After this lesson you should be able to
- Explain why 404 monitoring fails to detect lost links
- Run a snapshot comparison that finds losses reliably
- Prioritize reclamation requests by value rather than by date
- Build an ongoing calendar that survives busy quarters
The highest-yield work in the whole discipline
If I could give a link program one instruction, it would be this: before you build anything new, find out what you have lost.
Reclamation is cheaper than acquisition on every dimension. The site already linked to you once, so the relevance question is settled. There is a specific page and a specific former link to reference, so the message writes itself. The request is small — restore a link that used to be there — rather than a favor. And the person receiving it has already demonstrated, by linking once, that they are the kind of publisher who links.
Yet reclamation is the first thing dropped from every busy quarter, because it produces no story to present and no campaign to name. That is precisely why it is available: almost nobody does it consistently, so the losses accumulate quietly on profiles whose owners believe they are growing.
The scale is not marginal. On the profile behind this course, 49.6% of all links ever recorded are lost, and the median referring domain stops linking after 1,080 days. Half the work of a mature program is holding on to what it already earned.
Why 404 monitoring finds nothing
The standard advice for link maintenance is to monitor for broken pages: watch for 404s on the sites that link to you, and chase the ones that break. That advice is almost entirely wrong, and here is the evidence.
Across 645,202 lost links on a single profile, with the recorded reason at the moment each link was lost:
- 97.4% were lost while the source page was still perfectly reachable. The page returned a normal response. It simply no longer contained the link.
- Four links. Out of 645,202, exactly four were lost to a 404.
Read that ratio again. A monitoring strategy built on detecting dead pages would have caught four losses out of well over half a million. It is not a weak method; it is very nearly a null method.
The mechanism is obvious once you see the numbers. Links do not usually die because pages disappear. They die because pages get edited. A site redesigns and the sidebar goes. A resource list is refreshed and older entries are dropped. A content management system migration mangles inline links. An editor updates a two-year-old article and swaps your citation for a newer one. In every case the URL still resolves, still returns a healthy response, and no monitoring tool watching for errors will ever notice.
The biggest single risk to your existing links is other people's redesigns, which you cannot control and cannot predict — but can absolutely detect.
The snapshot comparison method
The method that catches everything is comparing referring-domain snapshots over time. It is unglamorous, it takes minutes, and it finds losses regardless of why they happened.
- Export your referring domains on a schedule. Monthly for an active program, quarterly at minimum. Save the file with the date in its name and keep every one.
- Compare the current export against the previous one. Domains present before and absent now are your loss list. A spreadsheet lookup does this; no special tooling is required.
- Prioritize by value, not by recency. Sort losses by Trust Flow and by relevance. On the profile behind this course, 717 referring domains at Trust Flow 21 and above have stopped linking entirely — that is a substantial recovery list hiding inside a profile that looks healthy in aggregate.
- Verify each one before writing. Fetch the source page and confirm the link is genuinely gone rather than moved, changed to nofollow, or missed by the crawler. Index data has lag and false negatives; a manual check takes seconds and prevents an embarrassing email.
- Write to the person. Reference the specific page and the specific former link, note politely that it appears to have been removed in a recent update, and ask whether it could be restored. Do not accuse anyone of anything; the removal was almost always accidental.
- Record the outcome in the same sheet you use for acquisition, so reclamation shows up as measurable work.
Response rates on this kind of request are far better than cold outreach, for the obvious reason that you are not a stranger and the ask is small.
The ongoing calendar
A program that depends on someone remembering will not run. Put it in a calendar with an owner and it will.
- Monthly: export referring domains, compare against last month, produce the loss list, verify the top losses by hand, send that batch of reclamation requests. Half a day for most sites.
- Monthly: check for new unlinked mentions and request links while the coverage is fresh. The request costs the writer thirty seconds when the piece is recent and is much harder a year later.
- Quarterly: full profile review. Net referring domains against the replacement rate, Trust Flow distribution of what arrived and what left, anchor drift, and a re-run of competitor link gap analysis to find newly reachable prospects.
- Quarterly: relationship maintenance. Contact people who have linked before with something genuinely useful rather than a request. This is what makes the next link cheap.
- Annually: re-audit the profile using the method in lesson 8.3, and revisit any disavow file that exists.
The monthly items are the load-bearing ones. If a quarter goes badly and something has to be cut, cut the acquisition campaign rather than the reclamation cycle, because losing links you already have is more expensive than not gaining new ones.
What the long view actually looks like
Two things become visible once you have several years of snapshots, and neither is obvious from inside a single quarter.
Quality buys durability. Survival scales steeply with Trust Flow: a Trust Flow 0 domain has a median life of 859 days, while a domain at Trust Flow 61 or above lasts a median of 3,353 days, roughly four times as long. A high-quality link is not just worth more while it exists, it exists about four times as long, so it needs replacing a quarter as often. That is a compounding advantage and it is the strongest practical argument for being selective.
Concentration is extreme and mostly outside your control. On the profile behind this course, the top 100 referring domains account for 84.3% of all links, and a single domain contributed 16.8% of the entire profile from one sitewide placement. You cannot plan for that. What you can do is keep acquiring genuine relationships, because the outlier arrives out of a large enough number of ordinary ones.
Which brings the module, and most of this course, to the same place. The work is not a campaign, it is a rate: acquire steadily, lose steadily, reclaim what you can, and prefer links that last. Do that for several years and the profile takes care of itself.
Questions
Why does monitoring for 404s not find my lost links?
Because losses do not come from dead pages. Across 645,202 lost links on one profile, 97.4% were lost while the source page was still perfectly reachable, and exactly four were lost to a 404. Links disappear when pages are edited, redesigned or migrated, which leaves the URL responding normally. Error monitoring cannot see that; snapshot comparison can.
How often should I check for lost links?
Monthly for an active program, quarterly at the absolute minimum. Export referring domains, compare against the previous export, and verify the highest-value losses by hand before writing. Frequency matters because a reclamation request lands far better while the editor still remembers making the change than it does two years afterward.
Do reclamation requests actually work?
Better than cold outreach, and by a wide margin. The site already chose to link to you once, the request references a specific page and a specific former link, and the ask is a small correction rather than a favor. Most removals were incidental to a redesign or an edit, so the tone should assume accident rather than intent.
Which lost links should I chase first?
Sort by value, not by date. Trust Flow and topical relevance first, then whether the page still gets traffic. On one profile, 717 referring domains at Trust Flow 21 and above had stopped linking entirely, which is a large recovery list sitting inside a profile that looked healthy in aggregate. Low-quality losses are rarely worth an email.