The measurements behind the Hosting Anatomy series

HostingBrain evidence brief · data snapshot 2026-08-23

Standfirst. The Hosting Anatomy series (published by Jonas Zimling) interprets the structure of the large hosting groups consolidating European markets. This page holds the instrument readings the articles draw on: what was measured, on which population, with which definitions, and where the measurements stop. The argument lives in the articles. The evidence lives here.

Data version: stated in the live figures panel below. Every figure on this page is recomputed from the currently served dataset on each site build.

The sample

The readings cover the hosting groups currently profiled, measured at brand grain — each consumer brand read separately, then rolled up — because group averages hide exactly the variation the series examines. The count of profiled groups is in the live figures panel below.

The profiled set is not a geographic classification. It is the mapped consolidator set our industry-map instrument serves — groups selected for their consolidation activity in the hosting market, including groups headquartered outside Europe. No instrument classifies groups by geography, so no figure on this page carries a geographic filter. Where the articles say "European", that is the series' editorial focus, not a served selection rule.

Two instruments, two sets. The structural readings come from two served instruments — the industry map (cohesion, own-network serving share) and the integration-depth scoreboard (platform hub share) — and their profiled sets are not identical: on the current dataset they differ by one group each way. Every range on this page names the instrument it is computed on, and the size of the membership difference is itself a published figure in the panel.

Counts and shares on this page always name the population they are computed on. The book types used here (active domains, the live-business book, its page-observed slice, the mail-carrying book) are defined once on the data methodology page; this page states which book each figure uses and does not redefine them.

Three kinds of statement

Every reading on this page is one of three things, and the wording keeps them apart:

The anatomy labels the articles use (vertically integrated, federated, orchestrator) are interpretation built on top of these metrics. They appear in the articles, not here.

The metrics

Platform concentration (hub share). Calculated, on the integration-depth scoreboard. The share of a group's book — its nameserver-signature domain book — resolving through its principal shared platform. Some groups have no identified principal platform at all: a book dominated by cloud-masked serving is a detection limit, not a zero, so those groups are counted beside the range and never ranked inside it. The range across profiled groups, and the count of no-identified-hub groups, are in the panel.

Brand-stack cohesion. Calculated, on the industry-map instrument, over the live-business book, size-weighted per brand. Within one brand, the degree to which the observed layers (DNS, web serving, mail) follow one coherent design. Range across profiled groups: in the panel.

Own-network serving share. Calculated, on the industry-map instrument. Of the web-provisioned domains on the group's own nameservers — the domains we observe because they were web-provisioned — the share served from network space the group itself operates. This base is a FLOOR: purely-parked or DNS-only domains the group controls but that never entered our web-provisioned book are not counted. Domains served from third-party clouds count toward the cloud, not the group: an operator serving customers on rented capacity reads low here by construction, which is a definition, not a defect. The range and each extremum's base count are in the panel.

Migration stage. Derived classification, per brand, on the integration-depth instrument. Where an acquired brand's stack stands relative to the group platform, in the instrument's own vocabulary: standalone, mail-migrated, hosting-migrated, or hosting-and-mail migrated — plus cloud-masked, which is a detection limit, not a stage: a brand whose serving is hidden behind cloud or CDN infrastructure cannot be placed either way and is never counted as migrated or as standalone. The residual standalone count is the calculated metric on top.

Layer build-or-buy. Derived classification, per brand and layer: whether the group operates that layer's machinery or buys it. Asked only where the evidence supports the question; a brand we cannot evidence either way is recorded as unresolved, never guessed.

SaaS attachment. Calculated. Fingerprints of monetizable services on fetched pages, divided by the page-observed slice of the live-business book. Every rate is published beside its measured base and its coverage.

Book quality (page states). Calculated, on the page-observed slice of the live-business book. The share of observed pages carrying no real site, split by what the page is: a platform's own default page on a provisioned, unbuilt domain (a placeholder), or a decayed page — expired, suspended, an error page, or parked for sale. The two halves are published separately, on the same base, because they describe different things. Pages whose fetch was answered by an anti-bot interstitial are counted in neither half: nobody has read the page behind them, so they are excluded from the base and published as a separate count — a floor, because interstitials are recognised from a curated list and new ones appear.

Email posture. Calculated, on the mail-carrying book. Two separate readings: presence (an SPF or DMARC record exists) and enforcement (SPF ends in hard-fail; DMARC policy rejects or quarantines). A record at monitoring-only counts toward presence and not enforcement.

Customer movement. Raw observation with a calculated incidence. DNS delegation changes between operators. Direction is measured; calibrated rates are not yet published, so every movement figure on this site is direction-only.

What each article uses

Article 1 — Why brand and domain counts mislead. Uses the ranges of hub share, cohesion and own-network serving share across the profiled set, against active-domain book sizes. Hub share is read from the integration-depth scoreboard; cohesion and own-network serving share from the industry map — the two instruments' sets differ by one group each way, and each range is computed on its own instrument's set. The measured point: groups of similar book size sit at opposite ends of each structural range.

Article 2 — The three anatomies. Uses per-brand stage classifications, hub share and residual standalone counts on multi-country groups. The measured point: platform concentration and residual standalone count vary independently of the number of national markets in which a group holds a leading brand.

Article 3 — Architecture and business signals. Uses attach rates beside their bases and coverage, against hub share and migration depth. The measured point: attach rates do not rise monotonically with concentration or migration depth across the profiled set.

Article 4 — Email and SaaS as secondary indicators. Uses the presence-versus-enforcement split and the attach readings against the primary architecture metrics. The measured point: both signal families can diverge from hub share and stage mix on the same groups.

Article 5 — What outside-in data can and cannot establish. Uses the full instrument list and the boundary statement below. The measured point is the boundary itself: what the wire records and what it cannot.

Coverage and comparability

Page-observation coverage differs across groups. Attach percentages are coverage-qualified: each is published beside its measured base and its coverage, and where coverage differs materially between two groups the percentages are published side by side with their coverage rather than compared directly. No fixed comparability threshold has been established.

Figures on different book types are never averaged or added. A rate on the live-business book and a count on the active-domains book are different instruments.

Every absolute count on this page carries its book type and its as-of date via the panel below.

Limitations

The instruments observe public infrastructure from the outside. They record DNS signatures, fetchable pages, network paths and mail-security records. They do not record:

Latency and traceroute can constrain or suggest likely ingress and network-path characteristics; they cannot establish a local presence, physical PoP, facility ownership or capacity.

Parked and idle domains without content sit outside the live-business readings. Movement figures are direction-only. Coverage bounds are published beside every rate they qualify.

The open questions

Because of that boundary, every group profile carries an open-question list beside its measurements — the questions the wire cannot answer, each attached to the observation that raises it. An answered question is recorded with its answer; it is not deleted. What we cannot answer stays on the list, visibly open.

Live figures

exact figures re-cut from the served release each build; the argument above is written to survive them.

FigureValueBaseAs of
hosting groups profiled (industry map)15the mapped consolidator set of the industry-map instrument; membership is instrument-specific2026-08-23
hosting groups on the integration-depth scoreboard15hosting-group rows of the integration-depth scoreboard — a set that differs from the industry map's2026-08-23
groups in exactly one of the two instruments' sets2symmetric difference of the two profiled sets2026-08-23
platform hub share — minimum19.2%share of the group's ns-signature domain book resolving through its principal shared platform — integration-depth scoreboard, hosting-group rows with an identified hub2026-08-23
platform hub share — maximum100%share of the group's ns-signature domain book resolving through its principal shared platform — integration-depth scoreboard, hosting-group rows with an identified hub2026-08-23
groups with no identified principal platform4hosting-group rows of the integration-depth scoreboard2026-08-23
brand-stack cohesion — minimum0.116brand-stack cohesion, live-business book, size-weighted per brand — industry-map instrument2026-08-23
brand-stack cohesion — maximum0.833brand-stack cohesion, live-business book, size-weighted per brand — industry-map instrument2026-08-23
own-network serving share — minimum0%of web-provisioned domains on the group's own nameservers, the share served from network space the group itself operates — industry-map instrument2026-08-23
own-network serving share — maximum86.7%of web-provisioned domains on the group's own nameservers, the share served from network space the group itself operates — industry-map instrument2026-08-23
the minimum group's base89,938web-provisioned domains on the group's own nameservers — minimum group2026-08-23
the maximum group's base558,731web-provisioned domains on the group's own nameservers — maximum group2026-08-23

Data version 2026-08-23 · ranges are extrema of the profiled set on the named instrument; membership moves between releases and the two instruments' sets differ — every range names its instrument. Every figure here is derived from the served release by this brief's own derivation script; nothing on this page is hand-typed.

Reproduce this

The figures above are served by the HostingBrain connector: the industry-map readings by the consolidation-landscape tool and the hub-share readings by the integration-depth tool (available on the connector's paid tiers — the tier table on the get-started page says which). Ask it:

Ask the connector

"For the profiled hosting groups, show platform concentration, cohesion and own-network serving share with book types and as-of dates."

"Show SaaS attachment for a market or group with its measured base and coverage."

Request a briefing

Group-level profiles and the per-group open-question lists are part of the briefing product.

Share this pageShare on LinkedIn