the complete process rather named components. You get stuck with blackboxes and implied workflow patterns that are no longer relevant. Like, "Articles are referred to by issue number and page number", distributed all over work flow, data models, and code. I'm sure a lot of you docheads recognize this situation. Years ago, digital publishing used to piggyback on paged media. Today, having paginated media as a prerequisite to digital publishing is a royal PITA. And, if identifiers in your workflow are implicitly depending on paged media, the pain is a lot worse than it need be. If you cannot abstract from previous assumptions, if you cannot talk about work, expression, manifestation, and item, you're stuck with whatever chance brought you. Schema are ways to make your assumptions visible and subject to change.
If in doubt, focus on the customer. If you don't know your customer, you are in trouble.
