The brief was simple: rebuild the marketing site. The risk was less simple: undo six months of search momentum in a single deploy. The team had been told the migration was a design and engineering project. I read the same brief and saw a content audit hiding in plain sight.
What 38 pages teaches you
Most sites have a long tail of pages nobody opens but search has indexed. Removing them feels like cleanup. It is also the fastest way to lose rankings on the queries those pages happened to satisfy.
I built a spreadsheet. One row per legacy URL. Columns for current rank, monthly impressions, target destination, redirect type, and what would change about the content on the way there. Boring work, but the kind that decides whether the migration is silent or expensive.
Redirect rules I'd reuse
Map *intent*, not URL structure. If a legacy /features/x page is becoming part of /platform/y, the redirect should land on the section that answers the same question — not the homepage of the new site.
Hold redirects for at least 12 months. Search engines need that long to fully transfer authority, and a stray internal link will outlive any deadline you set yourself.
The audit nobody asked for
Halfway through, I started flagging pages with copy that was no longer accurate. Nobody had asked for a content rewrite. They got one anyway. The migration was the right moment, and the redirect map was the only complete inventory that existed.