Read the backlink profile and its two different histories
Several link sources
Bounded source structure
Two history types
Separate summary, detail, and history sources
The backlink headline and profile facts come from Backlinks Summary. Anchors, referring domains, and indexed target pages are three separate detail modules. Organic and paid monthly growth comes from Rank History. A Domain profile can select each eligible module from a different successful run, so one screen can legitimately contain several provider dates and last-known sources.
Carry the module notice and persisted coverage with every claim. In the current detail implementation, the referring-domain aggregate also looks for target-page rows stored in that same source run. The target-page table and Site Audit health check follow the separately selected Indexed Pages source. If a composed profile splits those sources, their target summaries can have different bases.
Backlinks, referring domains, and main domains use different denominators
Backlinks counts observed links. Referring domains counts provider-level source domains, while referring main domains collapses host variation toward the main-domain level. Referring IPs is another diversity signal. Sitewide links, repeated links, and several hosts under one organization can make the backlink total grow much faster than source diversity.
Backlinks per domain divides the Backlinks Summary total by its referring-domain total. It is a concentration clue, not an average quality score. Authority uses the provider backlink rank transformed from 0 to 1,000 into the 0 to 100 display described in the core-metrics guide.
- First Seen is the provider's earliest observation in this source, not a guaranteed publication date.
- Broken backlinks and broken pages are provider observations that need a live target and source check before recovery work.
- Crawled pages, referring IPs, spam score, and main-domain totals describe different parts of the provider graph and must not be added together.
Follow and nofollow are attributes, not quality verdicts
The overview fact stripe calculates its visual split from follow plus nofollow counts. The detailed KPI for follow share divides follow by the headline backlink total. Those denominators can differ when provider totals and attribute-classified counts are not identical, so cite the counts and basis rather than copying a bar width.
Follow does not guarantee ranking value, and nofollow does not make a mention useless. Placement, page status, topical relevance, source legitimacy, sponsorship, user-generated context, redirects, and the current destination determine what the observation means.
The three link ledgers are bounded evidence tables
Top Anchors lists anchor text and observed backlink count. Referring Domains adds the provider backlink rank and link count. Most Linked Pages lists target title or URL and link count. Each ledger is paginated, its visible bars are relative to the strongest loaded row, and one shared query and sort state controls the active working view.
Current plan defaults retain up to 100 anchors, 100 referring domains, and 50 indexed pages. Code caps are 500, 500, and 200 before an effective policy can reduce them. The module manifest distinguishes provider total, returned rows, persisted rows, and limit where available. A missing row is not a measured absence from the full link graph.
| Evidence | Question it can support |
|---|---|
| Anchor text | Is the retained pattern branded, descriptive, generic, repetitive, empty, or suspicious enough to sample? |
| Referring domain and rank | How diverse is the retained source set, and where does its provider rank sit? |
| Indexed target page | Which retained destinations receive links and how concentrated are referring domains? |
| Loaded link bar | Which visible row has more observed links relative to this loaded set? |
Detail distributions describe the stored set or summary source
The authority chart aggregates every persisted referring-domain row with a rank, up to the bounded storage scope. On the stored 0 to 1,000 scale, its bands end at 200, 400, 600, 800, and above 800. The detail table displays that rank on the transformed 0 to 100 scale, so keep the scale visible when exporting a range.
Link Spread groups persisted referring domains by links contributed: one, two to five, six to twenty, and twenty-one or more. Top Targets ranks at most ten target rows by unique referring domains and then backlinks. Summary distributions for link type, platform, country, TLD, and semantic placement display at most eight positive categories in each donut. These are composition views, not independent totals to sum.
Target health crosses link evidence with a separate Site Audit inventory
The backlink target health query chooses up to fifty indexed-page rows by referring-domain count and then backlink count. It resolves them against the current workspace Site Audit URL inventory, counts matched and unmatched targets, sorts problems first, and returns at most twenty health rows. The UI calls out redirects and HTTP statuses of 400 or higher.
The backlink evidence date and Site Audit update date remain separate. An unmatched URL is not classified as healthy. A redirect can be intentional, and a broken target can be irrelevant or high value. Check canonical destination, link context, redirect chain, issue history, and ownership before changing routing or content.
A manual link review starts with a stratified sample
Monthly Growth is provider ranking history, not a chain of Crawl Foundry runs
The overview Growth band reads stored history-month rows from the Rank History module. The current plan defaults to 60 months and caps collection at 120. The window selector shows the newest 6, 12, 24, or 60 available months without buying or interpolating missing history.
Its stacked distribution groups Top 3, positions 4 to 10, 11 to 20, 21 to 50, and 51 to 100. The upper edge is the sum of these position bands. Six companion charts show organic ETV, organic keyword count, Top 10, Top 3, provider new versus lost counters, and paid ETV. An all-zero series becomes a quiet state rather than an invented trend.
- The history module has its own source run and provider month; it can differ from the current rank-overview headline.
- New and lost are provider monthly classes, not a complete explanation of content gains or losses.
- Paid ETV provides market context and cannot explain an organic change by itself.
The History tab derives ratios from the same stored months
The detailed History tab loads up to 121 dated month rows, sorts them chronologically, and defaults to a 24-month window. Month-over-month traffic rhythm is a percentage change; the first point and a previous value of zero remain undefined. Net keyword balance is new minus lost. Top 3 and Top 10 shares divide their buckets by the sum of all eight ranking-position buckets.
Traffic efficiency is ETV divided by organic keyword count. Value density is monthly estimated paid traffic cost in provider USD divided by ETV. Churn is new plus lost divided by organic keyword count. These are transparent derived ratios, not Crawl Foundry scores, causal models, or forecasts.
- A zero denominator yields no percentage or a displayed zero according to the stated ratio, not an infinite change.
- Year-over-year traffic requires at least thirteen dated rows and at least two matching calendar months; it compares the newest available period with the same months one year earlier.
- Seasonality, provider coverage, SERP changes, and model revisions remain alternative explanations.
Backlink history is a separate stored-analysis series
The History tab's backlink charts do not come from monthly Rank History. They select retained non-failed Crawl Foundry analyses for the same normalized domain, location, and language, order them by analysis creation time, and plot Backlinks Summary totals, referring domains, main domains, and transformed Authority.
Those observations can be irregularly spaced and can come from different collection runs or coverage conditions. A line between them aids orientation but does not create missing monthly measurements. Record each analysis date and source state when discussing link growth.
Triangulate a material change across clocks
Missing history and measured zero are not the same
A missing module, failed collection, unavailable provider field, compacted detail run, absent calendar month, or insufficient comparable analyses can all remove a chart. A stored zero can also be real. Read module state, source date, row coverage, window, and quiet-state message before reporting decline, stability, or no links.
Domain Overview supports evidence selection and comparison. It does not verify link quality, penalization, causation, revenue impact, or the success of a recovery action without source-page inspection and owned outcome evidence.