Indefinitely for any URL that carried search visibility or external inbound links, which on an industrial site means most of them. Search engines consolidate the signal within months, but the external references do not expire, so removing a working rule saves very little and risks breaking a link someone still uses.
The industrial case is stronger than the general one. Datasheet and manual URLs sit in distributor catalogues, association directories, tender packs and printed specification documents, and part-number URLs are quoted in procurement systems. None of those can be revised because a redirect rule was tidied up.
Search engines usually finish consolidating an old URL into its destination within a few months, so the ranking argument for keeping a rule weakens over the first year.
External references run on no clock at all. A datasheet link in a distributor catalogue, a URL printed in an installation manual or a path quoted in a procurement record stays exactly as it was written. The second clock is the one that determines how long the rule is needed.
An implemented redirect has a small, fixed maintenance cost and a negligible performance cost when it resolves in a single hop at the server or CDN. Rule count is rarely the constraint at the scale most industrial sites operate at.
Where redirects do become expensive is in how they are organised: undocumented rules spread across a configuration file, a plugin table and a CDN layer, with no record of why any of them exists.
A URL redirecting through two or three intermediate rules before reaching its destination is worth correcting, and that is a different task from removing the redirect. Flatten each chain so every legacy URL points directly at its current destination.
Chains accumulate naturally across successive migrations, which is why they are worth re-flattening after each one rather than allowing another layer to be added.
Anything with external inbound links, anything still receiving requests in the server logs, every document URL, every part-number URL, and every page that ranked. On an industrial site that describes most of the pre-migration inventory.
The categories where a case for removal exists are narrower: expired campaign URLs, faceted filter combinations, session parameters and staging paths that never held external references.
Confirm from server logs and backlink data that the URL has had no requests and no external links over a sustained period, ideally more than a year, and that no internal link or document still points at it. Record the decision and the date.
Removal at scale during a platform change carries more risk than individual removal, because the evidence is rarely gathered per URL when a whole rule set is dropped.
Hold the redirect map as a maintained record with the source URL, destination, reason and date, exported from the live configuration rather than reconstructed later. That record is what makes the next migration a mapping exercise rather than an archaeology exercise.
See website governance for where this sits alongside domain and ownership records.
Indefinitely for any URL that held search visibility or external inbound links. Search engines consolidate the signal within months, but distributor catalogues, association directories, tender packs and printed manuals keep pointing at the original URL for years. Since a working rule costs very little to retain, the balance favours keeping it in place.
A single-hop redirect resolved at the server or CDN adds a negligible amount, and rule count is rarely the limiting factor at normal site scale. Where redirects do cost performance is in chains, where a request passes through several rules before reaching the destination. Flattening chains addresses that without removing any rule.
Worth cleaning up, and a separate task from removal. Repoint each legacy URL directly at its current destination so no request passes through an intermediate rule. Chains build up across successive migrations, so re-flattening after each one keeps the set from deepening. Loops need correcting immediately, since they leave the URL unreachable.
Where evidence supports it. Confirm from server logs and backlink data that the URL has had no requests and no external links for a sustained period, that no internal link or published document still references it, then record the decision and the date. Removing at scale during a cleanup rarely meets that standard per URL.
Document redirects merit the longest retention of anything on an industrial site. Datasheets and manuals are deep-linked from distributor catalogues and quoted in specification packs and printed literature, and those references are never revised. See how to redirect legacy PDF URLs for how they are mapped in the first place.
Every redirect from it stops resolving at once, so a domain acquired in an acquisition or retired after a rebrand needs to stay registered and pointed for as long as the references matter. Registration is a small annual cost against the loss of every inbound link the domain accumulated. Keep it in the domain inventory with a named owner.
Tell us what you are considering removing and we will set out what the risk looks like.