Define the question and the denominator
An anchor distribution describes the supplied observations under a stated grouping rule. It is not a census of the web or a score for a healthy profile. Start with a question such as whether repeated wording comes from many domains or many pages on one domain. Specify the date, provider, filters and target scope.
This worksheet counts distinct recorded observations after exact duplicate removal. It does not weight them by traffic, authority or a provider’s aggregated link count. One observed row is not proof of one separate live HTML link.
Use a link-level export with a date
Required columns are Source URL, Target URL and Anchor; Rel and Nofollow are optional preserved evidence. One file must resolve to one target domain, using registrable domains with private suffixes. Separate hosted tenants remain separate. Invalid URLs, ambiguous columns, mixed targets and unreadable files are rejected.
A summarized list of anchor strings and totals cannot reconstruct source pages or domains. Do not invent URLs to make it importable. Search Console’s Links report is sampled and groups its data differently; its anchor summary is not a supported link-level input for this worksheet.
Normalize a view while preserving original evidence
The original anchor field remains unchanged in retained evidence. The grouping view applies Unicode NFC, collapses whitespace and trims outer whitespace. Choose case preservation or Unicode lowercase conversion; punctuation and accents remain distinct. Lowercase conversion is not full linguistic case folding or transliteration.
Exact duplicate identity uses parsed full source and target URLs, original anchor, Rel and Nofollow. Fragments and attribute differences remain distinct. Normalization may merge several original spellings into one group without merging their distinct observations. Blank text remains an unknown label, not an inferred image link.
Reconcile the twelve-record exercise
The fictional CSV has twelve input records: one internal record, one exact repeated external record and ten retained external observations. Lowercase grouping yields five anchor groups, six source domains and two target URLs. The Northstar group has four observations from alpha.example; its whitespace and capitalization variants remain inspectable.
Two descriptive observations from beta.example group together. The two Café spellings use composed and decomposed Unicode and join under NFC. A visible URL and a blank label each form another group. No example host was fetched, and the exercise does not describe a real publisher network.
Read the distribution from two perspectives
Northstar occupies 4/10, or 40%, of retained observations, yet appears on only 1/6 source domains, about 16.7%. These are different denominators answering different questions. A high page count concentrated on one source should prompt a context check, not an automatic over-optimization diagnosis.
A domain may use several anchor groups, so domain-coverage percentages can add to more than 100%. Observation shares use the selected target scope and total 100% apart from rounding. Search and pagination hide rows without recalculating those denominators.
Review one destination and its context
Select the exact /guide target: nine observations remain and Northstar falls to 3/9, about 33.3%. The home-page observation is no longer in the denominator. Source-domain grouping does not remove target fragments or query strings, so exact target URLs can remain separate.
Read a selected group’s original variants and source/target pairs. Check the destination and surrounding sentence manually before labeling its meaning clear or mismatched. A blank field, a repeated brand and a commercial phrase are questions to investigate; the worksheet does not infer intent, risk or a required replacement.
Record decisions and export your work
The current-review CSV includes all groups for the selected target; filtered export follows the search, not the visible ten-row page. Both preserve the full target-scope denominator. The evidence export includes every retained external observation across all targets, including original text, attributes, copies and collection context.
Notes and decisions live only in this browser session. Changing file, date, scope, case rule or target filter clears them; download first. Search and pagination preserve notes. CSV formula protection may prefix risky text, so retain the original file privately. Downloads are reports, not restorable sessions.
Repeat with comparable evidence
For later comparisons, keep the same data source, filters, grouping rule and target scope, and record actual export dates. A percentage can change because the denominator changes, even when the anchor count does not. Newly observed wording is not necessarily newly published wording.
Investigate specific source pages with the link monitor and use the reporting template to state evidence, limitations and next checks. No distribution alone proves ranking impact, a penalty or a need to disavow links. The writing guide handles clear labels; this workflow handles dated evidence.
Link writing guide · Backlink export analysis · Reporting template · Link monitor