Large content pruning fails when redirect mapping is treated as a URL list instead of a decision system

Large content pruning fails when redirect mapping is treated as a URL list instead of a decision system

Start with a complete export of URLs, then assign each page one action: keep, improve, merge, redirect, or return a not-found response. For every URL marked for removal, record the destination, reason, source page, and owner. A blank destination is a review task, not a completed mapping.

Choose redirects by search intent, not by URL similarity. Map a retired page to the strongest surviving page that satisfies the same need. If no close match exists, do not force the redirect to a broad category or the homepage; that creates a poor user experience and weak signals.

Before deployment, check four failure points:
— chains: old URL → intermediate URL → final URL
— loops: two or more URLs redirecting back to each other
— conflicts: one source assigned to multiple destinations
— targets: deleted, blocked, canonicalized, or irrelevant pages

Keep the mapping in a versioned file and preserve the old-to-new relationship after launch. Crawl the retired URLs, inspect response codes, follow every destination, and review internal links that still point to removed pages. Watch for traffic and rankings dropping on important destination pages; that can reveal a bad intent match even when the redirect technically works.

A good redirect map is conservative, explicit, and easy to audit. Redirect only when the replacement genuinely serves the original intent, and leave everything else out of the map until a human can make that decision.