After this lesson you should be able to
- Explain why total backlinks misleads and referring domains does not
- Recognize a sitewide placement from the shape of the two charts
- Track a short list of metrics that reflect real editorial decisions
- Ignore the metrics that describe an index rather than your site
The number that misleads most often
Total backlinks is the headline figure on every dashboard and the least useful number in link building. It counts individual link instances, which means it is dominated by placements that are repeated across a site rather than by editorial decisions.
Here is a concrete example from a real profile. The backlink history chart shows roughly 830 external referring URLs appearing in a single day in late July, against a baseline under 100. That looks like the best day of the year. Now look at the referring domains series for the same day: no spike at all. The domain series peaks at 26, months earlier.
A backlink spike with no matching referring-domain spike is one site adding a sitewide link. A footer, a sidebar widget, a blogroll entry, a template change. One person made one decision and it propagated across 830 pages. In a dashboard it is a triumph; in reality it is one link, repeated.
This is not an edge case. On the same profile, the single largest contributor sent 219,159 links, which is 16.8% of a 1.3 million link profile, from one sitewide placement following one post. The top 100 referring domains account for 84.3% of all links ever recorded, out of 22,260 domains. If you report link counts, you are mostly reporting the behavior of a hundred sites, and often the behavior of one.
Count domains, not links
The referring domain count approximates the number of independent decisions to link to you, which is what you actually care about. It is not perfect — a domain can link from ten genuinely separate articles, and that is worth more than one sitewide placement across ten thousand pages — but it is enormously more honest than a link count and it is far harder to inflate.
Two habits follow from this.
Read the two charts together, always. Links and referring domains, same period, same scale of attention. When they move together, something real happened. When links move and domains do not, a site changed a template. When domains move and links barely move, you gained a set of individual editorial links, which is usually the best news on the page and the least visually impressive.
Report domains as the headline and links as a footnote. This is not a cosmetic preference. It changes behavior: a team measured on link count will accumulate sitewide placements and directory entries, because those are the cheapest way to make the number move. A team measured on referring domains has to go and persuade people, which is the work.
The same logic applies when you look at competitors. A competitor with five times your link count and the same referring domain count has not outperformed you; they have a widget on somebody's site. Before you conclude that you are behind, compare the two numbers a competitor cannot inflate: referring domains, and how many of those domains sit above a meaningful trust threshold.
The short list worth tracking
Six measures cover almost everything a link program needs to know about itself. Anything beyond this list should have to justify its place.
- Referring domains, live. The headline. Track the number and the trend.
- Referring domains gained and lost per period. Net, not gross. A period with 20 gained and 18 lost is not a good period, and gross reporting hides that entirely.
- Trust Flow distribution of new domains. Not just how many, but what kind. A quarter that added 30 Trust Flow 0 domains and a quarter that added 8 domains at Trust Flow 30 and above are very different quarters, and the second is better on every dimension including durability.
- Topical relevance of new domains. Judged by hand on a sample, or via Topical Trust Flow categories. Volume without relevance is not progress.
- Links from pages that get traffic. A link on a page nobody visits and nobody links to is worth having but is not worth celebrating.
- Anchor distribution over time. Watched for drift, not managed to a ratio. Its main use is as an early warning of somebody else pointing links at you.
Notice that none of these is a single score. Proprietary composite metrics are useful for sorting a prospect list quickly, and misleading as a program measure, because a change in the score can mean the index recrawled something rather than that anything happened to your site.
What to ignore
Total backlinks as a headline. Covered above. Keep it in the report as context, never as the number people look at first.
Small movements in any authority score. These scores are recomputed as the index changes. A one or two point move usually means the crawler saw something different, not that your site changed. Watch the trend over quarters and ignore the week.
Toxicity and spam scores. These are third-party estimates presented with unearned confidence. On a profile where no link was ever bought, 69.3% of referring domains sit at Trust Flow 0 and would be flagged by most such tools. A score that flags the normal condition of the internet is not measuring risk.
Nofollow ratio as a target. There is no correct ratio, and the people who report one are usually reverse-engineering it from whatever their own profile happens to show. Coverage that arrives nofollowed still brings readers, still gets syndicated, and still gets seen by the people who link later.
Link velocity as a safety metric. The idea that acquiring links quickly is inherently suspicious does not survive contact with how real coverage behaves. A single story produces a burst; that is what a story does. Velocity matters only in combination with the other signals in lesson 8.3.
Tying links to outcomes without overclaiming
The honest position on attribution is that you can demonstrate correlation carefully and you cannot demonstrate causation cleanly, because you cannot run the counterfactual. Anyone who shows you a clean causal chart of links to revenue has hidden something.
What you can do is build a defensible case in layers. Start with the campaign metric, which is referring domains gained and their quality. Move to the page-level metric, which is ranking and organic traffic movement on the specific URLs that received links, compared against comparable URLs on the same site that did not. Then the site-level metric over quarters. And finally referral traffic, which is small in absolute terms but is the one number that is unambiguously real, since it can be attributed to a specific link.
Two cautions. Lag is genuine and variable: links take an unpredictable time to be crawled, weighed and reflected, so a monthly reporting cycle will show you noise. And a link's value is not confined to search at all. A placement in a publication your buyers read has value that no link metric captures, and I would rather have that link than five that score better.
Questions
Why did my backlink count jump by hundreds overnight?
Almost certainly one site added a sitewide link: a footer, sidebar or blogroll entry that appears on every page. Check the referring domain chart for the same day. If the domain count did not move, you gained one link repeated across many pages. On one real profile, roughly 830 referring URLs appeared in a single day with no corresponding domain spike at all.
Are referring domains always better than links?
As a headline measure, yes. Referring domains approximate independent decisions to link, while link counts are dominated by repetition. The refinement worth making is that several genuinely separate articles from one domain are worth more than one sitewide placement, so for important domains it is worth looking at distinct linking pages rather than assuming one domain means one link.
Should I worry about my spam or toxicity score?
Not on its own. These scores are third-party estimates, and they flag conditions that are entirely normal. On a twenty-year profile where no link was ever bought, 69.3% of referring domains are Trust Flow 0 and would trip most toxicity thresholds. Use the score to sort a list for manual review if you like, never as evidence that something is wrong.
How long after a link should I expect to see a ranking effect?
Unpredictably long, which is why monthly reporting on rankings produces mostly noise. The link has to be crawled, the page reassessed, and the effect combined with everything else changing on the site and in the results. Judge over quarters, compare linked pages against similar unlinked pages on the same site, and treat any confident short-term causal claim with suspicion.