What a redirect chain does to a crawler
A browser hides redirects. A crawler experiences every hop as a separate request against your crawl allowance, and each one is a chance for something to answer wrong. The chain, not the destination, is where the problems are.
Why the chain matters and not just the destination
You click a link, you land on a page, and the two or three round trips in between never reach your attention. A link audit that only records the final URL misses the interesting part for the same reason.
The status code on every hop is worth recording because the codes mean different things. A 301 says the move is permanent and the new URL should replace the old one everywhere. A 302 says the original is still the right address and this detour is temporary. A 307 says the same as a 302 while promising the request method survives, which matters for anything that is not a plain GET. Mixing them up does not break the page for a human and does change which URL ends up in the index.
The chains that cost you something
The usual accident is stacking. Someone adds a rule to move an old category to a new one. A year later the new one moves again and a second rule goes on top. Then the site gains an apex to www rule and a trailing slash rule, and a single old link now takes four hops before anything renders. Nothing is broken, the page loads, and every visit pays for it. Rewrite the old rules to point straight at the current destination rather than letting them chain.
Worse is the circle. Two rules disagree, or an application redirect fights a server one, and the URL never arrives anywhere. A browser gives up with an error and a crawler drops the URL from its index, which is how a page disappears with nobody having deleted it. Any URL that appears twice in the walk is the tell, and what a loop does to a crawler covers how to trace the two rules that are arguing.
Then there is the redirect that should have been a status code. Retiring a product and sending every dead URL to the home page looks tidier than a wall of errors, and search engines treat it as a kind of lie, because the destination has nothing to do with what was asked for. Return a real 404 or a 410 instead. Why 404s are healthy makes the case better than a redirect rule can.
After the chain is clean
Once a URL resolves in one hop or none, check that the destination agrees with itself. A page that redirects correctly and then names a different address in its canonical tag sends two conflicting signals, and the canonical URL entry explains which one wins and when.
Walking a chain yourself
Start from the URL people actually have, the one in the old link or the printed advert. Starting from the destination hides the hops you are looking for. Send the user agent of the crawler you care about while you do it, because a CDN rule can redirect one agent and not another, and each crawler page carries the string.
Where redirects sit in the order a crawler works through is on how a crawler resolves a page. A chain that ends on a 429 is a rate limit rather than a routing problem, and it is read differently again.
Common questions
How many redirects are too many?
One is fine, two is tolerable, three starts costing you. Google follows up to about five hops in a single pass and picks the chain up later if it runs out, which delays the move rather than stopping it. Browsers usually allow around twenty.
Does a 302 pass ranking signals?
Treat it as temporary and expect the original URL to stay indexed. Search engines can decide a long-lived 302 is really permanent, but that is their inference on their schedule. For a move that is not coming back, return a 301.
Why does a chain show an extra hop nobody created?
Usually the host or the trailing slash. A request to the apex is redirected to www, or a path without a slash is redirected to one with it, and each of those is a separate hop that stacks on top of whatever rule you wrote.
What counts as a redirect loop?
Any URL that appears twice in the chain. Once a URL repeats, the rest of the walk tells you nothing new, and the repeat itself is the evidence you need to find the two rules that disagree.