Accuracy

How do we calculate the score, and how far can you trust it?

The authority score is a single number from 0 to 100. This page explains in plain language where that number comes from, why it is the same everywhere, and what the range next to it tells you.

1 · Consistency

The same score on every surface.

You see a domain's score in the overview, in the comparison, in the bulk query and in the reference lists. They all ask the same single calculation service; the pages have no calculation of their own.

So you never see two different numbers for the same domain on two pages. If you do, that is a bug; write to us.

2 · Stability

The score does not change unless the data changes.

The score is not produced by rolling the dice again every time you open the page. As long as the data underneath stays the same, the number you see today is the same tomorrow.

The score moves in two cases: when the link index or the domain's data is refreshed, or when we update the calculation method. If the method changes, all scores move to the new method at the same time; half are never left old and half new.

3 · Range

An approximate score comes with its measured range, if there is one.

We do not have data of the same depth for every domain. When we calculate a score with data missing, we call it approximate. Where we have measured the error, we also put a range next to it: "the score is most likely between these two numbers".

The width of the range is not a guess; it is set by measuring the method's own error. Where we have not measured, we do not write a range, we only mark the score as "approximate".

Components

The score looks at three things.

All three are measurable and are shown separately in the mini report. The raw total goes through a scaling curve; that way the strongest sites approach 100, but hundreds of them do not all show a flat 100.

When more detailed data for a domain (the real number of referring domains, the backlink count, estimated traffic) reaches us, the score is recalculated with that data. Then the approximate label goes away and the range narrows.

  1. Network positionHow central is the domain in the link network? A domain that can be reached by a short path from many sites sits closer to the center.
  2. Strength of linking domainsWho links to it? A link from strong sites counts for more than many links from weak sites.
  3. Popularity indexWhere does the domain stand in the ranking of the most visited sites in the world? For a domain with a weak place in the link network, this component alone does not carry the score.

Other numbers

How to read the rest of the report

Number of referring domains

On every page you see a single number in the heading. Next to it, the source of the number is written. The referring domains we can list may be only part of this number; we also tell you how many we listed.

Live referrers

We check how many of the referring domains are still live. "Not live" is said only for domains whose registration does not resolve. A domain we have not crawled yet is counted as live; we do not call something "dead" when we do not know.

Spam heading

In the mini report, the spam heading appears only if an audit was done earlier for that domain. Otherwise we leave it empty; we do not say "clean" about a domain we have not looked at.

Limits

What the score does not say.

  • It is not a promise of rankings. A high score does not mean a page will appear at the top of search results.
  • It is not the same scale as other tools' scores. Every tool calculates with its own data and its own formula. Putting the numbers of two different tools side by side is misleading; comparing two domains inside the same tool is meaningful.
  • It is not a quality verdict. The score summarizes link and popularity signals. It does not say whether the content is good or whether the domain's history is clean; for that there is the spam history audit.
  • Uncertainty is higher for small and new sites. For a domain with little data about it, the range is wider.

If you think a score looks wrong, send us the domain. If we find the error we fix the method, and we apply the fix to all scores together.

Kosto, in a miner's helmet, with a calm expression
I do not make up numbers for what I do not know.

Try it on your own domain.

See the score without signing up, its range if measured, and the three components.