A redirect is a promise that the address still means something. Getting that promise wrong is worse than leaving the page missing, because a redirect to an unrelated page is a soft 404 that a search engine records as a success.
The three decisions per rule
From and to. The from side is a path on this site with no domain, and the leading slash is optional. The to side is a path or a full address, and leaving it empty makes the rule answer 410, which says the page is gone on purpose.
How it matches. This exact path matches one address. This path and everything under it matches the path and every child, which is how a whole retired section moves in one rule.
Which code. 301 for a permanent move, 302 for a temporary one, 410 for something that is not coming back. A 302 that stays for a year is a 301 nobody updated.
Priority, and why two rules can both match
Rules are checked lowest number first. Two rules can match the same address, and the earlier one wins, with the rule id breaking a tie so the answer is always the same one. That determinism matters more than the numbers themselves.
A loop is refused when you save it rather than served when somebody hits it. A chain is collapsed, so a to b to c answers with one 301 rather than two.
The 404 log is the source of the rules
The screen lists addresses that were requested and did not exist, with a hit count and the last time each was seen. Each one carries a verdict.
Break means the address used to work here and no longer does, which is the one worth redirecting. Wrong URL means it looks like a mistyped link. Probe means it was never published and the path looks like a scanner looking for a hole, and redirecting those is how a site ends up with rules for paths it never had.
Redirect this on a row starts a rule with the from side already filled.
What a redirect must not break
Redirecting a page that is still published is the mistake that costs traffic twice. The rule wins over the page, the page stops being reachable, and the sitemap still lists it. The plugin disables canonical redirection for its own sitemap and key file for the same class of reason.
Retire the page first, then redirect the address. If the old address ranked for something the new page answers, the redirect keeps that. If it did not, a 410 is the honest answer.
Check the live behavior with a request rather than trusting the table.
curl -sI https://yoursite.com/old-blog/ | head -5
The Blue Raven plugin page lists the rest of the search tools, and the sitemap, robots.txt and IndexNow covers what the crawler is told afterwards.
EnterraHost