Hosting concentration is national, not continental: measured Europe-wide the market looks fragmented, yet several countries run on only a handful of effective owners of the customer relationship. Across these thirteen markets, about 5.4 to 25.5 effective hosting infrastructures collapse to about 3.1 to 23.1 effective owners — a gap of up to 2.9× that no storefront shows — though in Spain it runs the other way, with the customer book spread across more owners than infrastructures. Each page states which concentration question every number answers, and the words local buyers actually type.
Every figure is queryable on the free tier via market_concentration,
market_digitization, technology_adoption, market_pricing and
email_security. Named provider shares, per-operator price ladders and the quarterly named
operator ranking live in the product.
Reading across the set rather than one market at a time: European web hosting statistics collects the same concentration readings beside domain populations, website technology, email posture and the consolidation record.
Every market page states its own numbers and its own caveats. The definitions behind them are the same everywhere, so they live here once rather than at the foot of all thirteen.
The denominator is the active-domains book on each national ccTLD — domains that resolve and carry verified activity (operating email, live web content or managed-platform provisioning), with parking and dead pages excluded. It is a domain count, not a website count: a domain qualifies on operating email alone. It is deliberately smaller than a registry total, and it is a floor that rises as measurement progresses, so it is never a count of "all domains in the country". Technology figures use a further subset: the sites whose pages we measure and fingerprint. Email figures use mail-carrying domains. Each page names which denominator each number sits on — they are not interchangeable.
HHI is the sum of squared percent market shares, running 0–10,000; 1,500 and 2,500 are the conventional moderate and high thresholds. Effective competitors = 10,000 / HHI: how many equal-sized players would produce the observed concentration.
"How concentrated is this market" is really three questions with three answers, and mixing them is a category error:
The gap between them is itself the finding: customers can meet many brands and many infrastructures while the economic relationship behind them has already concentrated. All readings are ownership-folded via the validated acquisition ledger plus nameserver brand mapping, which makes every concentration figure a lower bound — owners that are unledgered, or hidden behind misconfigured nameservers, are missed, so true concentration is at least as high as stated. The three readings do not share one base: each is renormalised over its own actor set, so a share from one reading must never be quoted against another.
Adoption reads HTML and CNAME fingerprints: it measures presence, not billing. The website-builder share is a floor — the catalog catches some hoster-native editors via generator tags, but native builders that emit no markup fingerprint are undercounted, so a low builder number can mean "built local", not "less digital". Detection differences are also why builder shares should not be compared across markets without that caveat attached.
Prices come from weekly storefront captures taken from verified in-country vantage points, shared-hosting category only, normalised to ex-VAT monthly EUR at ECB FX. The tracked set is a coverage subset, not a census, so each page states what share of its market's customer relationships the tracked storefronts represent. Medians are suppressed below five priced operators, and a renewal cliff is stated only where five or more operators publish a measurable renewal ladder — a median over three storefronts is an anecdote, not a statistic. Teaser share is a floor: storefronts whose renewal price is never disclosed sit in the denominator. Renewal cliffs are read at three grains — the measured median over storefronts that publish a renewal, a structure-adjusted median at one vote per storefront, and a customer-weighted median — because flat-displaying storefronts are often numerous but small: a storefront-grain figure is a merchandising statistic, not the customer picture. Renewal visibility is reported as three shares of a market's captured customer relationships, never two: storefronts whose second-year price we can read against their first-year price, storefronts that publish a second-year price we are still holding — stated on their page, but on a basis we cannot yet line up, so it is re-read before we compare it — and the rest, which publish no second-year price at all. The middle share is printed on a market page whenever it is anything other than zero, because a measurement gap of ours is not a finding about a storefront.
Prices carry their own date. Storefront captures are re-priced far more often than the market book behind the rest of a market page is rebuilt, so the two would be stated under one date only by accident. Every price figure we publish is dated to the price capture it came from — the price snapshot beside it — and a page's headline snapshot date covers concentration, market size, technology and email, never pricing.
SPF and DMARC shares are of each market's mail-carrying business domains. Adoption is not
enforcement: only a hard-fail SPF policy and DMARC p=reject actually reject forged mail,
while p=none is monitoring only. The gap between publishing a record and rejecting forged
mail is where the risk lives.
Each market's word cloud is the vocabulary local hosting providers put in their own page titles and headings, measured from live storefronts, one vote per provider. Term size encodes how many providers use a term; the counts themselves are not printed. It is evidence of what providers optimise for, which is not the same as search volume.
Everything here is aggregate-level. Named provider shares, per-operator price ladders and the quarterly named operator ranking are the product, not the public page. Percentages are of the stated denominator and of the stated snapshot date — not of a registry, and not of a complete customer book.