Editorial policy

We do not invent first-hand experience

Nobody at TransferSelect has ridden these transfers, and no page here says otherwise. You will not find "we took the coach and it was busy", "our driver met us at arrivals" or a photograph presented as ours. Those sentences are easy to write, impossible for you to check, and they are the single most common way a comparison site becomes fiction.

Where this site tells you something is worth doing, it is derived from figures it can show you: a published tariff, a measured distance, an operator's own timetable. If the derivation cannot be shown, the claim is not made.

The source hierarchy

A source record states what kind of source it is, and that decides how much weight a figure taken from it carries. The strongest are the ones with an obligation behind them, in this order:

• official
• operator
• provider
• affiliate api
• affiliate feed
• published tariff

A regulator publishing a tariff it is bound by, an operator publishing its own fare, an airport describing its own ground transport: these are preferred over anything that merely reports them. The weakest, used only where nothing better exists and always labelled as such, are calculated, editorial estimate, secondary reference.

We do not take figures from competitors. An aggregator's price is a portrait of that aggregator, not of the route, and copying it would make this site a slower copy of theirs.

There are 66 source records behind the site today, each with its own URL and status. When a source moves or goes offline, the record says so rather than quietly continuing to be cited.

How a fact is verified

Provenance is stored per figure, not per page. A distance, a fare and a timetable on the same page were each checked on their own day, against their own source, and each carries that separately. A single page-level "last updated" date would let one small edit imply that everything on the page had just been confirmed.

Each figure records:

• the kind of source it came from
• which specific source record, by ID
• the date it was checked
• how confident we are in it
• how it was verified: by hand, from an API, or calculated
• how long it stays trustworthy before it should be checked again

That last one differs by kind. A road distance is stable for a year or more; a regulated tariff for a few months; a commercial price from a provider for weeks at most. A provider price that stops being refreshed disappears from the comparison on its own rather than hardening into a claim nobody is checking. There is more on price methodology.

Review before publishing, and the checks that do not rely on us

Editorial copy is written against the verified data rather than alongside it, so a paragraph cannot describe a coach the data says does not run. Where a figure is derived, the derivation is shown on the page and recomputed on every build, so the working and the number cannot disagree.

Judgement is not the only safeguard, because judgement is the thing that slips. The build refuses to publish a price with no stated type and basis, a figure that disagrees with the research it was recorded from, a page whose sitemap entry contradicts its own robots tag, or a commercial link missing its disclosure. Those are not warnings; they stop the build.

Search research, and why none of it appears on the site

Which routes get built is decided partly by search research: a keyword dataset of roughly seventy thousand queries, from which each route entity records a handful of internal metrics such as its primary keyword and an opportunity score.

Those metrics decide build order and appear on no page. They are estimates from a third-party tool, not measurements of the world, and printing "3,500 people search for this every month" would be dressing somebody else's model as a fact about the route. The build re-reads the dataset and fails if a recorded metric has drifted from it, because a number somebody remembered is worse than none: it survives, it looks sourced, and it quietly reorders what gets built next.

Corrections

Fares change, timetables change, and a page that was right in one season can be wrong in the next. If something here is wrong, tell us through contact and say which figure and which page.

A correction is treated as new research, not as an edit. We go back to the source that the figure was originally taken from, or find one if the original has moved, and the figure is rewritten with a new checked date and a new source reference. If a claim turns out to have had no adequate source behind it, it is removed rather than softened.

Independence

Commission cannot decide what this site recommends, and the mechanisms that make that true rather than merely stated are on how we make money. Two figures from today's build say it more plainly than a promise would. On 21 of the 81 routes that carry a bookable product, the option we can earn from is not the one the page names as cheapest. And 5 routes here have no bookable product at all and still carry a full comparison, because a route nobody pays us for is still a route somebody has to travel.

What we compare, and where a route deliberately has no overall winner, is on how we compare. What this site is and is not is on about.