Hosting customers don't leave
Every hosting provider can measure its own churn. None of them can measure faithfulness — because the question lives outside any one provider's book. When a business owns several domains, does it keep them all in one place, or spread them across providers and countries? Only someone who can see the owner across every provider can answer that. We can.
A majority are faithful — but it's not universal
Across 2,212,091 owners of two or more live business domains, 64.7% keep all of them with a single provider, and the average owner concentrates 85.3% of their domains with their primary one. That still leaves a measurable minority — 35.3% — who deliberately multi-home, and 3.7% who spread across three or more providers. For a provider, that minority is the entire upsell-and-defend map: the customers already half-somewhere-else.
Churn tells you who left. Faithfulness tells you who was never fully yours to begin with.
The surprise: borders barely matter
You would expect a business operating across several countries — a national domain here, a regional one there, a .com — to pick a local provider in each market. It mostly doesn't. Among the 1,259,904 owners whose domains span more than one country-code, 62.0% still keep every one of them with a single provider — set that beside the 64.7% across all multi-domain owners and the whole effect of crossing a border is the gap between those two numbers. Cross-border complexity, which should fragment the relationship, barely does. The provider that wins a multinational's first domain tends to keep the rest, wherever they operate.
Faithfulness is a book-quality signal — and it varies
Measured per provider, the share of a provider's multi-domain customers who use it exclusively ranges from 64.9% at the top down to 40.5% for a pure cloud/DNS layer. Those are the ten books charted below — the eight largest hosting books and the two largest CDN/DNS layers, one denominator for every figure in this paragraph. Across them, no hosting book falls below 46.5%, while the layers top out at 48.7% - so the two fields overlap: the top layer sits above the lowest hosting book. That is book quality, orthogonal to book size: a smaller book of exclusive customers is stickier, more upsell-ready and more defensible than a larger book of always-shopping ones — exactly the distinction an acquirer or lender wants and cannot get from a raw domain count.
The ten largest books we track, as archetypes rather than named providers (named scores are a product-tier feature), ordered by exclusivity. Faded bars are CDN/managed-DNS providers — a layer run alongside a host, where low exclusivity is expected, not weakness.
One honest read of the same data: CDN and managed-DNS providers show low exclusivity, and that is correct — they are a layer run alongside a real host, not instead of one. We flag them, because reading their low score as churn risk would be a mistake.
And it's conscious — customers are loyal to the brand, not accidentally to the owner
A fair challenge: consolidators own many brands, so a customer with domains at two sibling brands would count as "loyal" to the group without knowing it. Measured: that case is nearly empty. Of group-loyal owners, 99.3% keep everything within a single brand — the loyalty is a relationship the customer actually chose, not an artifact of hidden ownership. The flip side is a finding for the consolidators themselves: cross-brand wallet expansion inside a group is close to nonexistent — the remaining 0.7% is every loyal owner who spans two brands of one group. Multi-brand portfolios run as silos, and the second domain goes to the brand the customer already knows.
Revisions. 2026-07-06: the figures first published that day were deflated by an unattributed-provider bucket — domains whose provider we could not identify counted as a distinct "provider", overstating multi-homing — and were corrected upward the same day. 2026-08-02: every figure on this page was re-pulled from the served release and is now held against it by the publishing gate, after a period in which the page quoted an older release than the tool its own prompt sends you to. The superseded values are recorded in the changelog, which is where they belong; this page states only what
customer_faithfulness returns today, and the reproduce-prompt below is the
check. 2026-10-04: until this release the exclusivity paragraph said every hosting book sits above both
layers. On this data the lowest hosting book came level with the top CDN/DNS layer, a tenth of a point
below it, so the paragraph now says the two fields touch and the chart draws that layer above the lowest
hosting book. 2026-10-11: on this data the two fields separated again - the lowest hosting book sits
clear of the top layer - so the paragraph says every hosting book sits above both layers and the chart
draws both layers below the lowest hosting book. 2026-10-25: on this data the top layer sat above the
lowest hosting book for the first time. Three re-cuts in three releases told us the gap between the two
fields is smaller than the week-to-week movement of the measurement, so from this release the clause that
says how the two fields sit (apart, level within a point, or overlapping) is written from the data on
every release, and the chart labels and fades each bar by the kind of book it is rather than by its
position.Reproduce this — or ask it yourself
Every figure here is queryable through the HostingBrain connector. The aggregate loyalty distribution is
on the free tier via the customer_faithfulness tool; per-provider exclusivity scores are analyst
tier. In Claude or any MCP-compatible assistant, this is the whole brief in one prompt:
“Using HostingBrain, tell me how faithful hosting customers are: what share of owners with multiple domains keep them all with one provider, the average share they hold with their primary provider, and whether it changes for customers operating across multiple countries. Then rank providers by how many of their multi-domain customers use them exclusively, flag which are CDN/DNS layers where low exclusivity is expected — and check whether loyal customers actually sit within a single brand of the group, or are only loyal after ownership folding.”
Resolves to customer_faithfulness (distribution, free) and its per-provider leaderboard
(analyst). Ask definitions('faithfulness') for the bias control and folding.
Ask who's really loyal — and who's already shopping. HostingBrain answers inside Claude and any MCP-compatible assistant, with the denominator and caveats attached to every number.