Link building metrics and reporting

By Bill Hartzer · Hartzer Consulting · 25+ years in search, domains and expert witness work

The reporting problem

Link building reports fail in a specific and predictable way: they report the numbers that export easily rather than the numbers that describe what happened. Total backlinks exports easily. Whether a link still exists in six months does not.

The result is a genre of report that is technically accurate and substantively misleading. Totals go up. Charts trend right. Nobody notices that a single site added three thousand template pages, or that half of last year's placements have quietly been edited out.

A report is a document that someone may later read adversarially — a client deciding whether to renew, a finance director asking what the money bought, an acquirer's diligence team, occasionally a court. Build it so it survives that reading. In practice that means three commitments: state your definitions, lead with the stable numbers, and report loss as prominently as gain.

There is a second failure worth naming, because it is the one that damages relationships rather than just accuracy. Reports drift toward activity. Emails sent, prospects contacted, articles published, pitches made — all easy to count, all evidence of effort, none evidence of result. A report built mostly from activity numbers is describing what the agency did rather than what the client got, and every client eventually notices. Activity belongs in a report, but at the bottom of it.

The rest of this page is the reference version of what to include, what to lead with, and what to keep out. Module 8 has the full worked structure with a sample layout.

Referring domains over links

The most important reporting decision is which unit you count. Total backlinks is the least useful number in link building, and it is the one almost every report leads with.

The reason is concentration. On a real twenty-year profile of 1,301,839 links from 22,260 referring domains, the top 100 referring domains accounted for 84.3% of all recorded links. One single domain sent 219,159 links — 16.8% of the entire profile — from one sitewide link in a template.

Think about what that means for a report. If that one relationship ends, the site loses a sixth of its total backlink count overnight, with essentially no change in how many distinct sites vouch for it. If the same site redesigns and adds pages, the count rises for the same non-reason. The total is a measurement of someone else's page count.

Referring domains does not have this problem. It counts distinct sites that decided to link, which is the thing a link is supposed to represent. It is also far more stable across different backlink tools, because vendors disagree wildly about link totals and much less about domain counts.

The practical rule for any report: lead with referring domains, show total links only as a secondary figure, and annotate any large movement in the total with its cause. If your total link count doubled and your referring domain count did not move, nothing happened.

Distinct over duplicate

Related but separate: within a link count, decide explicitly whether repeated links from the same source count once or many times, and state the decision.

On a real Fresh index, a headline figure of 11,815 backlinks resolved to 9,772 classified links, of which 7,395 were duplicates — leaving roughly 2,400 genuinely distinct links. That is a fivefold difference in the headline number produced entirely by a counting convention. No editorial decision sits behind it. Nobody built or removed anything.

Two agencies can audit the same site, both report accurately, and produce numbers five times apart. If your report does not say which convention it used, the number in it cannot be checked, compared, or defended.

The definitions worth writing into every report:

• Referring domain: a distinct registered domain with at least one live link to the site.
• Distinct linking page: a unique URL containing at least one link.
• Total links: every recorded link instance, including duplicates and sitewide repetitions.
• Index and view: which vendor, which index, and the date pulled.

This looks like pedantry until the first time a number is challenged, at which point it is the difference between a defensible report and an argument.

Reporting loss as honestly as gain

Links are not permanent, and a report that counts acquisitions without accounting for attrition describes a snapshot as though it were a balance.

The real numbers from that twenty-year profile:

• The median referring domain stops linking after 1,080 days — under three years.
• 49.6% of all links ever recorded were eventually lost. Roughly half.
• Survival scales with quality: domains at Trust Flow 0 lasted a median of 859 days; those at Trust Flow 61 or above lasted 3,353 days — about four times as long.
• 97.4% of lost links disappeared while the source page was still perfectly reachable. Of 645,202 lost links, only 4 were lost to a 404.

That last figure is the one that changes behavior. Links do not mostly die because pages die. They die because pages get edited, redesigned, re-templated, migrated or pruned — while remaining live. Which means the source is still there, still publishing, and still contactable.

So a competent report includes a net position: domains gained, domains lost, and net change. It also includes a recovery list. On that profile, 717 referring domains at Trust Flow 21 or higher had stopped linking entirely — a warm prospect list of sites that already decided once that the site was worth citing.

A program that reports only gains will eventually be running flat while claiming growth.

What to report

A link report that survives scrutiny contains a short list of things, each of which someone could act on.

• Referring domains, current and net change for the period. The headline.
• New referring domains, listed individually, with the linking page and the reason it exists. If you cannot name why a link happened, you did not earn it.
• Lost referring domains, listed individually, with whether the source page is still live.
• Quality distribution of new domains, banded by your chosen authority metric, with an explicit note on what that metric measures.
• Relevance judgment. Manual, subjective, and worth more than the metric band. A relevant link from a modest site beats an irrelevant link from a strong one.
• Survival of prior periods' links. What percentage of last quarter's placements are still live.
• Referring pages to the pages that matter commercially, not just to the home page.
• Activity, in a subordinate position. Emails sent, pitches placed, assets published. This is input, not outcome, and should never lead.

Module 8 has a full worked report structure. The reference version is the list above, in roughly that order.

What to leave out

Some numbers actively make a report worse, either because they are noise or because they imply a precision nobody has.

• Total backlinks as the headline. For every reason in the section above.
• Average domain authority of new links. The distribution is heavily skewed; an average across it is arithmetic without meaning. Show the bands.
• Toxicity scores. Vendor model outputs presented as findings. They over-flag, they have no ground truth, and putting one in a report invites a client to demand action on it.
• Projected ranking impact from links. Nobody can compute this. Presenting it as a figure is the fastest way to lose credibility when the projection misses.
• Estimated traffic value of coverage. An advertising-equivalence figure dressed as an outcome. It measures nothing that happened.
• Anchor text charts as a percentage of links. One sitewide link with a single anchor can dominate the entire chart. Report anchor distribution over referring domains.
• Screenshots without dates. A link report that cannot be reproduced is not evidence of anything.

The instinct behind including these is understandable: reports feel thin without them, and clients ask for a number that translates links into money. The professional answer is that the translation does not exist, and a report that pretends otherwise is a liability for whoever signs it.