MY APPROACH
I started from the submission lifecycle itself: open a window, receive work, triage it, assign readers, record decisions, send responses. Each step had to have a single owner in the system. Blind review could not be a checkbox. It had to be enforced at the data layer so no reader could reach author identity through any screen or export. Writers needed their own portal with history, status, and withdrawal before a decision landed. Journals running more than one publication had to share one account without paying again for each title.