Dead Link Recovery: Step-by-Step

Capture the original row

A dead link is not a lost item, it is a lost address. The row still holds a title, a price band and often a stale thumbnail; the page it pointed at has been pulled, relisted or moved to a new identifier. Everything needed to find the item again is usually still in the row, provided the row is captured before the list is refreshed and the entry disappears from view. A row captured after the list has been refreshed is often a row that cannot be recovered at all, because there is nothing left to search on.

Capture immediately, in text rather than in a screenshot. Copy the row description exactly, including spelling errors, because sellers repeat their own typos across relists and a typo is a surprisingly strong search key. Copy the price band, the storefront or seller name, the category label and the date you saw the row. Screenshots are useful as attachments but useless as search input.

Where the row links through a short code, keep the code as well. Short codes sometimes resolve after the destination changes, and a code that has already been reused will point at whatever occupies it now, which is itself a finding. Record what the code resolves to today alongside what the row claimed it resolved to, and date both observations.

Time matters more than care here. Dated lists are refreshed, forum posts are edited and entries are removed once a row stops resolving, so the description that was available last week may be gone today. Capture on the day the link fails, even where the capture is rough, because a rough record taken early beats an accurate record taken after the list has moved on.

Probe the seller storefront

The storefront is the first place to look. Most dead rows belong to a seller who is still trading and who has relisted the same item under a new identifier. Open the storefront rather than the dead link, then search inside it for the distinctive words from the original row: the model name, the colourway, the odd spelling, the material claim. Two or three of those words are usually enough to find the item again inside a single storefront.

Read the storefront as a structure rather than a list. Note how many pages repeat the same title pattern, whether the newest entries cluster in one category, and whether the item you want has siblings under a renamed title. A storefront that has been reorganised will usually contain the item; a storefront that has been emptied will not, and an emptied storefront is a different problem with a different answer.

Where the storefront is gone, look for the seller's other identifiers. The same photographs often reappear under a different storefront name, sometimes with the same option list order and the same price ladder. Reused photographs are not proof of the same seller, but they are a lead strong enough to justify the next step, and they are cheap to check.

Work the storefront before spending time on search engines. Inside a storefront the item is one or two filters away, while a general search returns the whole category and buries the row that was lost. Where the storefront holds hundreds of pages, sort by newest first and walk back through the period since the row was last known to resolve, which is usually a short window.

Rebuild search keywords

Rebuild the search from the row rather than from memory. Combine the most distinctive noun in the row description with the most distinctive modifier, then add the category label. Strip words that appear on every page in the category, because a query built from common words returns the whole catalogue and tells you nothing. Two or three terms is usually the right length: a longer query fails because relists reorder the wording, and a shorter one fails because it matches everything in sight.

Run the rebuilt query in more than one form. Use the exact row wording first, then a version with the obvious synonyms substituted, then a version with the brand or model name removed, because relists frequently drop a name to avoid attention. Compare the three result sets; the item you want usually appears in at least two of them, and a single appearance in one set is worth less.

Sort the results by the things a relist changes. A relist usually keeps the photographs and the price band while changing the identifier and sometimes the title. Search on the stable parts and read the changing parts. Where a candidate matches the photographs and the price band but not the wording, treat it as a probable match and move to verification rather than accepting it.

Keep a copy of every query that worked, not only the result it produced. A search string that surfaced the replacement will surface it again when the same row dies in three months, and the wording of that query is easier to reuse than to reinvent. Paste the query into the same note as the row, so that the recovery becomes repeatable rather than remembered.

See current listings

Verify replacement identity

Verification is the step that stops a recovery becoming a second mistake. A candidate page must match the original row on at least two independent identifiers: the photographs, the price band, the option structure, the measurement table or the packaging description. One match is a coincidence; two matches in different fields is a recovery. Choose the identifiers from different parts of the page where possible, since two facts drawn from the same gallery are one fact photographed twice.

Compare the option lists side by side where both are available. A relist that kept the same colourways in the same order with the same relative prices is very likely the same supply. A relist that inverted the ladder, dropped the cheapest option or renamed every variant has changed something, and the changed part is what you need to ask about explicitly.

Where a candidate fails verification, do not force it. A dead link is an absence of evidence, and the honest outcome is a note saying the item could not be re-found at that price. Two candidates that each match on one field are not a match between them; they are two unrelated pages, and treating either as the original is how buyers end up disputing an item they never actually ordered.

Price the replacement before accepting it, and compare the total rather than the headline figure. A relist that matches the photographs but sits in a higher freight band is not the same purchase, and a relist that dropped its cheapest option has removed the configuration that made the original attractive. Verification covers identity first and cost second, in that order.

Log the substitution

Log the substitution in the same place as the original row. Record the dead identifier, the replacement identifier, the fields that matched, the fields that changed, and the date of the swap. Six entries take a minute and turn a private guess into a checkable record that survives the next relist, which is likely to arrive within a season. Write the entries as plain searchable text, because a log that cannot be searched is a log that will not be read.

A logged substitution also changes how a later problem is argued. If the replacement ships something different, the log shows which attributes carried over from the original and which were introduced by the relist. That distinction is the difference between a claim about the item you chose and a claim about an item you were quietly moved onto. A claim naming two fields that matched and one that broke is easier to act on than a claim that something feels wrong.

Keep the log short and keep it in one file. Recovery attempts spread across three apps and two chat threads cannot be compared later, and comparison is the only thing the log is for. The desk updates its own substitutions on the same schedule it checks row survival, so that a replacement is re-checked at least as often as the original was.