After this lesson you should be able to
- Build a tracking system simple enough that it survives contact with real work
- Record relationships and permissions, not just acquired links
- Run a re-contact cycle on people who said not now
- Hand a campaign over without losing its accumulated context
The CRM you will not maintain
Every link building team I have observed goes through the same cycle. Outreach is messy, so somebody proposes a proper system. A CRM is configured with custom fields, pipeline stages and automation. It is used properly for three weeks. Then a campaign gets busy, the updates lag, the data goes stale, and within two months everyone is working from a spreadsheet again while the CRM sits there being technically the source of truth.
The failure is not laziness. It is that the maintenance burden was set higher than the value returned to the person doing the maintaining. An outreach specialist updating twelve fields per prospect gets nothing back from field nine. So field nine goes empty, then wrong, and once any field is wrong the whole record stops being trusted.
The design principle that actually works: record the minimum that a future human needs, and make every field one that somebody uses. If you cannot name who reads a field and when, delete it. A five-column spreadsheet that is accurate beats a thirty-field CRM that is 60% populated, because the spreadsheet can be trusted and the CRM cannot.
Choose the tool last, and choose it based on whether replies land somewhere the record can see them. A shared inbox plus a shared sheet is a perfectly respectable system and it is what many good teams actually run.
The minimum viable record
These are the fields I would defend. Each one has a named consumer and a moment it gets read.
| Field | Why it exists |
|---|---|
| Domain | The key everything joins on; one row per domain, always |
| Contact name and address | You are corresponding with a person, not a site |
| Where the address came from | Tells a future sender whether it was published or inferred |
| Why this prospect is here | One sentence. This is what the person writing the email needs |
| Target URL | What you were asking them to link to |
| Date first contacted | Anchors the follow-up policy and prevents double-pitching |
| Outcome | A fixed vocabulary, not free text |
| Their words | The reason they gave, quoted, when they gave one |
| Live URL of the link | Only populated when a link actually exists |
| Next action date | The single most valuable field in the sheet |
Two rules make this work. Outcome uses a fixed vocabulary — linked, agreed not live, refused with reason, refused, wrong contact, contact later, do not contact, no response, bounced — because free text cannot be counted and will not be. And one row per domain, ever. The moment a domain appears twice, two people pitch it, and one of them is following up on a conversation they did not have.
Record relationships, not just links
Most tracking systems record the artifact and discard the relationship. That is backwards, and I have data explaining why.
In an analysis of my own profile — 1,301,839 links from 22,260 referring domains, none bought — the median referring domain stops linking after 1,080 days, roughly three years. Nearly half of all links ever recorded, 49.6%, are already gone. And 717 referring domains at Trust Flow 21 or above have stopped linking to me entirely. Those are not junk sites. They are established domains that once decided I was worth citing.
The mechanism matters more than the totals. Of the lost links, 97.4% disappeared while the source page was still perfectly reachable. Only four out of 645,202 lost links were lost to a 404. The pages did not die. They were edited, redesigned, migrated to a new CMS, consolidated, or rewritten by someone who did not know what the link was for — and the link went with the edit.
Read that as a management problem rather than a technical one. Links disappear because of human decisions inside organizations, and the counter to a human decision is a human relationship. If somebody at that publication knows you, the migration that drops your link is the migration where someone thinks to check. If nobody knows you, you are a URL in a database.
So record the relationship. Who they are, what they cover, what they said, what they asked for, what they did not want. A person who refused politely and explained why is a contact. A person who linked once and vanished into the sheet as a green cell is not.
The re-contact cycle
The highest-return activity in mature link building is contacting people you have already spoken to. It is also the one almost nobody does, because campaigns end and their spreadsheets get archived.
Build a cycle around three queues, each driven by the next action date field.
- Contact later. Everyone who said not now, ask again after the study is out, or try me in the new financial year. These convert at a far higher rate than cold prospects because the relationship exists and the objection is known. Diary them at the date they named, and honor it.
- Agreed but not live. Links that were promised and never appeared. Check at two weeks and again at six. A meaningful share of agreed links simply never get published — the editor got busy, the CMS update was deferred, the page was in a redesign freeze — and a single polite check recovers many of them.
- Was live, now gone. Run a periodic check against your known link list. Given the decay figures above, this queue fills steadily whether you attend to it or not. Recovering a lost link from someone who once agreed to link is far cheaper than earning a new one, and the conversation is easy: your page was updated and this reference came out, was that deliberate?
Quarterly is a sensible cadence for all three. The check itself is largely automatable — a scheduled crawl of known linking URLs — but the email is not, and should not be. A message that says our monitoring detected a change is a message from a system. A message from the person they dealt with is a message from a person.
Scaling without losing the thread
The system has to survive more than one person using it, and it has to survive that person leaving.
Ownership per domain, not per campaign. One named person owns the relationship with a domain across all campaigns. Without this, the same publication receives pitches from three colleagues in a quarter, which is exactly how a warm contact turns cold.
Replies must be visible. If outreach goes out from individual mailboxes and replies stay there, the record is fiction the moment somebody takes a holiday. A shared mailbox, or an outreach tool that syncs threads to a shared view, is the minimum.
Suppression lists are global and permanent. Anyone who asked not to be contacted goes on a list checked before every send, across every domain and every campaign, forever. This is a legal requirement in most jurisdictions and it is also the thing that most reliably goes wrong when a new tool is introduced and the old list is not imported.
Write down the context that is not a field. One free-text note per domain: they only cover the enterprise market, their editor hates statistics without methodology, they will not link to anything commercial but will quote an expert. That paragraph is the actual institutional knowledge, and it is what walks out of the door when somebody resigns.
Do the handover before you need it. A campaign handed over with a spreadsheet and no context is a campaign restarted from cold, and the prospects can tell.
Common mistakes
Tracking only successes. A sheet containing only the domains that linked discards the refusals with reasons, which is where the intelligence is. Failed outreach that explains itself is more useful than a link that arrived unexplained.
Letting the tool dictate the fields. Configuring every field a CRM offers guarantees most will be empty. Start from what a human needs to read and add nothing else.
Archiving at the end of a campaign. The contact-later queue is worth more than the links, and it dies in an archived folder. Relationship records should outlive campaigns and clients.
Never re-checking live links. Half of all links eventually vanish, almost always through page edits rather than page deaths. If you do not check, you will not know, and you will report a number that has not been true for a year.
Automating the human part. Automate the crawl, the reminder and the report. Do not automate the message. The whole value of a relationship record is that it lets a person write something a system could not.
Questions
Do I need a CRM for link building outreach?
No, and most teams that buy one abandon it within two months because the maintenance burden exceeds the value returned. What you need is one row per domain, a fixed outcome vocabulary, a next-action date, and replies visible to everyone. An accurate shared spreadsheet plus a shared mailbox beats a half-populated CRM that nobody trusts.
How often should I check whether my links are still live?
Quarterly is sensible for most programs. In my own profile the median referring domain stopped linking after 1,080 days and 49.6% of all links are already gone, so the decay is continuous rather than occasional. Automate the crawl, but write the recovery emails by hand — a message from a person outperforms one from a monitoring system.
Why do links disappear if the pages are still there?
Because pages get edited. In my data 97.4% of lost links vanished while the source page was still perfectly reachable, and only four of 645,202 losses were caused by a 404. Redesigns, CMS migrations, content consolidation and rewrites remove links routinely, and usually nobody involved knows what the link was for.
What is worth recording besides whether they linked?
The reason they gave in their own words, whether they asked to be contacted later and when, who at your end owns the relationship, and any free-text context about how they work. Those notes are the institutional knowledge, and they are what walks out the door when the person who did the outreach leaves.