Methodology: how this index is built and what it leaves out

A number appears on this site only if it resolves to a named source with a URL and the date that source was read. Nothing is estimated, rounded to look tidy, or carried forward from memory. A check runs before publication that fails if any figure references a source that does not exist, which is the mechanism rather than the intention.

The rule has a visible consequence: some fields are empty. An empty field means no source was found, which is more useful than a plausible number would be.

Firms are found through published comparisons and trade press, then their own sites are read for the markets they claim and the services they advertise. Presence in a market is taken from the firm's own page for that market, never from a directory listing or a third-party claim.

Some firms return an automated refusal to any request, which makes their pages unreadable by machine. Where that happened, the pages were located through the public search index instead and the citation remains the firm's own live URL. A refusal was never read as absence, because the same refusal is returned for pages that do not exist at all.

Firms in this category rarely publish prices, so most figures come from third-party reporting. Different publications describe different tiers, and the resulting ranges frequently do not overlap.

Where that happens every reported range is kept, attributed to whoever published it. Averaging them would produce a single confident number that no source supports, and the disagreement between published figures is itself the useful finding for anyone trying to budget.

Each market's calendar covers recurring fixtures that measurably change local availability. Dates come from organisers first, official tourism bodies second, and reference works third. One-off events are excluded, as is anything that does not move the market.

Every fixture carries two layers: the recurring season, which is stable, and the most recent confirmed edition, which is not. Where an organiser has not published a date, only the season appears. A calendar that quietly ages into a list of past events is worse than one that admits what it does not yet know.

Only firms that publish enough to be verified are indexed, which excludes the single-operator businesses that make up much of the supply in smaller markets. An empty firm count understates a market rather than describing it, and the market pages say so where it applies.

Fee coverage is thinner still, because publishing a price is unusual in this category. And the calendars cover the markets tracked here rather than every market that exists. None of these gaps is closed by guessing at what belongs in them.

The entry threshold for the firm index is publication: a firm appears where it publishes enough about itself to be checked. That rule keeps every entry verifiable and it systematically excludes a large part of real supply, because much of this market is single operators and small partnerships who publish almost nothing and take clients by referral. A market showing a low count here is a market with little published presence, which is not the same as a thin one.

Events are filtered the same way and for the same reason. The calendars cover recurring fixtures that measurably change local availability; one-off events are left out because a page built around them ages into a list of things that already happened, and anything below the availability threshold is left out because including it would make the lead-time figures meaningless.

Both exclusions are stated on the pages they affect rather than only here, since a count is easy to misread as a census.

Several numbers on this site are not read from a source but computed from the rows: median lead times, the count of markets whose calendar fits inside a four-month span, the crossover volume between paying per request and holding a retainer. Each is a stated operation on published inputs rather than an estimate, and each is described in the sentence that carries it so the arithmetic can be repeated.

Nothing is weighted, smoothed or adjusted. Where a derived figure would need an assumption to produce, it is not produced: there is no attempt, for example, to convert an hourly model into a per-request one, because no source publishes a conversion and any figure would be invented.

A check runs before publication that fails the build if any figure references a source that does not exist. That is a completeness test rather than a truth test, and it is the reason the rule about sourcing holds across several hundred figures rather than only where someone remembered.