How OAR is computed

The data source

Common Crawl publishes a web-scale hyperlink graph with each crawl release, roughly quarterly. The domain-level ranks file lists over 120 million registrable domains with two independent signals: a harmonic-centrality rank and a PageRank rank within that release’s graph. The files are openly downloadable from commoncrawl.org, so any score here can be checked against the raw data.

The formula (OAR v1)

Each rank maps to a logarithmic position between 0 and 1, so rank 1 scores highest and the long tail decays smoothly. The two signals combine with harmonic centrality as the primary weight, because it is much harder to inflate with link farms than PageRank:

f(rank) = 1 - ln(rank) / ln(N)
OAR = round(100 x (0.7 x f(hc_rank) + 0.3 x f(pr_rank)))

N is the total number of domains in the release’s graph. The result is clamped to 0 to 100. The algorithm version is stored with every score, and v1 is frozen: any change to the formula, the weights, or the missing-data policy gets a new version number rather than silently rewriting history.

Refreshes

When Common Crawl publishes a new release, every known domain is re-scored against it and a new point is added to its history. Scores are per release, so a trend chart shows how a domain moved across crawls, not day-to-day noise.

Honesty rules

• A domain that is provably absent from a release’s graph gets a gap in its history, never an estimated number.
• If we cannot prove absence (for example while a release is still being indexed), the page says so instead of guessing.
• Scores are never adjusted manually, bought, or boosted.

What OAR is not

OAR is not Ahrefs Domain Rating and not Moz Domain Authority. Those metrics are built on proprietary crawlers and private link indexes, and their numbers are not comparable to OAR or to each other. OAR is an independent scale; compare OAR to OAR.

Questions this page doesn’t answer are probably on the FAQ.