How to rank company websites by popularity using public data

You can rank company websites by popularity using public data from three published rankings: Google's Chrome UX Report, Cloudflare Radar and the Tranco research list. They count different people in different units, so a sound ranking starts from knowing what each one observes. This post sets out each ranking's population, grain, cadence and licence from its publisher's own pages, then works an illustrative example on five company websites.

The glossary entry on traffic rank defines the term. This post is about method.

How do you rank company websites by popularity using public data?

Pick one public ranking, match each company to the website origins or domains that ranking lists, take each site's rank or band for the same period, and sort. Keep sites with no entry apart from the ranked ones, record the period and the list version, and never mix positions from two rankings in one ordering.

Seven steps make the result repeatable.

1. Reduce each company to its websites. Take the registrable domain of each website, in lower case and without a leading www, as in domain matching, and keep its www and bare hostnames.
2. Choose one ranking for the whole comparison. Choose by the question: page loads by Chrome users, users of one public DNS resolver, or a research list that blends several providers.
3. Look up the key the ranking uses. The Chrome UX Report ranks origins, and an origin is a scheme, a host and a port together, as RFC 6454 defines it, so https://www.example.com and https://example.com are separate entries. Radar and Tranco list domains. Decide in advance which variant stands for the company.
4. Read one period for every site. The same month, the same week of buckets or the same daily list.
5. Sort by band, then by position where the ranking gives one. Sites in one band are tied; leave the tie standing.
6. Keep unlisted sites in their own group. A site with no entry fell outside what the ranking published for the period. Report it as unranked, not as last.
7. Record the provenance. The ranking, the period, the list's identifier, and the licence with its attribution.

The public rankings side by side

Each row is taken from the publisher's own methodology, documentation and licence pages, read on 7 October 2026.

The Chrome UX Report popularity rank

The Chrome UX Report is Google's dataset of how real Chrome users experience popular websites. Its methodology draws on users who have turned on usage statistic reporting, sync their browsing history and have no sync passphrase, on desktop Chrome for Windows, macOS, ChromeOS and Linux and on Chrome for Android. Chrome on iOS, Android WebView and other Chromium browsers such as Microsoft Edge are outside it. An origin enters when it is publicly discoverable and reaches a minimum number of visitors, a number the methodology does not disclose.

The metrics page defines the popularity rank as a relative order of origins within the dataset by total navigations, published as a band on a base-ten log scale in half steps. Each band excludes the bands above it, so the top 5,000 band holds the origins placed from 1,001 to 5,000, and the largest band moves as the dataset grows. Google presents the rank as a guide for broad analysis.

The rank arrived in BigQuery as an experimental metric with the February 2021 dataset, worldwide and for each country, as the announcement explains; a user's country is inferred from the IP address (dimensions). The BigQuery guide calls the rank column a coarse popularity ranking. The datasets are licensed CC BY 4.0, which permits commercial use with credit, a link to the licence and a note of any changes.

Cloudflare Radar domain popularity ranking: methodology and data

Cloudflare Radar ranks domains by an estimate of how many people reach each one, relative to the others, computed from aggregated, anonymised DNS queries to Cloudflare's public 1.1.1.1 resolver. It publishes an ordered top 100, worldwide and per country, refreshed daily over the last 24 hours, and unordered global buckets up to the top one million, refreshed weekly.

The post that introduced the rankings ties popularity to people: the more people interested in a domain, the higher it stands. Below the top 100 a bucket is a set with no order inside it; the bucket after the top 10,000, for example, holds the domains placed from 10,001 to 20,000. Cloudflare explains the buckets by the smaller numbers behind each domain further down the list.

The ranking datasets page lists the bucket sizes and the Radar API endpoints for the top domains and for dataset downloads. A request takes an API token with read access to Radar, as the guide to a first request shows. The Radar documentation states that data from the API is made available under CC BY-NC 4.0, whose NonCommercial term rules out use for commercial purposes.

The Tranco research list

Tranco is a top-sites ranking built for research. Its methodology combines the rankings of the providers it names, the Chrome UX Report and Cloudflare Radar among them, over the past 30 days. Each source list scores its domains by the Dowdall rule, one point for first place, a half for second and so on down to 1/N, and the resulting score sets the order. A bucketed source is converted first: each bucket takes the geometric mean of its two boundaries as a stand-in rank.

The Tranco home page states that a new list of one million domains is published each day by 0:00 UTC, names terms for most of its source lists, and gives every list a permanent ID, so a study can cite the exact list it used. A position in the combined list reflects several sources at once, so read it as a blend and not as the count of any one population.

Free website traffic ranking tools with an API

Three public rankings can be read by program. Tranco's API returns a domain's ranks in recent daily lists with no account, at one query a second. Cloudflare Radar's API serves its top domains and bucket datasets to a Cloudflare API token with Radar read access. The Chrome UX Report rank sits in Google's publicly available BigQuery dataset.

The limits of the Tranco API are in its documentation. Read the terms on the data before you build: CC BY 4.0 allows commercial use with attribution, and CC BY-NC 4.0 does not allow use for commercial purposes. The three rights a buyer of company data usually needs are set out in what a data licence covers.

The Similarweb website traffic ranking methodology and public domain popularity data

Similarweb's Global Rank orders domains by the company's own estimates, a mix of estimated monthly unique visitors and monthly pageviews across desktop and mobile traffic. Its data page lists four kinds of input that its models combine: direct measurement through websites' analytics, data from its own products installed on devices, public data and partner data.

Similarweb's page on its data names the partners as including internet service providers and measurement companies, and describes normalising all the sources into a single view and extrapolating from the signals gathered each day. The Global Rank reference returns one position for a domain, monthly or over the last 28 days, to a request made with a Similarweb API key.

For a ranking of company websites the difference is in kind: the Chrome UX Report and Radar each draw on one population that one publisher observes, and Similarweb's rank draws on several kinds of input. Keep the two in separate columns and name the one you sorted on.

Why two public rankings disagree

The same website can hold different places in the three public rankings, and the methodologies above account for it.

• Different people. Opted-in Chrome users on desktop and Android; users of one public DNS resolver; a blend of providers.
• Different units. Navigations to an origin; an estimate of the people reaching a domain; a score built from positions.
• Different keys. An origin separates the www host from the bare host, and https from http; Radar and Tranco list domains.
• Different windows. A calendar month; the last 24 hours or seven days; 30 days combined.
• Different grain. Sites placed at 10,001 and at 50,000 share a Chrome UX Report band.

So compare a site with itself over time within one ranking, or several sites within one ranking and one period. Two rankings belong side by side as two columns, never merged into one sorted list.

What a band can and cannot support

A band places a site inside a range of ranks. That supports grouping company websites into tiers for account scoring, filtering to the sites in the top 10,000 of a month, counting the companies you track in each band, and spotting a site that crosses into another band between two periods. A band cannot order two sites inside it, show movement within it, or stand for a count of visits.

Worked example: five company websites ranked in one month

This example is illustrative. Acme Robotics wants to see where its website stands among four rivals, using a band ranking and a combined position list.

Assumptions:

• The sites are acme-robotics.example and four rivals, rival-b.example to rival-e.example. All five, and every value below, are invented.
• The month is September 2026. Each site has a band in a ranking built like the Chrome UX Report's, each band excluding those above, taken for the better of its www and bare origins.
• Each site also has a position in a combined daily research list, read on one day of the same month.
• To set a band beside a position, a band takes the geometric mean of its two boundaries as a stand-in rank, the treatment Tranco's methodology gives bucketed sources.

Step 1: stand-in ranks. The top 10,000 band gives the square root of 5,000 × 10,000 = 50,000,000, which is 7,071. The top 50,000 band gives the square root of 10,000 × 50,000 = 500,000,000, which is 22,361. The top 500,000 band gives the square root of 100,000 × 500,000 = 50,000,000,000, which is 223,607.

Step 2: order on the band ranking. rival-b.example comes first at 7,071. Acme Robotics and rival-c.example share 22,361 and are tied. rival-d.example follows at 223,607. rival-e.example has no band and goes into an unranked group.

Step 3: set the positions beside the bands. Three positions fall inside their band's range: 8,400, 31,200 and 44,900. The combined list separates the tied pair, with Acme Robotics at 31,200 ahead of rival-c.example at 44,900. The position of rival-d.example, 88,000, places it higher than its band's range of 100,001 to 500,000. The two differ by design: one counts navigations by one browser's users over a month, the other combines several providers over 30 days. Report both columns and sort on the one you chose.

Step 4: read the next month. In October 2026, assume Acme Robotics moves into the top 10,000 band, rival-c.example moves from 44,900 to 38,000 in the combined list while staying in its band, and nothing else changes. Acme Robotics' stand-in rank falls from 22,361 to 7,071, a change of 15,290, and it now ties with rival-b.example. The 6,900 places gained by rival-c.example show only in the position list. A band records a crossing and is silent about movement inside it.

The monthly traffic tier in Fokals data

Web Traffic, in the announcements and scale product, delivers the monthly traffic tier of each website: the smallest tier the website falls in that month, where a smaller tier means more traffic. Each row names the month, the website's domain, its company and company ID, and carries an attribution. The sourcing statement sets out sourcing and licensing, and the data dictionary defines every field.

Because every row carries the company ID, the tier sits beside firmographic data, technology, hiring and intent on one key, ready to sort by tier, compare across months, and count the companies you track in each tier. The weekly Market Series include the traffic up and traffic down families over 30-day and 180-day windows, each over a same-store cohort.

Frequently asked questions

What this page says about the products it names was checked against their public documentation on 7 October 2026. Product and company names are trademarks of their owners. Fokals is not affiliated with them or endorsed by them.