Link metrics explained

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

The one thing to understand before any of them

No search engine uses any of these metrics. Not one. Trust Flow, Domain Rating and Domain Authority are third-party estimates built by companies that crawl the web independently and model what they think matters. Google has its own internal signals, does not publish them, and has said repeatedly that the public scores are not inputs to its ranking.

That does not make the metrics useless. It makes them proxies, and a proxy is valuable exactly to the extent that you understand what it is standing in for. A metric built from an index of ten billion pages and a metric built from an index of thirty billion will disagree about the same site, and neither is wrong — they are measuring different maps.

The practical consequence: use metrics to sort and to disqualify, never to decide. A metric can tell you that a prospect list of four thousand contains three hundred worth a human look. It cannot tell you whether a specific link is worth having. That judgment is relevance, and no vendor sells it.

Trust Flow and Citation Flow

Citation Flow is Majestic's measure of link volume, scored 0 to 100 on a logarithmic scale. It counts how much link equity flows into a page or domain without regard to where it came from. It is a quantity signal, and quantity is cheap to manufacture.

Trust Flow is Majestic's measure of link quality, also 0 to 100, computed from proximity to a manually reviewed seed set of trusted sites. The closer a site sits in the graph to those seeds, the higher its Trust Flow. Because the seed set is curated by humans, Trust Flow is much harder to inflate than Citation Flow — you would have to acquire links from sites that are themselves close to trusted seeds.

What Trust Flow misses is topic. A high Trust Flow domain in an unrelated field is still a high Trust Flow domain, and the number gives you no warning. It also inherits the biases of the seed set: certain sectors, languages and regions are better represented than others, so a perfectly reputable niche site in an underrepresented market can score low for reasons that have nothing to do with its quality.

The other thing worth knowing is how ordinary low scores are. On the twenty-year, never-bought profile used throughout this site, 69.3% of the 22,260 referring domains were Trust Flow 0. Zero. On a completely clean profile. Anyone treating TF 0 as evidence of a toxic link is misreading the metric badly.

The TF/CF ratio

Dividing Trust Flow by Citation Flow produces the single most useful derived number in link analysis, because it compares quality against quantity on the same scale from the same index.

A ratio near or above 1.0 means the site's link volume is matched by trusted sources. A ratio around 0.5 is unremarkable — plenty of legitimate sites sit there. A ratio well below 0.3, particularly with a high Citation Flow, means the site has accumulated a large amount of link volume from sources that are nowhere near trusted seeds. That is the classic signature of a domain whose authority has been built rather than earned.

Two cautions. First, the ratio is unstable at low numbers: TF 1 against CF 2 gives you 0.5 and means nothing at all. Ignore the ratio below roughly CF 15. Second, a good ratio is not a certificate. Ratios can be gamed by someone who understands the metric, and I have seen networks built specifically to produce flattering flow numbers.

Topical Trust Flow

Topical Trust Flow breaks the Trust Flow score down by subject category, telling you not just how trusted a site is but what it is trusted about. A site with Topical Trust Flow concentrated in Health is being cited by health sites; the same overall score concentrated in Computers means something completely different for a medical client.

This is the closest any public metric gets to measuring relevance, and it is genuinely useful for that. The specific way it misleads is rank order. The categories are ranked, and people read the top category as a verdict when the numbers behind ranks two and three are often nearly identical, or when the top category reflects one enormous historical link rather than the site's actual subject. Always read the values, not the ordering.

It also inherits the taxonomy's limitations. The category tree is fixed and coarse, so a site in a specialty that does not map cleanly onto a category gets sorted into the nearest available bucket, which can be misleading in both directions.

Domain Rating and Domain Authority

Domain Rating (Ahrefs) scores a domain 0 to 100 on the strength of its inbound link profile, on a logarithmic scale, calculated from the number of unique linking domains weighted by those domains' own DR and dampened by how many sites each links out to. It is fundamentally a volume-and-strength model. It does not attempt to model trust, and it does not model topic at all.

Domain Authority (Moz) is different in kind. It is a machine-learning model trained to predict how likely a domain is to rank in search results, using link data as its input features. That makes it a prediction rather than a measurement, and it means the score can shift when Moz retrains the model even if nothing about the site changed. DA is also the metric most polluted by the link-selling market, because it has been the headline number in link pricing for over a decade — which gives vendors a direct financial incentive to inflate it, and inflating it is not especially difficult.

Majestic Million is a different animal again: not a score but a ranked list of the top million websites by referring subnet count, published freely and updated daily. Its value is as a sanity check. If a domain a vendor describes as authoritative does not appear anywhere in a list of a million sites, that is worth knowing.

How the metrics disagree, side by side

The metrics disagree constantly, and the disagreements are informative once you know what drives them.

Three common disagreement patterns, and what each usually means:

• High DR, low TF. Many linking domains, few of them near trusted seeds. Often a site that has been the target of bulk link building — sometimes its own, sometimes not.
• High TF, modest DR. Few links but excellent ones. Frequently a small institutional, academic or long-established niche site. These are usually better prospects than the numbers suggest.
• High DA, low everything else. The pattern I trust least. DA is the most actively manipulated of the public scores because it is the one link sellers quote.

Using them without being fooled

My working rules, after fifteen years as a Majestic Brand Ambassador and rather longer arguing about this.

Never use a single metric as a threshold. The moment you tell a team to acquire links above DR 40, you have told them what to buy, and a market exists to supply exactly that number at exactly that price.

Always pull two metrics from different vendors. Agreement raises confidence; disagreement is a prompt to look manually, which is where the real judgment happens.

Check the page, not just the domain. All of these scores are usually quoted at domain level. A link from an orphaned page nobody links to internally, on a DR 80 domain, passes close to nothing. Domain-level metrics say nothing about the specific page your link will sit on.

Remember the base rate. Two thirds of a clean twenty-year profile was Trust Flow 0, and only 446 domains — 2.0% — ever reached Trust Flow 41 or higher. That is roughly twenty-two strong domains a year from constant publishing. Calibrate your expectations to that, not to a vendor's prospect list.

Module 2 of the course covers judging link quality end to end, including a scoring rubric that weights relevance above every metric on this page. Lesson 3.3 goes deeper on the flow metrics specifically.